阅读说明：本技术 一种油罐车底部装油口安全监控方法及系统 (Safety monitoring method and system for oil filling port at bottom of oil tank truck ) 是由 袁顺昌 孟证 葛凯 于 2021-04-08 设计创作，主要内容包括：本发明实施例提出了一种油罐车底部装油口安全监控方法及系统,其中方法包括：建立分析系统、识别系统和网络识别模型；启动所述分析系统对付油作业开始时间进行监测；付油作业开始时,所述分析系统发送识别指令至所述识别系统；所述识别系统接收指令,截取图像采集设备对鹤管对接过程采集的视频图像；所述网络识别模型接收图像采集设备监测时产生的图像数据,输出对象分类识别结果；根据所述对象分类识别结果,结合提油单信息,判断鹤管对接结果；可视化鹤管对接结果。本发明在不增加装车作业操作步骤的情况下,实现对鹤管对接的正确性地判断,不增加车辆成本,系统安装和维护也较简单。(The embodiment of the invention provides a method and a system for safely monitoring an oil loading port at the bottom of an oil tank truck, wherein the method comprises the following steps: establishing an analysis system, an identification system and a network identification model; starting the analysis system to monitor the starting time of the oil-applying operation; when the oil-loading operation is started, the analysis system sends a recognition instruction to the recognition system; the identification system receives the instruction and intercepts a video image acquired by the image acquisition equipment in the process of butt joint of the loading arm; the network identification model receives image data generated during monitoring of the image acquisition equipment and outputs an object classification identification result; judging a loading arm butt joint result according to the object classification and identification result and by combining oil extraction list information; and visualizing the butt joint result of the crane pipe. The invention realizes the judgment of the accuracy of the butt joint of the oil filling riser without increasing the operation steps of loading operation, does not increase the cost of vehicles, and has simpler system installation and maintenance.)

1. A method for monitoring the safety of an oil loading port at the bottom of an oil tank truck is characterized by comprising the following steps:

establishing an analysis system, an identification system and a network identification model;

starting the analysis system to monitor the starting time of the oil-applying operation;

when the oil-loading operation is started, the analysis system sends a recognition instruction to the recognition system;

the identification system receives the instruction and intercepts a video image acquired by the image acquisition equipment in the process of butt joint of the loading arm;

the network identification model receives image data generated during monitoring of the image acquisition equipment and outputs an object classification identification result;

judging a loading arm butt joint result according to the object classification and identification result and by combining oil extraction list information;

and visualizing the butt joint result of the crane pipe.

2. The method for safely monitoring the oil loading port at the bottom of the oil tank truck as claimed in claim 1,

the analysis system enters a ready state by running the identification registration file on the analysis server;

the recognition system enters a ready state on the recognition server by inputting a start command at the terminal.

3. The method for safely monitoring the oil loading port at the bottom of the oil tank truck as claimed in claim 1,

and the recognition system receives the recognition command sent by the analysis system and enters a response state.

4. The method for safely monitoring the oil loading port at the bottom of the oil tank truck as claimed in claim 1,

performing recognition classification on the video images by using the trained network recognition model;

the training process of the network recognition model comprises the following steps:

acquiring a sample data set;

preprocessing the sample data set;

receiving a preprocessed sample data set;

extracting characteristic information of a video image;

converting, transferring and fusing the characteristic scale, and further extracting the characteristic;

feature parameter simplification is performed by omitting half of the feature detectors;

and identifying the nonlinear object class of the object in the sample data set by adopting an activation function.

5. The method for safely monitoring the oil loading port at the bottom of the oil tank truck as claimed in claim 4, wherein,

sample data in the sample library is formed by acquiring an original image, augmenting data, marking the data, dividing a data set and modifying training parameters;

the preprocessing is further to mark the canning interface at the bottom of the oil tank truck with concentrated sample data.

6. The method for safely monitoring the oil loading port at the bottom of the oil tank truck as claimed in claim 1,

in the safety monitoring process, when the operation of the recognition server and the analysis server is abnormal, an alarm prompt is given.

7. The method for safely monitoring the oil loading port at the bottom of the oil tank truck as claimed in claim 1,

and data query and problem tracing are performed by recording data information generated in the safety monitoring process.

8. A safety monitoring system for an oil filling port at the bottom of an oil tank truck is used for realizing the method of any one of claims 1 to 7, and is characterized by comprising the following steps:

an analysis system configured to: monitoring an oil distribution system, triggering an identification system and analyzing the butt joint condition of a loading arm by combining oil extraction list information;

the identification system comprises a network identification model and information acquisition equipment;

a hair oil system configured to: recording card swiping information of a user, and receiving a final butt joint and error state of the loading arm;

a database configured to: recording data information generated in the operation process and being used as a basis for problem tracing;

a data transmission module configured to: and transmitting data among all the modules of the system according to requirements.

9. The system for monitoring the safety of the oil loading port at the bottom of the oil tank truck as claimed in claim 8,

the information acquisition device is configured to: the data transmission module is connected with the data acquisition module, acquires real-time image data and transmits the acquired image data to the analysis system; the network identification model is set to receive image data obtained by the information acquisition equipment and analyze the butt joint state of the oil filling riser in the image data.

10. A computer readable storage medium, characterized in that the computer readable storage medium stores computer program instructions, and the computer program instructions when executed by a processor implement the method for monitoring the safety of the bottom oil loading port of the oil tank truck according to any one of claims 1 to 7.

Technical Field

The invention relates to a method and a system for monitoring the safety of a bottom oil filling port of an oil tank truck, in particular to the field of image data processing.

Background

In the highway oil supply operation of the oil depot, because the multi-bin oil tank truck is provided with a plurality of canning interfaces, especially in the bottom loading mode which is widely used at present, the bottom canning interfaces are concentrated together, and when an operator connects a lower loading crane pipe to the bottom canning interfaces, the interfaces are easy to be mistakenly connected once negligence happens, so that the delivery quantity errors can be caused, and serious accidents such as oil mixing, roof collapse, fire hazard, explosion and the like can be easily caused.

In the prior art, in order to deal with the risks, a sensor signal comparison mode is adopted, and the accuracy of the butt joint of the loading arm is automatically judged. This requires the installation of corresponding electronic sensing devices on the vehicle and on the loading arm, and requires special attention by personnel during operation to ensure that the sensors are in place and that the detection signals are transmitted to the controller by wireless means. However, this method requires installation of a plurality of devices, is cumbersome to implement and maintain, and involves a cost increase due to the fact that the sensors on the vehicle are easily lost and damaged; meanwhile, signal transmission passes through a plurality of links, resulting in low reliability.

Disclosure of Invention

The purpose of the invention is as follows: a method and a system for monitoring the safety of an oil loading port at the bottom of an oil tank truck are provided to solve the problems in the prior art.

The technical scheme is as follows: in a first aspect, a method for monitoring the safety of an oil loading port at the bottom of an oil tank truck is provided, which is characterized by comprising the following steps:

firstly, establishing an analysis system, an identification system and a network identification model;

step two, starting the analysis system to monitor the starting time of the oil-applying operation;

step three, when the oil-loading operation is started, the analysis system sends an identification instruction to the identification system;

step four, the identification system receives the instruction and intercepts a video image acquired by the image acquisition equipment in the process of butt joint of the oil filling riser;

the network identification model receives image data generated during monitoring of the image acquisition equipment and outputs an object classification identification result;

step six, judging a loading arm butt joint result according to the object classification and identification result and by combining oil extraction list information;

and seventhly, visualizing the butt joint result of the crane pipe.

In some implementations of the first aspect, the analytics system enters the ready state by running an identifying registry file on an analytics server;

the recognition system enters a ready state on the recognition server by inputting a start command at the terminal.

In some implementations of the first aspect, the recognition system receives a recognition command sent by the analysis system and enters a response state.

In some implementations of the first aspect, the video images are classified by recognition using the trained network recognition model;

the training process of the network recognition model comprises the following steps:

acquiring a sample data set;

preprocessing the sample data set;

receiving a preprocessed sample data set;

extracting characteristic information of a video image;

converting, transferring and fusing the characteristic scale, and further extracting the characteristic;

feature parameter simplification is performed by omitting half of the feature detectors;

and identifying the nonlinear object class of the object in the sample data set by adopting an activation function.

In some realizations of the first aspect, the sample data in the sample library is formed by original image acquisition, data augmentation, data tagging, data set partitioning, and training parameter modification;

the preprocessing is further to mark the canning interface at the bottom of the oil tank truck with concentrated sample data.

In some implementation manners of the first aspect, in the security monitoring process, when the operation of the recognition server and the analysis server is abnormal, an alarm prompt is performed.

In some implementations of the first aspect, data information generated in the security monitoring process is recorded for data query and problem tracing.

In a second aspect, a system for monitoring the safety of an oil loading port at the bottom of an oil tank truck is provided, which specifically comprises:

an analysis system configured to: monitoring an oil distribution system, triggering an identification system and analyzing the butt joint condition of a loading arm by combining oil extraction list information;

the identification system comprises a network identification model and information acquisition equipment;

a hair oil system configured to: recording card swiping information of a user, and receiving a final butt joint and error state of the loading arm;

a database configured to: recording data information generated in the operation process and being used as a basis for problem tracing;

a data transmission module configured to: and transmitting data among all the modules of the system according to requirements.

In some realizations of the second aspect, the information-gathering device is configured to: the data transmission module is connected with the data acquisition module, acquires real-time image data and transmits the acquired image data to the analysis system; the network identification model is set to receive image data obtained by the information acquisition equipment and analyze the butt joint state of the oil filling riser in the image data.

In a third aspect, a computer readable storage medium is provided, wherein computer program instructions are stored on the computer readable storage medium, and when the computer program instructions are executed by a processor, the computer program instructions implement a method for monitoring the safety of a bottom oil filling port of a tank truck.

Has the advantages that: compared with the prior art which adopts a sensor signal comparison mode, the method and the system for safely monitoring the oil filling port at the bottom of the oil tank truck intelligently recognize the actual butt joint condition of the oil filling riser and the bottom filling port in real time by utilizing a video image recognition technology and adopting an AI deep learning algorithm, and judge the butt joint accuracy under the conditions that the operation steps of loading operation are not increased and special attention is not needed during operation. In addition, the original loading operation flow is not influenced, the vehicle cost is not increased, and the system installation and maintenance are simpler.

Wherein, there is a record in every identification process, can inquire afterwards to carry out the problem and trace back. In addition, the monitoring system has a self-checking function, whether the system server works normally or not is monitored and identified at any time, and once abnormity occurs, alarm prompt is carried out.

Aiming at a camera video stream, the identification algorithm provided by the invention can feed back an identification result within 1 second, the identification rate of the idle bottom canned interface and the crane pipe number with the shielding part smaller than 1/5 can reach 95%, the camera placing position is good, and the identification rate of the clear image without shielding can reach 100%.

Drawings

FIG. 1 is a flow chart of data processing according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating training sample generation according to an embodiment of the present invention.

Detailed Description

According to the safety monitoring method and system for the oil filling port at the bottom of the oil tank truck, the actual butt joint condition of the oil filling riser and the bottom filling port is intelligently identified in real time, and the butt joint correctness is judged, so that the loading safety is guaranteed. The present invention will be further described in detail with reference to the following examples and accompanying drawings.

In one embodiment, a method for monitoring safety of an oil loading port at the bottom of an oil tank truck is provided, which is used for automatically judging the butt joint correctness of a loading arm in the highway oil loading operation of an oil depot, and as shown in fig. 1, the method specifically comprises the following steps:

firstly, establishing an analysis system, an identification system and a network identification model;

step two, starting the analysis system to monitor the starting time of the oil-applying operation;

step three, when the oil-loading operation is started, the analysis system sends an identification instruction to the identification system;

step four, the identification system receives the instruction and intercepts a video image acquired by the image acquisition equipment in the process of butt joint of the oil filling riser;

the network identification model receives image data generated during monitoring of the image acquisition equipment and outputs an object classification identification result;

step six, judging a loading arm butt joint result according to the object classification and identification result and by combining oil extraction list information;

and seventhly, the oil delivery system receives the butt joint result of the loading arm.

In a further embodiment, the analysis system is first run on the analysis server, the recognition system is run on the recognition server, and the function of monitoring the pay-per-view job in real time is activated when the analysis system is ready. When the oil supply operation is started, the analysis system monitors the cargo space card swiping information received by the oil distribution system and immediately sends a corresponding identification command to the identification system. After the recognition command is received by the recognition system, the video image collected by the camera is intercepted, the oil filling port recognition is started, the recognition result is returned to the analysis system, the analysis system combines the recognition result with the oil extraction list information for comprehensive analysis, the conclusion whether the oil filling port is correctly butted is obtained, and the conclusion is sent to the oil sending system.

Specifically, the operating system of the recognition server is ubuntu18.04lts, and only one application system of the recognition system is operated in the recognition server, and the dependent support environment required for operation is GPU driver, Anaconda, Python, tenserflow, Keras, Pytorch, and OpenCV. The operating system version of the analysis Server is at least Windows Server2008, and the analysis system and other application systems, such as a data platform, a cargo space card reading, database management and the like, are operated on the Server, and the operation needs to depend on support environments such as Microsoft network Framework 4.0 and Microsoft SQLServer 2008. The operating environment presented in the examples herein may be adapted to different code implementation platforms, as this is not a further limitation of the present invention.

In a further embodiment, identifying the server configuration comprises:

CPU (Intel to strong 4110), memory (32G DDR 4), hard disk (1 TB), two display cards (RTX 2080 Ti), network port (1 GE power port) and power supply (1000W).

Identifying the server configuration includes:

CPU (Intel to strong 4110), memory (8G DDR 4), hard disk (1 TB), display card (GTX 1060), network interface (1 GE power port) and power supply (650W).

The analysis system acquires the cargo space card swiping information in the oil delivery system, and can generate an identification command and send the identification command to the identification system by combining parameters stored in the database to identify the specified cargo space, so that the oil delivery system and the SQLServer database management system are in normal operation states. Exe file is double-clicked on the analysis server, and after the operation is successful, the analysis system is ready.

The step of starting the identification system is that firstly, a command terminal is opened; secondly, inputting a corresponding instruction code; and finally, completing input through an enter key, thereby successfully starting the identification system. Wherein, inputting corresponding instruction code is specifically

conda activate tf

cd PycharmProjects/detect

python interface_ detectio.py

After the identification system is started successfully, the identification system starts to wait for the identification command sent by the analysis system and responses, and relevant information is displayed in the terminal without manual operation.

In a further embodiment, the bottom canning interfaces of the oil tank truck in the obtained loading video image are marked, then the YOLO convolution model is used for training, a special convolution neural network, namely a network identification model, is finally formed, each bottom canning interface on the oil tank truck can be automatically identified in real time by utilizing the neural network, whether the oil tank truck is connected with a loading arm or not is identified, which loading arm is connected, whether the loading arm is connected correctly or not can be judged by combining oil extraction list information, and finally a judgment result is output to an oil delivery system.

In a further embodiment, in order to improve the accuracy of the object classification recognition by the network recognition model, the object classification recognition is trained.

Specifically, the sample data used in the training process is sample data stored in the sample library at a previous time. The sample data adopted by training forms a sample library through original image acquisition, data augmentation, data marking, data set division and training parameter modification.

The original image obtaining process includes that video images of the surveillance video are captured and stored according to specified frame numbers. In the preferred embodiment, the specified frame number is 16, the target video file and the video frame number are read, meanwhile, the variable of the frame number is divided by the specified frame number, whether the multiple of the specified frame is reached is judged, if the multiple of the specified frame is reached, a picture of the current frame is stored, and the operation is performed in a circulating mode until the video is completely read.

The data augmentation process is to remove invalid pictures, rename the data-cleaned files according to the number 1,2,3 …, and then perform augmentation for one time; subsequently, the second amplification is performed on the basis of the first amplification. In the preferred embodiment, one augmentation uses horizontal and mirror flipping to expand the data, as shown in FIG. 2. On the basis of the primary amplification, the secondary amplification adopts methods of mean value disturbance, convolution, sharpening, edge detection in a specific direction, Gaussian noise addition, Gaussian disturbance, contrast change, radiation transformation, pixel-by-pixel addition and the like to expand data again.

The data labels the port and loading arm numbers, in the preferred embodiment, labelImg labels the monitoring targets, the labeling is shown in table 1 below.

TABLE 1 classification of monitoring targets

Data set partitioning divides the acquired data into a training set and a validation set. In the preferred embodiment, the training set and the verification set are divided according to the proportion of 0.91:1, and then the annotation data file is converted into the yolo format.

In the preferred embodiment, the total sample size of algorithm training is 38551 images, covers the numbers 0, 1 and 2 of the protective cover, the lower oil filling port, the multiple oil gas recovery ports and the oil filling riser, the samples are diversified fully, the sample library is established reasonably, and the number and the types of the samples are covered sufficiently.

In a further embodiment, the network recognition model uses the YOLOv3 algorithm, YOLO is at speed, processing speed can reach 45fps, and its fast version (smaller network) can even reach 155 fps. Compared with the previous generation series, the YOLOv3 not only makes a 'compromise' design, balances the monitoring speed and the accuracy rate, but also increases the layer number of the neural network. The network design combining identification and positioning is adopted, so that training and prediction can be carried out end to end, namely, the whole monitoring sample is input, after the characteristics of the middle layer are extracted, and finally, the coordinate information and the category information of the prediction box are regressed in the logics of the output layer. In the preferred embodiment, Darknet19 is changed into Darknet53, so that the object features are extracted more deeply, and the accuracy is further improved.

Specifically, because the analysis system needs to feed back the identification result quickly after the driver swipes the card, usually not more than 1 second, in addition, the situation that a plurality of goods places swipes the card simultaneously often appears, which means that the system needs to have the capability of identifying and processing a plurality of images simultaneously, and when the system identifies the images, the system not only needs to detect which types of objects are in the images simultaneously, such as an oil filling port, an oil filling riser and the like, but also needs to identify the specific positions of the objects in the images, and the calculation amount is large. According to the method, the speed is improved under the condition that the accuracy is guaranteed, the network design integrating recognition and positioning is realized through the YOLOv3 algorithm, the object features are extracted more deeply, and the accuracy is further improved.

Due to objective factors such as the parking position of the oil tank truck, the angle of a camera and the like, monitored targets are often close and mutually shielded, and errors such as missing detection, false detection and the like can be caused due to poor detail discrimination of the YOLOv3 network, so that the algorithm is improved aiming at the conditions. In the preferred embodiment, multi-scale prediction is added, bounding boxes with different sizes are provided, and the implementation process is similar to that of a feature image adopted by YOLOv2 for multi-scale projection on training data samples, and the bounding boxes are set.

In order to better divide the feature information extracted from the image and reduce the training time and the classification error rate, a relu function is adopted as an activation function in the embodiment. The phenomenon of overfitting is easy to occur under the conditions that the model parameters are excessive and the training samples are insufficient, soOptimizing the model by adopting dropout, and reducing overfitting and error rate and improving classification precision by neglecting half of the feature detectors; meanwhile, regularization constraints are added for improving network performance. In the preferred embodiment, the input received by the neuron in the neural network is used as an input signal, the activation value passing through the activation function is used as an output signal, and a constant false alarm rate is set at the same time, so that when the distribution of input data and output results is not uniform, the threshold value can be adaptively adjusted, and the classification effect is improved. Wherein, the relu function is defined asInput isHas a weight ofThe output value is expressed by the following expression:

in the formula (I), the compound is shown in the specification,representing an activation function;representing weightsThe conjugate transpose of (1);representing input data;a correction term is represented. Thus, the activation values of the nodes are:

in the formula (I), the compound is shown in the specification,representing an activation function;representing weightsThe conjugate transpose of (1);representing input data;representing a correction term; and M represents a binary mask matrix, obeys the value of Bernoulli distribution, and is set according to the activation value. Specifically, according to a set threshold, when the activation value of a node is greater than the threshold, the value of an element in the M binary mask matrix is 1, and when the activation value of the node is less than the threshold, the value is 0. And selecting a numerical value from the set formed by the node activation values as a threshold value for judging the interest degree, and deleting the nodes lower than the threshold value through judgment so as to realize optimization. Wherein, the deleted node set is the set obtained by subtracting the reserved node from all the node sets, that is:

in the formula (I), the compound is shown in the specification,indicating a deleted node;represents all nodes;represents the retention ratio;representing nodes above a threshold;indicating nodes below a threshold;representing the ratio of reserved nodes.

On the other hand, under the condition that the sample data is insufficient, in order to improve the expression performance of extracting the features of the convolutional neural network, the feature scale is converted, migrated and fused under the condition that the original data of the retained features are not damaged. In the preferred embodiment, the feature graphs are scaled down, migrated and fused in the backbone network, so as to reuse the features.

In order to solve the problem of single sample image, the image is expanded by combining sample classification in the aspect of processing the network camera video stream, and the unbalance degree of the sample is analyzed. Specifically, in the expansion process, when the size and the rotation angle of the image are converted, the corresponding rotation angle, the scaling factor and the contrast factor are respectively in the interval、[Andare randomly generated.

And aiming at the problem that the IOU value of the YOLOv3 model is not ideal, clustering optimization is carried out on the anchor value. Firstly, obtaining a plurality of classes by using a DBSCAN clustering algorithm under the condition of neglecting a central point; secondly, in order to obtain local and overall information of the detection target, clustering is carried out by increasing multiple scales, and further multi-scale clustering is carried out on the outline information of the detection target; thirdly, a primary characteristic diagram can be obtained through multi-scale clustering and convolution operation; and finally, the classes are used as input data, and partition clustering is carried out through a K-means algorithm, so that the accurate position of the central point can be obtained. The DBSCAN clustering algorithm can obtain a good clustering effect for the first time under the condition of neglecting a central point, and the method combined with the K-means algorithm can effectively accelerate data set convergence and improve the classification accuracy of small targets.

In a further embodiment, the identification process performs identification of the actual image of the cargo space by using a trained network identification model. Specifically, the successfully trained network recognition model is placed in a recognition server, and the anchor value is updated at the same time. During identification, the actual goods space image to be identified is introduced into a network model, a series of candidate objects can be predicted by a network, and redundant candidate objects are removed by using a non-maximum suppression algorithm, so that the problem that one target is detected for multiple times is solved. In the preferred embodiment, the non-maximum suppression algorithm uses 0.5 as the threshold. And finally, the remaining candidate objects are recognition results, and the categories and the coordinates of the candidate objects are marked in the image and then output to finish the recognition process. The identified result comprises bottom canning interfaces, oil gas recovery interfaces and oil filling riser numbers of all non-filled oil filling risers in the image. By adopting the image identification method, aiming at one camera video stream, the identification result can be fed back by an identification algorithm within 1 second, the identification rate of the idle bottom canned interface and the crane pipe serial number with the shielding part smaller than 1/5 can reach 95%, the camera is placed at a good position, and the identification rate of the clear and non-shielding image can reach 100%.

In a further embodiment, when the butt joint result of the oil filling riser is judged, the identification object is analyzed and corrected by further combining the data information of the oil extraction list on the basis of the output result of the network identification model, so that the final butt joint relation between the oil filling riser and the oil filling port is obtained.

In a further embodiment, a UDP transport protocol is used between the fueling system, the identification system, and the analysis system to facilitate data exchange between the systems. And receiving an identification command of the oil distribution system in an integrated mode with the oil distribution system, and finishing the functions of immediately intercepting the video image of the specified camera to carry out canning interface identification, sending an identification result, correcting the system time, feeding back the running state of the system and the like.

In a further embodiment, when the system monitors that the server is abnormal, an alarm prompt is given.

In a further embodiment, the data information generated at each step in the identification process is recorded for information query, and problem tracing.

In a further embodiment, the equipment configuration is carried out according to the oil depot situation, and in a preferred embodiment, the number of the cameras installed on each island, the related network parameters and the like are configured according to the number of oil sending islands.

In one embodiment, a system for monitoring the safety of an oil filling port at the bottom of an oil tank truck is provided, which is used for realizing a method for monitoring the safety of the oil filling port at the bottom of the oil tank truck, and the system specifically comprises:

an analysis system configured to: monitoring the oil distribution system, triggering the recognition system and analyzing the butt joint condition of the oil filling riser by combining the information of the oil extraction list.

The identification system comprises a network identification model and information acquisition equipment; the information acquisition device is configured to: and the data transmission module is connected with the data acquisition module, acquires real-time image data and transmits the acquired image data to the analysis system. The network identification model is set to receive image data obtained by the information acquisition equipment and analyze the butt joint state of the oil filling riser in the image data.

A hair oil system configured to: and recording card swiping information of a user, and receiving the final butt joint and wrong state of the loading arm.

A database configured to: and recording data information generated in the operation process to be used as a basis for tracing the problems.

A data transmission module configured to: and transmitting data among all the modules of the system according to requirements.

In a further embodiment, the information collecting device is a camera, and in a preferred embodiment, the configuration of the camera is as shown in table 2 below.

TABLE 2 configuration information of the cameras

In one embodiment, a computer readable storage medium having computer program instructions stored thereon, the computer program instructions when executed by a processor implement a tank wagon bottom fill port safety monitoring method.

As noted above, while the present invention has been shown and described with reference to certain preferred embodiments, it is not to be construed as limited thereto. Various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.