阅读说明：本技术 基于ai驱动的高分遥感森火风险普查以及预警防控方法 (AI (Artificial intelligence) -drive-based high-resolution remote sensing fire risk general investigation and early warning prevention and control method ) 是由 章豪 温剑锋 徐刚 朱安峰 徐海燕 苏宗奎 范全龙 郑志鹏 程文 刘鸿鹄 李洪兵 于 2021-05-21 设计创作，主要内容包括：本发明涉及基于AI驱动的高分遥感森火风险普查以及预警防控方法,基于高分遥感影像以及人工智能技术识别影像,该方法包括林区的可燃物致灾要素识别、建筑物承载要素识别、道路与水体减灾要素识别以及森林火灾监测预警传感网系统；实现实时、远程监测监控,降低人员现场巡查监测的安全风险,降低森林火灾预防成本,可在森林火灾发生时有效掌握相关信息,并进行决策指挥辅助,提升森林火灾应急反应速度和处理能力,降低因救灾慢救灾难造成的社会经济财产损失。(The invention relates to a high-resolution remote sensing forest fire risk general investigation and early warning prevention and control method based on AI drive, which is based on a high-resolution remote sensing image and an artificial intelligence technology identification image and comprises combustible disaster causing element identification, building bearing element identification, road and water disaster reduction element identification and a forest fire monitoring and early warning sensor network system; the system realizes real-time and remote monitoring, reduces the safety risk of personnel on-site patrol monitoring, reduces the cost for preventing forest fires, can effectively master related information when forest fires occur, carries out decision command assistance, improves the emergency response speed and the processing capacity of forest fires, and reduces the social, economic and property loss caused by slow disaster relief.)

1. AI-drive-based high-resolution remote sensing fire risk general investigation and early warning prevention and control method is characterized by comprising the following steps: the risk general investigation and early warning prevention and control method is based on high-resolution remote sensing images and artificial intelligence technology identification images, and comprises combustible disaster-causing element identification, building bearing element identification, road and water disaster reduction element identification and forest fire monitoring and early warning sensor network systems in forest areas; the combustible disaster element identification is based on forest combustible identification combined with an attention mechanism, forest combustible types and distribution ranges thereof are collected, combustible remote sensing information extraction and special analysis are carried out, combustible type loading capacity distribution data are obtained, and a combustible type chart is generated; the building bearing element identification is based on forest area building identification considering ground feature multi-scale property, buildings and infrastructure in a forest area are used as main disaster bearing bodies for extraction and investigation, the space distribution and disaster attribute characteristics of the disaster bearing bodies are obtained, and a disaster bearing body investigation result geographic information database is established; the identification of the road and water disaster reduction elements comprises forest region road water body identification based on topological connectivity, forest region road water body distribution data are obtained, a forest fire comprehensive disaster reduction resource database is established, and a corresponding forest region road water body disaster reduction body chart is generated; the forest fire monitoring and early warning sensing network system comprises a plurality of internet of things sensors and an online early warning prevention and control platform, the internet of things sensors are all flexibly thrown into a forest area based on an unmanned aerial vehicle, and Beidou modules and N-R communication modules are arranged in the internet of things sensors and are in AD-HOC network communication based on an AD-HOC networking technology; the online early warning prevention and control platform is provided with a forest fire prevention monitoring function and is in signal connection with the Internet of things sensor.

2. The AI-driven-based high-resolution remote sensing fire risk general survey and early warning prevention and control method according to claim 1, wherein: the forest combustible recognition combined with the attention mechanism comprises a channel attention mechanism and a space attention mechanism, a visual system information processing mechanism is called by simulating human eyeball movement based on an artificial intelligence model, attention, rejection and combustible object selection are carried out, a group of real-time weight coefficients are generated, and the real-time weight coefficients are endowed to different combustible element areas.

3. The AI-driven-based high-resolution remote sensing fire risk general survey and early warning prevention and control method according to claim 1, wherein: the forest region building identification considering the multi-scale property of the ground features comprises the steps of establishing a dynamic multi-scale feature module, wherein the dynamic multi-scale feature module comprises a plurality of Dynamic Convolution Modules (DCMs) which are arranged in parallel, and the dynamic convolution modules learn semantic feature expressions of specific scales by utilizing different context-aware convolution kernels; based on the parallel branches formed among the dynamic convolution modules with different perception ranges, the characteristics of different branches are fused, and the high-effectiveness self-adaptive multi-scale characteristic expression is obtained.

4. The AI-driven-based high-resolution remote sensing fire risk general survey and early warning prevention and control method according to claim 1, wherein: the forest road water body identification based on topological connectivity comprises the steps of establishing a boundary and a topological perception module, wherein the topological perception module sequentially uses convolution operation and nonlinear activation on rows and columns of an element characteristic diagram from four different directions, models spatial context information and establishes complete topological connection.

5. The AI-driven-based high-resolution remote sensing fire risk general survey and early warning prevention and control method according to claim 1, wherein: the forest fire prevention monitoring function is provided with a disaster relief unmanned aerial vehicle and an amphibious disaster relief ship, and comprises the following steps of 1) acquiring forest area environment data acquired by an Internet of things sensor, analyzing, calculating and establishing a mathematical model; 2) setting an early warning threshold value of the environment-related parameter; 3) acquiring smoke concentration and temperature information acquired by an internet of things sensor, analyzing whether a fire disaster occurs or not by combining a multi-terminal cooperative edge algorithm, and alarming; 4) analyzing the situation and the spreading condition of the fire disaster based on an edge calculation algorithm, and acquiring the geographical position information of the fire disaster through a Beidou module in the sensor of the Internet of things; 5) starting the unmanned aerial vehicle and the land and water disaster relief ship to perform on-site exploration and inspection according to the geographical position information; 6) after the fire alarm is verified, the disaster relief unmanned aerial vehicle and the land and water disaster relief ship are operated and controlled to conduct fire extinguishing treatment.

6. The AI-driven-based high-resolution remote sensing fire risk general survey and early warning prevention and control method according to claim 5, wherein: the amphibious disaster relief ship comprises a ship body (1), a controller, a water gun (2), a pipe coiling disc (3), a sensor group and a digital wireless network bridge, wherein the ship body (1) is provided with an engine (4), a movable wheel (5) and a movable water gun frame (6), the engine (4) is fixedly arranged at the tail part of the ship body (1), and the engine (4) is connected with a propulsion propeller (7) for matching and linkage; the moving wheel (5) is connected and arranged at the bottom of the ship body (1), and the moving wheel (5) is connected and provided with a driving motor fixed in the ship body (1) and is matched and linked; the movable water gun frame (6) and the pipe coiling disc (3) are fixedly arranged at the top of the ship body (1), and the movable water gun frame (6) is provided with a fixed buckle adaptive to the water gun (2); the hose coiling disc (3) is connected with a water delivery hose (8) and a hose coiling motor (9), and the hose coiling motor (9) is fixedly connected with a rotating shaft of the hose coiling disc (3) to be matched and linked; one end of the water delivery hose (8) is fixedly connected with the water gun (2), the other end of the water delivery hose (8) is connected with a water delivery pump (10), and a water inlet valve is arranged in the water delivery pump (10); the sensor group comprises a liquid level sensor and a plurality of laser sensors, the liquid level sensor is fixedly arranged at the bottom of the ship body (1), and the laser sensors are respectively and fixedly arranged at two sides of the ship body (1); the digital wireless network bridge is fixedly arranged in the ship body (1), and is in signal connection with the controller, the sensor group, the engine (4), the driving motor, the pipe coiling motor (9) and the water delivery pump (10).

7. The AI-driven-based high-resolution remote sensing fire risk general survey and early warning prevention and control method according to claim 6, wherein: the bow part of the ship body (1) is also provided with a guide plate (11), the ship body (1) is also provided with a pipe discharging chute (12) which is matched with the position of the pipe coiling disc (3), and the pipe discharging chute (12) is communicated with the side surface of the ship body (1); the bottom end of the guide plate (11) inclines towards one side where the ship body (1) is located, and a supporting spring (13) is connected between the guide plate (11) and the ship body (1); the water gun (2) comprises a direct-flow water gun and a spraying water gun, and one end of the water delivery hose (8) provided with the water delivery pump (10) is also connected with a positioning anchor (14).

Technical Field

The invention belongs to the technical field of forest fire prevention and control, and particularly relates to a high-resolution remote sensing forest fire risk general investigation and early warning prevention and control method based on AI drive.

Background

The forest fire disaster involvement range is wide, the forest fire disaster cannot be found in time, the processing difficulty is high, the harmfulness is serious, and the forest fire disaster involvement range is one of eight natural disasters in the world; the method can burn a large amount of forest resources, is a main disturbance factor of the forest ecological environment, and directly destroys the natural environment and ecological balance; is also a potential threat source of animals and plants, and directly influences the industrial and agricultural production and even threatens the safety of people's lives and properties. The traditional forest fire risk monitoring method mainly obtains the state information of a non-key area through group measurement and group defense, obtains the state information of a key area through a conventional monitoring means, consumes a large amount of manpower, and is low in timeliness of data acquisition. Therefore, the AI-driven high-resolution remote sensing forest fire risk general survey and early warning prevention and control method which can be developed by combining the artificial intelligence technology to develop remote sensing application research, establish and optimize related remote sensing models and assist in completing forest fire risk general survey becomes a technical problem which needs to be solved urgently.

Disclosure of Invention

The invention provides a high-resolution remote sensing fire risk general investigation and early warning prevention and control method based on AI drive for solving the problems.

The technical scheme of the invention is that the method is based on AI-driven high-resolution remote sensing forest fire risk general survey and early warning prevention and control method, the method is based on high-resolution remote sensing image and artificial intelligence technology identification image, the method comprises combustible disaster-causing element identification, building bearing element identification, road and water disaster reduction element identification and forest fire monitoring and early warning sensing network system in forest; the combustible disaster element identification is based on forest combustible identification combined with an attention mechanism, forest combustible types and distribution ranges thereof are collected, combustible remote sensing information extraction and special analysis are carried out, combustible type loading capacity distribution data are obtained, and a combustible type chart is generated; the building bearing element identification is based on forest area building identification considering ground feature multi-scale property, buildings and infrastructure in a forest area are used as main disaster bearing bodies for extraction and investigation, the space distribution and disaster attribute characteristics of the disaster bearing bodies are obtained, and a disaster bearing body investigation result geographic information database is established; the identification of the road and water disaster reduction elements comprises forest region road water body identification based on topological connectivity, forest region road water body distribution data are obtained, a forest fire comprehensive disaster reduction resource database is established, and a corresponding forest region road water body disaster reduction body chart is generated; the forest fire monitoring and early warning sensing network system comprises a plurality of internet of things sensors and an online early warning prevention and control platform, the internet of things sensors are all flexibly thrown into a forest area based on an unmanned aerial vehicle, and Beidou modules and N-R communication modules are arranged in the internet of things sensors and are in AD-HOC network communication based on an AD-HOC networking technology; the online early warning prevention and control platform is provided with a forest fire prevention monitoring function and is in signal connection with the Internet of things sensor.

After the method is adopted, the high-resolution remote sensing image and the artificial intelligence technology are used for identifying the image, so that the high-efficiency and accurate identification of disaster-related elements of forest fire risk investigation and fire disaster prevention and control is realized, the disaster hidden danger area is visualized, and key technical support is provided for forest fire hidden danger early warning and post-disaster treatment; the combustible substance type chart is generated through combustible substance disaster-causing element recognition, so that accurate forest combustible substance types and distribution ranges thereof can be mastered conveniently, data and technical support are provided for the pre-disaster warning and prediction process of forest fires, and normalized monitoring and emergency response of the forest fires are realized; by extracting and investigating disaster-bearing body spatial distribution and disaster attribute characteristics and establishing a database, people's lives and properties which are possibly affected by forest fires can be conveniently and timely emergently transferred and safely evacuated, and basic data and scientific decision basis are provided for ecological civilization construction work; by acquiring forest region road water body distribution data and generating a forest region road water body disaster reduction body chart, targeted and accurate release of manpower and material resources during disaster relief is ensured, the maximum utilization of disaster relief resources is realized, manpower cost and public resources are saved, and personnel and property loss caused by forest fires is effectively reduced; the online early warning and prevention and control platform is connected with the sensor signal of the Internet of things, the integrated sensor group collects various data in the forest area in real time, and the online early warning and prevention and control platform carries out early warning on abnormal data, so that the abnormal conditions of the forest area can be found in time; the Internet of things sensors are flexibly put in the unmanned aerial vehicle, so that the construction cost is reduced, the deployment time is shortened, and the operation and maintenance pressure is relieved; the positioning data of the sensor of the Internet of things is acquired through the Beidou module, and the environment and the positioning data are sent to the online early warning prevention and control platform through the N-R communication module, so that the forest environment monitoring and the automatic detection of fire hidden dangers are realized; through the AD-HOC networking technology, AD HOC networking among devices is realized, a mobile communication network among sensors of the Internet of things is quickly constructed, and the adaptability to the complex forest area environment is improved; the system realizes real-time and remote monitoring, reduces the safety risk of personnel on-site patrol monitoring, reduces the cost for preventing forest fires, can effectively master related information when forest fires occur, carries out decision command assistance, improves the emergency response speed and the processing capacity of forest fires, and reduces the social, economic and property loss caused by slow disaster relief.

As a further improvement of the invention, the forest combustible recognition combined with the attention mechanism comprises a channel attention mechanism and a space attention mechanism, a visual system information processing mechanism is called by simulating human eyeball movement based on an artificial intelligence model, attention, rejection and combustible object selection are carried out, a group of real-time weight coefficients are generated, and the real-time weight coefficients are endowed to different combustible element areas.

After the method is adopted, the correlation among different characteristic channels can be fully considered through a channel attention mechanism, the spatial dependence among different positions can be considered through a spatial attention mechanism, the artificial intelligence model is combined to simulate the movement of human eyes to process visual system information, the combustible and non-combustible regions are distinguished, and the forest combustible region with the risk is focused; meanwhile, attention is paid to different combustible material types, the problems of foreign matter co-spectrum and boundary fuzzy of different combustible materials are relieved, and the combustible material types in forest regions are efficiently and accurately identified.

As a further improvement of the present invention, the forest zone building identification considering the multi-scale property of the ground objects comprises establishing a dynamic multi-scale feature module, wherein the dynamic multi-scale feature module comprises a plurality of Dynamic Convolution Modules (DCMs) arranged in parallel, and the dynamic convolution modules use different context-aware convolution kernels to learn the semantic feature expression of a specific scale; based on the parallel branches formed among the dynamic convolution modules with different perception ranges, the characteristics of different branches are fused, and the high-effectiveness self-adaptive multi-scale characteristic expression is obtained.

By adopting the method, the parallel branches are formed among the dynamic convolution modules, the characteristics of different branches are fused, and the building characteristic expressions of different scales are dynamically selected, so that the corresponding characteristic expressions are conveniently extracted aiming at the building ground objects of different scales when a mathematical model is established, the accurate self-adaptive learning of the building characteristics is realized, and the identification precision and effect of the building are improved.

As a further improvement of the invention, the forest road water body identification based on topological connectivity comprises establishing a boundary and a topological perception module, wherein the topological perception module sequentially uses convolution operation and nonlinear activation on rows and columns of the element characteristic diagram from four different directions, models spatial context information and establishes complete topological connectivity.

By adopting the method, a complete topological communication structure is constructed by capturing the feature expressions of the ground features in different directions, the effective connectivity of the road and the water body is kept, the complete and effective road and water body information is conveniently extracted, the efficiency of planning and disaster relief work is improved, and the life and property loss is reduced.

As a further improvement of the invention, the forest fire prevention monitoring function is provided with a disaster relief unmanned aerial vehicle and an amphibious disaster relief ship, and the method comprises the following steps of 1) acquiring forest area environment data acquired by an internet of things sensor, analyzing, calculating and establishing a mathematical model; 2) setting an early warning threshold value of the environment-related parameter; 3) acquiring smoke concentration and temperature information acquired by an internet of things sensor, analyzing whether a fire disaster occurs or not by combining a multi-terminal cooperative edge algorithm, and alarming; 4) analyzing the situation and the spreading condition of the fire disaster based on an edge calculation algorithm, and acquiring the geographical position information of the fire disaster through a Beidou module in the sensor of the Internet of things; 5) starting the unmanned aerial vehicle and the land and water disaster relief ship to perform on-site exploration and inspection according to the geographical position information; 6) after the fire alarm is verified, the disaster relief unmanned aerial vehicle and the land and water disaster relief ship are operated and controlled to conduct fire extinguishing treatment.

By adopting the method, the environmental data of the forest area can be acquired in real time through the system, so that the system can trigger early warning in time according to actual conditions, and the forest fire prevention monitoring efficiency is improved; disaster situation geographical position information is acquired through the Beidou module, so that the disaster relief unmanned aerial vehicle and the land and water disaster relief ship can be operated to perform timely fire extinguishing treatment according to specific longitude and latitude information, and the expansion of fire accidents is prevented.

As a further improvement of the invention, the amphibious disaster relief ship comprises a ship body, a controller, a water gun, a pipe coiling disc, a sensor group and a digital wireless network bridge, wherein the ship body is provided with an engine, a movable wheel and a movable water gun frame, the engine is fixedly arranged at the tail part of the ship body, and the engine is connected with a propulsion propeller to perform matching linkage; the moving wheels are connected and arranged at the bottom of the ship body and are connected and provided with driving motors fixed in the ship body and are matched and linked; the movable water gun frame and the pipe coiling disc are fixedly arranged at the top of the ship body, and the movable water gun frame is provided with a fixed buckle adaptive to the water gun; the pipe coiling disc is connected with a water delivery hose and a pipe coiling motor, and the pipe coiling motor is fixedly connected with a rotating shaft of the pipe coiling disc for matching linkage; one end of the water delivery hose is fixedly connected with the water gun, the other end of the water delivery hose is connected with the water delivery pump, and the water delivery pump is internally provided with a water inlet valve; the sensor group comprises a liquid level sensor and a plurality of laser sensors, the liquid level sensor is fixedly arranged at the bottom of the ship body, and the laser sensors are respectively and fixedly arranged at two sides of the ship body; the digital wireless network bridge is fixedly arranged in the ship body and is in signal connection with the controller, the sensor group, the engine, the driving motor, the pipe coiling motor and the water delivery pump.

After the method is adopted, a signal is sent to the digital wireless network bridge through the controller to control the ship body and other parts on the ship body, the propeller is driven to rotate through the engine to push the ship body to move along a river channel, when a fire disaster occurs, the pipe coiling disc is driven to rotate by starting the pipe coiling motor, one end of the water delivery hose connected with the water delivery water pump is placed into a water body, the water inlet valve is opened, the water delivery water pump is started, water is delivered to the water gun through the water delivery hose, the water gun is clamped in the fixed buckle, the movable water gun frame is adjusted to enable the water spray port of the water gun to face a fire source, and the water gun sprays water to the fire source; when the fire source is far away from the water body, the height difference between the water surface of the water body and the shore is detected through the laser sensors on the two sides of the ship body, the ship body is rushed to land when the proper position of the height difference is detected, meanwhile, the front end of the ship body is detected through the liquid level sensor at the bottom of the ship body to be away from the water body, the driving motor is controlled to be started to drive the movable wheel to rotate, the ship body is driven to be close to a fire starting point through the movable wheel, then the water gun sprays water to extinguish the fire source, and the water gun can be taken out of the fixed buckle and is operated by being held by the hand; the fire fighting ship can move and patrol along the water body and the river channel in daily life, can move to a nearby water body in time in case of fire, draws water from the water body to irrigate a fire extinguishing source, can find a position suitable for the height difference of the water surface of the water body to flush the land in case of fire, and can drive the fire fighting ship to be close to the fire fighting point through the moving wheel and lay a water conveying hose along the way when facing the fire fighting point to be far away from the water source, so that the water can be conveyed to the fire fighting point in the shortest time, and the disaster reduction efficiency of a forest area is improved.

As a further improvement of the invention, the bow part of the ship body is also provided with a guide plate, the ship body is also provided with a pipe-releasing chute which is adaptive to the position of the pipe-rolling disc, and the pipe-releasing chute is communicated with the side surface of the ship body; the bottom end of the guide plate inclines towards one side where the ship body is located, and a supporting spring is connected between the guide plate and the ship body; the water gun comprises a straight-flow water gun and a spraying water gun, and one end of the water delivery hose, which is provided with the water delivery pump, is also connected with a positioning anchor.

After the method is adopted, the guide plate is arranged at the bow part of the ship body, the bottom of the guide plate is inclined towards the ship body, when the ship body is ashed, the guide plate is firstly contacted with the shore to lift the bow upwards, and a moving wheel at the bottom of the ship body is convenient to contact the land; the pipe placing chute communicated with the side surface of the ship body is arranged, so that the water delivery hose can be guided into water when the pipe coiling disc rotates to release the water delivery hose, and the water delivery hose is prevented from being accumulated on the ship body; the positioning anchor is arranged at the end, provided with the water delivery pump, of the water delivery hose, so that the water delivery hose can be fixed by the positioning anchor when water is drawn in a river channel, and the water delivery hose is prevented from being taken out of a water body when a relief ship is ashore, and the influence on the water drawing of the water delivery pump is avoided.

Drawings

Fig. 1 is a schematic structural view of an amphibious disaster relief ship of the invention.

Fig. 2 is a schematic structural view of the support bump of the flattening apparatus.

The device comprises a ship body, a water gun, a 3-pipe coiling disc, a 4-engine, a 5-moving wheel, a 6-movable water gun frame, a 7-propelling propeller, an 8-water delivery hose, a 9-pipe coiling motor, a 10-water delivery pump, an 11-guide plate, a 12-pipe placing chute, a 13-supporting spring and a 14-positioning anchor.

Detailed Description

The AI-driven high-resolution remote sensing forest fire risk general survey and early warning prevention and control method is based on high-resolution remote sensing images and artificial intelligence technology recognition images and comprises combustible disaster-causing element recognition, building bearing element recognition, road and water disaster reduction element recognition and forest fire monitoring and early warning sensor network systems in forest areas as shown in figures 1-2; the combustible disaster element identification is based on forest combustible identification combined with an attention mechanism, forest combustible types and distribution ranges thereof are collected, combustible remote sensing information extraction and special analysis are carried out, combustible type loading capacity distribution data are obtained, and a combustible type chart is generated; the building bearing element identification is based on forest area building identification considering ground feature multi-scale property, buildings and infrastructure in a forest area are used as main disaster bearing bodies for extraction and investigation, the space distribution and disaster attribute characteristics of the disaster bearing bodies are obtained, and a disaster bearing body investigation result geographic information database is established; the identification of the road and water disaster reduction elements comprises forest region road water body identification based on topological connectivity, forest region road water body distribution data are obtained, a forest fire comprehensive disaster reduction resource database is established, and a corresponding forest region road water body disaster reduction body chart is generated; the forest fire monitoring and early warning sensing network system comprises a plurality of internet of things sensors and an online early warning prevention and control platform, the internet of things sensors are all flexibly thrown into a forest area based on an unmanned aerial vehicle, and Beidou modules and N-R communication modules are arranged in the internet of things sensors and are in AD-HOC network communication based on an AD-HOC networking technology; the online early warning prevention and control platform is provided with a forest fire prevention monitoring function and is in signal connection with the Internet of things sensor.

The high-resolution remote sensing image and the artificial intelligence technology identification image are used for realizing efficient and accurate identification of disaster-related elements of forest fire risk investigation and fire disaster prevention and control, visually presenting a disaster hidden danger area and providing key technical support for forest fire hidden danger early warning and post-disaster treatment; the combustible substance type chart is generated through combustible substance disaster-causing element recognition, so that accurate forest combustible substance types and distribution ranges thereof can be mastered conveniently, data and technical support are provided for the pre-disaster warning and prediction process of forest fires, and normalized monitoring and emergency response of the forest fires are realized; by extracting and investigating disaster-bearing body spatial distribution and disaster attribute characteristics and establishing a database, people's lives and properties which are possibly affected by forest fires can be conveniently and timely emergently transferred and safely evacuated, and basic data and scientific decision basis are provided for ecological civilization construction work; by acquiring forest region road water body distribution data and generating a forest region road water body disaster reduction body chart, targeted and accurate release of manpower and material resources during disaster relief is ensured, the maximum utilization of disaster relief resources is realized, manpower cost and public resources are saved, and personnel and property loss caused by forest fires is effectively reduced; the online early warning and prevention and control platform is connected with the sensor signal of the Internet of things, the integrated sensor group collects various data in the forest area in real time, and the online early warning and prevention and control platform carries out early warning on abnormal data, so that the abnormal conditions of the forest area can be found in time; the Internet of things sensors are flexibly put in the unmanned aerial vehicle, so that the construction cost is reduced, the deployment time is shortened, and the operation and maintenance pressure is relieved; the positioning data of the sensor of the Internet of things is acquired through the Beidou module, and the environment and the positioning data are sent to the online early warning prevention and control platform through the N-R communication module, so that the forest environment monitoring and the automatic detection of fire hidden dangers are realized; through the AD-HOC networking technology, AD HOC networking among devices is realized, a mobile communication network among sensors of the Internet of things is quickly constructed, and the adaptability to the complex forest area environment is improved; the system realizes real-time and remote monitoring, reduces the safety risk of personnel on-site patrol monitoring, reduces the cost for preventing forest fires, can effectively master related information when forest fires occur, carries out decision command assistance, improves the emergency response speed and the processing capacity of forest fires, and reduces the social, economic and property loss caused by slow disaster relief.

The forest combustible recognition combined with the attention mechanism comprises a channel attention mechanism and a space attention mechanism, a visual system information processing mechanism is called by simulating human eyeball movement based on an artificial intelligence model, attention, rejection and combustible object selection are carried out, a group of real-time weight coefficients are generated, and the real-time weight coefficients are endowed to different combustible element areas.

The correlation among different characteristic channels can be fully considered through a channel attention mechanism, the spatial dependence relation among different positions can be considered through the spatial attention mechanism, the artificial intelligence model is combined to simulate the movement of human eyes to process information of a visual system, combustible and non-combustible regions are distinguished, and the forest combustible region with the risk is focused; meanwhile, attention is paid to different combustible material types, the problems of foreign matter co-spectrum and boundary fuzzy of different combustible materials are relieved, and the combustible material types in forest regions are efficiently and accurately identified.

The forest region building identification considering the multi-scale property of the ground features comprises the steps of establishing a dynamic multi-scale feature module, wherein the dynamic multi-scale feature module comprises a plurality of Dynamic Convolution Modules (DCMs) which are arranged in parallel, and the dynamic convolution modules learn semantic feature expressions of specific scales by utilizing different context-aware convolution kernels; based on the parallel branches formed among the dynamic convolution modules with different perception ranges, the characteristics of different branches are fused, and the high-effectiveness self-adaptive multi-scale characteristic expression is obtained.

Parallel branches are formed among the dynamic convolution modules, the features of different branches are fused, building feature expressions of different scales are dynamically selected, so that corresponding feature expressions are conveniently extracted for building ground objects of different scales when a mathematical model is established, accurate self-adaptive learning of building features is realized, and the identification precision and effect of buildings are improved.

The forest road water body identification based on topological connectivity comprises the steps of establishing a boundary and a topological perception module, wherein the topological perception module sequentially uses convolution operation and nonlinear activation on rows and columns of an element characteristic diagram from four different directions, models spatial context information and establishes complete topological connection.

The complete topological communication structure is constructed by capturing the feature expressions of the ground features in different directions, the effective connectivity of the road and the water body is maintained, the complete and effective road and water body information is conveniently extracted, the efficiency of planning disaster relief work is improved, and the life and property loss is reduced.

The forest fire prevention monitoring function is provided with a disaster relief unmanned aerial vehicle and an amphibious disaster relief ship, and comprises the following steps of 1) acquiring forest area environment data acquired by an Internet of things sensor, analyzing, calculating and establishing a mathematical model; 2) setting an early warning threshold value of the environment-related parameter; 3) acquiring smoke concentration and temperature information acquired by an internet of things sensor, analyzing whether a fire disaster occurs or not by combining a multi-terminal cooperative edge algorithm, and alarming; 4) analyzing the situation and the spreading condition of the fire disaster based on an edge calculation algorithm, and acquiring the geographical position information of the fire disaster through a Beidou module in the sensor of the Internet of things; 5) starting the unmanned aerial vehicle and the land and water disaster relief ship to perform on-site exploration and inspection according to the geographical position information; 6) after the fire alarm is verified, the disaster relief unmanned aerial vehicle and the land and water disaster relief ship are operated and controlled to conduct fire extinguishing treatment.

The system can acquire the environmental data of the forest area in real time, so that the system can trigger early warning in time according to actual conditions, and the forest fire prevention monitoring efficiency is improved; disaster situation geographical position information is acquired through the Beidou module, so that the disaster relief unmanned aerial vehicle and the land and water disaster relief ship can be operated to perform timely fire extinguishing treatment according to specific longitude and latitude information, and the expansion of fire accidents is prevented.

The amphibious disaster relief ship comprises a ship body 1, a controller, a water gun 2, a pipe coiling disc 3, a sensor group and a digital wireless network bridge, wherein the ship body 1 is provided with an engine 4, a movable wheel 5 and a movable water gun frame 6, the engine 4 is fixedly arranged at the tail part of the ship body 1, and the engine 4 is connected with a propulsion propeller 7 for matching and linkage; the moving wheels 5 are connected and arranged at the bottom of the ship body 1, and the moving wheels 5 are connected and provided with driving motors fixed in the ship body 1 and are matched and linked; the movable water gun frame 6 and the pipe coiling disc 3 are fixedly arranged at the top of the ship body 1, and the movable water gun frame 6 is provided with a fixed buckle adaptive to the water gun 2; the hose coiling disc 3 is connected with a water delivery hose 8 and a hose coiling motor 9, and the hose coiling motor 9 is fixedly connected with a rotating shaft of the hose coiling disc 3 for matching linkage; one end of the water delivery hose 8 is fixedly connected with the water gun 2, the other end of the water delivery hose 8 is connected with a water delivery pump 10, and a water inlet valve is arranged in the water delivery pump 10; the sensor group comprises a liquid level sensor and a plurality of laser sensors, the liquid level sensor is fixedly arranged at the bottom of the ship body 1, and the laser sensors are respectively and fixedly arranged at two sides of the ship body 1; the digital wireless network bridge is fixedly arranged in the ship body 1 and is in signal connection with the controller, the sensor group, the engine 4, the driving motor, the pipe coiling motor 9 and the water delivery pump 10.

The digital wireless network bridge is used for sending signals to a controller to control the ship body 1 and other components on the ship body 1, an engine 4 is used for driving a propulsion propeller 7 to rotate, the ship body 1 is pushed to move along a river channel, when a fire disaster occurs, a pipe coiling motor 9 is started to drive a pipe coiling disc 3 to rotate, one end of a water delivery hose 8 connected with a water delivery water pump 10 is placed into a water body, a water inlet valve is opened, the water delivery water pump 10 is started, water is delivered to a water gun 2 through the water delivery hose 8, the water gun 2 is clamped in a fixed buckle, a movable water gun frame 6 is adjusted to enable a water spraying port of the water gun 2 to face a fire source, and the water gun 2 sprays water to the fire source; when the fire source is far away from the water body, the height difference between the water surface of the water body and the shore is detected through the laser sensors on the two sides of the ship body 1, the ship body is rushed to the land when the proper position of the height difference is detected, meanwhile, when the front end of the ship body 1 is detected to be away from the water body through the liquid level sensor at the bottom of the ship body 1, the driving motor is controlled to be started to drive the moving wheel 5 to rotate, the ship body 1 is driven to be close to a fire point through the moving wheel 5, then the water gun 2 sprays water to water the fire source, the water gun 2 can be taken out of the fixed buckle, and the operation is carried out by holding; the daily life-saving ship can move and patrol along the water body and the river channel, can move to the nearby water body in time in case of fire, draws water from the water body to irrigate a fire extinguishing source, can find a position with a proper height difference with the water surface of the water body to flush the land in case of a fire point far away from a water source through the laser sensor, then is driven to be close to the fire point through the movable wheel 5, and lays the water delivery hose 8 along the way, so that water can be conveniently transported to the fire point in the shortest time, and the forest area disaster reduction efficiency is improved.

The bow part of the ship body 1 is also provided with a guide plate 11, the ship body 1 is also provided with a pipe placing chute 12 which is matched with the position of the pipe coiling disc 3, and the pipe placing chute 12 is communicated with the side surface of the ship body 1; the bottom end of the guide plate 11 inclines towards one side where the ship body 1 is located, and a supporting spring 13 is connected between the guide plate 11 and the ship body 1; the water gun 2 comprises a straight-flow water gun and a spraying water gun, and a positioning anchor 14 is further connected to one end, provided with a water delivery pump 10, of the water delivery hose 8.

The guide plate 11 is arranged at the bow part of the ship body 1, the bottom of the guide plate 11 is inclined towards the ship body 1, when the ship body 1 is ashed, the guide plate 11 is firstly contacted with the shore to lift the bow upwards, and the moving wheel 5 at the bottom of the ship body 1 is convenient to contact the land; the pipe-placing chute 12 communicated with the side surface of the ship body 1 is arranged, so that the water hose 8 can be guided into water when the pipe-coiling disc 3 rotates to release the water hose 8, and the water hose 8 is prevented from being accumulated on the ship body 1; the positioning anchor 14 is arranged at one end, provided with the water delivery pump 10, of the water delivery hose 8, so that the water delivery hose 8 can be fixed by the positioning anchor 14 when water is drawn in a river channel, and the water delivery hose 8 is prevented from being taken out of a water body when a relief ship is ashore, and the water drawing of the water delivery pump 10 is influenced.