阅读说明：本技术 基于视觉识别的室内地图基础数据生成方法 (Indoor map basic data generation method based on visual identification ) 是由 邵勤 孔维君 于 2019-10-11 设计创作，主要内容包括：基于视觉识别的室内地图基础数据生成方法,包括如下步骤：管理员通过管理端上传室内地图至服务器,并对室内地图进行数据归一化处理；AI服务器调用AI识别算法识别上传的室内地图中的道路和室内部件,对识别出的道路进行Mask掩模数据的生成；业务服务器将Mask掩模数据转换为点阵数据并储存；用户终端加载室内地图,加载道路Mask掩模数据,加载点位数据,加载室内部件数据；用户选择导航目的地；用户终端调用服务器中的应用层,应用层根据当前点位数据、目的地点位数据和道路数据通过Floyd路径算法计算最短路径；用户终端根据最短路径生成有向引导线,并将有向引导线与室内地图合并显示,完成室内地图数据的生成。(The indoor map basic data generation method based on visual identification comprises the following steps: the method comprises the following steps that a manager uploads an indoor map to a server through a management end, and data normalization processing is carried out on the indoor map; the AI server calls an AI recognition algorithm to recognize roads and indoor components in the uploaded indoor map, and Mask data is generated on the recognized roads; the business server converts Mask data into dot matrix data and stores the dot matrix data; loading an indoor map, road Mask data, point bit data and indoor component data by the user terminal; the user selects a navigation destination; the user terminal calls an application layer in the server, and the application layer calculates the shortest path through a Floyd path algorithm according to the current point bit data, the destination point bit data and the road data; and the user terminal generates a directional guide line according to the shortest path, and combines and displays the directional guide line and the indoor map to complete the generation of indoor map data.)

1. The indoor map basic data generation method based on visual identification is characterized by comprising the following steps:

the data generation method comprises the following steps:

step 1, a manager uploads an indoor map to a server through a management terminal and performs data normalization processing on the indoor map;

step 2, the server comprises an AI server, a service server and an application layer, the AI server calls an AI recognition algorithm to recognize roads and indoor components in the uploaded indoor map, the indoor components comprise walls, compartments and the like, Mask data is generated on the recognized roads, the indoor components generate intermediate components, non-access closed loop areas are formed in the road Mask generation process, and the areas are deleted in the Mask in a crossed mode, so that the road Mask only comprises the areas where the road accesses accurately avoid the indoor components; the business server converts Mask data into dot matrix data and stores the dot matrix data;

step 4, loading an indoor map, road Mask data, point bit data and indoor component data such as walls and compartments by the user terminal;

step 5, selecting a navigation destination by a user;

step 6, the user terminal calls an application layer in the server, and the application layer calculates the shortest path through a Floyd path algorithm according to the current point bit data, the destination point bit data and the road data;

and 7, generating a directional guide line by the user terminal according to the shortest path, and combining and displaying the directional guide line and the indoor map to finish the generation of indoor map data.

2. The vision recognition-based indoor map basic data generation method according to claim 1, characterized in that: the interaction between the user terminal and the server adopts a C/S client-server architecture; in the framework, the server background shares an AI server and a service server, and the user terminal forms an interface display by Android and interacts with a user.

3. The vision recognition-based indoor map basic data generation method according to claim 2, characterized in that: the server and an Android terminal communication protocol used as a user terminal select an Internet of things IoT lightweight protocol MQTT.

4. The vision recognition-based indoor map basic data generation method according to claim 1, characterized in that: the management end adopts a B/S browser-server framework, the server background in the framework shares an AI server and a service server, and the management end provides an interface through a browser and realizes operation.

5. The vision recognition-based indoor map basic data generation method according to claim 4, characterized in that: the management terminal provides a management operation interface based on the Web browser by using an Http and Http protocol.

6. The vision recognition-based indoor map basic data generation method according to claim 1, characterized in that: in step 2, the AI server returns the Mask data of the identified road in Json format and http protocol.

7. The vision recognition-based indoor map basic data generation method according to claim 1, characterized in that: the AI recognition algorithm is constructed on an AI server, AI recognition is carried out on objects in a map, OpenCV is constructed in a service cloud server, data extraction after the AI recognition algorithm is carried out, and a Floyd path algorithm is constructed on an application layer, and path planning based on the recognized road data is carried out.

8. The vision recognition-based indoor map basic data generation method according to claim 1, characterized in that: the AI server selects an Nvidia GPU GTX1060 and a calculation board with the calculation capacity of more than 6.0, and an AI recognition algorithm is trained in the AI server to ensure the recognition effect.

9. The vision recognition-based indoor map basic data generation method according to claim 1, characterized in that: the AI identification algorithm selects two VGG16 and Unet; the VGG16 is used for classifying objects in the images, the classification standard is two classifications, namely, road and indoor components, and a two-classification sorter with the sorting success rate larger than 95% is trained according to the road and indoor components in the training set; the Unet is used to generate Mask data for roads.

Technical Field

The invention belongs to the field of visual algorithms, and particularly relates to an indoor map basic data generation method based on visual identification.

Background

The method comprises the steps that in traditional work, complicated work of indoor map road basic point location data needing manual marking is carried out, two modes of marking are carried out in traditional work, a first mode user needs to upload pictures to an application service, then road data begin to be marked, after the basic road data are marked manually, the basic data need to be communicated, the data weight is calculated and marked, an indoor main channel is determined through the data weight, otherwise, the problem that a road access is abnormal and disconnected or the problem that the road access penetrates through the wall and the like can occur in the process of needing rich experience, the basic road marking needs to be very careful and intensive, and the problem that fixed-point navigation cannot be carried out can occur if the marking is sparse. The second way is to perform manual route planning on point-to-point navigation, which only realizes limited fixed navigation routes and cannot realize navigation between any two points, and the manual workload is multiplied with the increase of the point-to-point quantity.

Disclosure of Invention

The purpose of the patent is to provide a method for generating basic data of an indoor map based on visual identification, which is used for automatically identifying roads and indoor compartments on the indoor map, generating basic point location data required by a path planning algorithm, connecting user point locations and destination point locations through a shortest path algorithm, automatically generating a route guide, merging the route guide to an original guide map, and guiding a user route.

The indoor map basic data generation method based on visual identification comprises the following steps:

step 1, a manager uploads an indoor map to a server through a management terminal and performs data normalization processing on the indoor map;

step 2, the server comprises an AI server, a service server and an application layer, the AI server calls an AI recognition algorithm to recognize roads and indoor components in the uploaded indoor map, the indoor components comprise walls, compartments and the like, Mask data is generated on the recognized roads, the indoor components generate intermediate components, non-access closed loop areas are formed in the road Mask generation process, and the areas are deleted in the Mask in a crossed mode, so that the road Mask only comprises the areas where the road accesses accurately avoid the indoor components; the business server converts Mask data into dot matrix data and stores the dot matrix data;

step 4, loading an indoor map, road Mask data, point bit data and indoor component data such as walls and compartments by the user terminal;

step 5, selecting a navigation destination by a user;

step 6, the user terminal calls an application layer in the server, and the application layer calculates the shortest path through a Floyd path algorithm according to the current point bit data, the destination point bit data and the road data;

and 7, generating a directional guide line by the user terminal according to the shortest path, and combining and displaying the directional guide line and the indoor map to finish the generation of indoor map data.

Further, the interaction between the user terminal and the server adopts a C/S client-server architecture; in the framework, the server background shares an AI server and a service server, and the user terminal forms an interface display by Android and interacts with a user.

Further, an internet of things (IoT) lightweight protocol MQTT is selected as a communication protocol of the server and an Android terminal serving as the user terminal.

Furthermore, the management end adopts a B/S browser-server architecture, the server background in the architecture shares an AI server and a service server, and the management end provides an interface through the browser and realizes operation.

Further, the management terminal provides a management operation interface based on the Web browser by using an Http and Http protocol.

Further, in step 2, the AI server returns Mask data of the identified road in a format of Json and a protocol of http.

Furthermore, the AI recognition algorithm is constructed in an AI server, AI recognition is carried out on objects in the map, OpenCV is constructed in the service cloud server, data extraction after the AI recognition algorithm is carried out, and the Floyd path algorithm is constructed in an application layer, and path planning based on the recognized road data is carried out.

Further, the AI server selects an Nvidia GPU GTX1060 and a calculation board with the calculation capacity of more than 6.0, and the AI recognition algorithm is trained in the AI server to ensure the recognition effect.

Further, the AI identification algorithm selects two VGG16 and Unet; the VGG16 is used for classifying objects in the images, the classification standard is two classifications, namely, road and indoor components, and a two-classification sorter with the sorting success rate larger than 95% is trained according to the road and indoor components in the training set; the Unet is used to generate Mask data for roads.

The invention achieves the following beneficial effects: the invention solves the problem that a large amount of manual work is needed to participate in manually marking the key points of the path in the traditional indoor map path generation system, the path data generation process is fully automatic, the user can generate basic path point location data only by uploading the indoor map, the path algorithm can navigate any bidirectional guide path between any two points according to the basic data, and the use is convenient. The user only needs to upload an indoor map, the system automatically calculates the road access mask and generates basic data for navigation, the user does not need to invest in manual marking of detailed road points and manually planning navigation lines, and the user only needs to select a two-point system on the map to automatically calculate the access path and automatically navigate and plan the sight. The workload of manual marking is greatly liberated, and one-key full-automatic indoor navigation data generation is realized.

Drawings

Fig. 1 is a schematic structural diagram of a hardware system according to an embodiment of the present invention.

Fig. 2 is a flowchart of a management end in the embodiment of the present invention.

Fig. 3 is a schematic view of an interaction flow of a user terminal in an embodiment of the present invention.

Fig. 4 is a schematic view of a visualization process of the data generation method in the embodiment of the present invention.

Detailed Description

The technical scheme of the invention is further explained in detail by combining the drawings in the specification.

The indoor map basic data generation method based on visual identification comprises the following steps:

step 1, an administrator uploads an indoor map to a server through a management terminal and performs data normalization processing on the indoor map.

Step 2, the server comprises an AI server, a service server and an application layer, the AI server calls an AI recognition algorithm to recognize roads and indoor components in the uploaded indoor map, the indoor components comprise walls, compartments and the like, Mask data is generated on the recognized roads, the indoor components generate intermediate components, non-access closed loop areas are formed in the road Mask generation process, and the areas are deleted in the Mask in a crossed mode, so that the road Mask only comprises the areas where the road accesses accurately avoid the indoor components; the AI server returns Mask data of the identified road in a Json format and an http protocol; and the business server converts the Mask data into dot matrix data and stores the dot matrix data.

And 4, loading an indoor map, road Mask data, point bit data and indoor component data such as walls and compartments by the user terminal.

And 5, selecting a navigation destination by the user.

And 6, calling an application layer in the server by the user terminal, and calculating the shortest path by the application layer through a Floyd path algorithm according to the current point bit data, the destination point bit data and the road data.

And 7, generating a directional guide line by the user terminal according to the shortest path, and combining and displaying the directional guide line and the indoor map to finish the generation of indoor map data.

Referring to fig. 1, the hardware of the present invention includes a server, a firewall, a cloud service, a user terminal, and a management terminal. The management terminal and the user terminal are connected with the server through cloud service. The server comprises an AI server, a service server and an application layer, wherein the AI server is connected with the service server, the service server is connected with the application layer, and a firewall is arranged between the application layer and the cloud service.

The interaction between the user terminal and the server adopts a C/S client-server architecture; in the framework, the server background shares an AI server and a service server, and the user terminal forms an interface display by Android and interacts with a user. The server and an Android terminal communication protocol used as a user terminal select an Internet of things IoT lightweight protocol MQTT.

The management end adopts a B/S browser-server framework, the server background in the framework shares an AI server and a service server, and the management end provides an interface through a browser and realizes operation. The management terminal provides a management operation interface based on the Web browser by using an Http and Http protocol.

An AI recognition algorithm is constructed in an AI server, AI recognition is carried out on objects in a map, OpenCV is constructed in a service cloud server, data extraction after the AI recognition algorithm is carried out, and a Floyd path algorithm is constructed in an application layer, and path planning based on the recognized road data is carried out.

The AI server selects an Nvidia GPU GTX1060 and a calculation board with the calculation capacity of more than 6.0, and an AI recognition algorithm is trained in the AI server to ensure the recognition effect.

The AI identification algorithm selects two VGG16 and Unet; the VGG16 is used for classifying objects in the images, the classification standard is two classifications, namely, road and indoor components, and a two-classification sorter with the sorting success rate larger than 95% is trained according to the road and indoor components in the training set; the Unet is used to generate Mask data for roads.

The map data generation method provided by the invention has the visualization process shown in fig. 4, and comprises the steps of firstly processing through an AI (artificial intelligence) recognition algorithm, then synthesizing and storing data through OpenCV (open circuit vehicle), and finally calculating the shortest path through a Floyd path algorithm to generate a directed guide line, and merging and displaying the directed guide line and an indoor map to finish the generation of indoor map data.

The above description is only a preferred embodiment of the present invention, and the scope of the present invention is not limited to the above embodiment, but equivalent modifications or changes made by those skilled in the art according to the present disclosure should be included in the scope of the present invention as set forth in the appended claims.