Grain depot grain transport vehicle operation supervision method
By binding the IC card and the grain truck track ID in the grain warehouse, and combining video surveillance data and IC card reporting information, the operation process of the grain truck is supervised in real time, and the loopholes in the grain truck supervision are solved, achieving the correctness and stability Comprehensive supervision and abnormal detection of grain truck operations have improved the intelligent management level of grain warehouses.
Patent Information
- Application Number
- CN202410938607.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-14
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-07-14
AI Technical Summary
The existing technology has loopholes in the supervision of grain trucks in the grain warehouse. IC cards or labels are prone to falling off or disassembled by unauthorized persons, resulting in the operation process of grain trucks not being effectively supervised, and the reliability and maintenance costs of electronic tags are high. .
By binding the card number of the IC card with the grain truck track ID, the grain truck track data is obtained in real time, the rest point and rest time interval are extracted, and the grain warehouse operation points are matched, and the operation time is compared using the IC card to report the data. 进行异常检测和报警提示,确保运粮车作业过程的监管。
Achieve one-to-one correspondence between IC cards and grain transport trucks, supervise abnormal behaviors during the operation of grain transport trucks, improve the systematic and intelligent management of grain warehouses, and prevent "changing cars and cards"-type cheating behaviors , strengthened the information management of smart grain warehouse construction.
Smart Images

Figure CN119048545A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of intelligent monitoring of grain transport vehicles, and in particular relates to a method for supervising the operation of grain transport vehicles in a grain depot. Background Art
[0002] At present, my country's policy grain management has achieved full coverage of informatization, ensuring that information such as purchase, sales, and rotation can be queried and traced. Similarly, grain transport vehicles that transport grain in and out of grain depots also need to be effectively tracked and managed. At present, when grain depots temporarily manage grain transport vehicles, most of them issue IC cards or attach and stick disposable labels. The problem with using IC cards or labels for temporary management of grain transport vehicles is that the correspondence between IC cards or labels and grain transport vehicles cannot be guaranteed. When grain transport vehicles perform business operations, disposable labels are prone to falling off and loss, as well as unauthorized human disassembly, movement, and damage. The use of IC cards is prone to cheating by unauthorized human exchange, and there are loopholes in the supervision of grain depot entry and exit business.
[0003] The Chinese invention patent with patent number CN201610849841.9 proposes an electronic tag that can be adsorbed on the object to be tested, such as vehicles, turnover boxes and freight containers, through the permanent magnet inside it. It is used to store electronic information such as vehicles, personnel, and goods, and actively detect the association status between the object to be tested and the tag. However, it requires a power module and network support. The power module causes its single service life to be short, reliability to be reduced, and maintenance costs to increase. The network support causes it to not work properly in places with poor networks. IC cards have the advantages of long service life, reusability, small size and easy to carry, high reliability, and strong confidentiality. Therefore, this electronic tag cannot better replace IC cards.
[0004] The Chinese invention patent with patent number CN202010233879.X proposes a method for compressing video data of grain depots based on target detection and trajectory analysis. The real-time video surveillance data in the grain depot is collected by a surveillance camera, and the video surveillance data obtained by the camera is analyzed by a target detection algorithm to extract valid frames with moving targets and the location and sub-graph information of the target; a convolutional neural network model is used to perform target recognition operations on the target sub-graph to obtain the category information of the target; target matching and tracking are performed based on the target information collected from multiple consecutive frames of pictures collected from the same camera to obtain the motion trajectory data of the target in the current scene. Based on the information collected by each surveillance camera, a storage architecture for image data, label data and trajectory data is constructed to achieve efficient data storage. The method of the invention can extract and store valid data related to moving targets from massive surveillance video data, thereby improving the utilization rate of storage resources. However, the invention focuses on the compressed storage of surveillance video data, and does not further utilize the extracted moving target trajectory data.
[0005] The Chinese invention patent with patent number CN202210344357.6 proposes a vehicle detection method, system and storage medium. The vehicle detection method includes: obtaining video surveillance information in the grain storage area; judging whether there is a moving vehicle in the video surveillance information, and if so, judging the type of the moving vehicle; in the case where the moving vehicle is a grain transport vehicle, judging whether the grain transport vehicle is a heavy vehicle; if so, obtaining the movement trajectory of the heavy vehicle, and judging whether the heavy vehicle is entering or leaving the warehouse. In this way, the time of manual measurement is saved and labor costs are reduced. However, the invention only judges whether the grain transport vehicle is a heavy vehicle and whether the grain transport vehicle is entering or leaving the warehouse, and the extracted grain transport vehicle trajectory is not fully utilized. The provided vehicle detection system has little practical application significance in the supervision of grain transport vehicle operations in grain storage. Summary of the invention
[0006] The technical problem to be solved by the present invention is to provide a method for supervising the operation of a grain transport vehicle in a grain depot, so as to supervise the whole operation process of the grain transport vehicle in the grain depot.
[0007] In order to solve the above technical problems, the present invention provides a method for supervising the operation of a grain transport vehicle in a grain depot, comprising the following steps:
[0008] S100, binding the card number of the IC card with the track ID of the grain transport vehicle;
[0009] S200, obtaining the grain transport vehicle trajectory in real time from the surveillance video data in the grain storage area;
[0010] S300, extracting the stationary point and stationary time interval of the grain transport vehicle from the trajectory of the grain transport vehicle;
[0011] S400, matching the stationary point with the grain depot operation point area, and taking the stationary time interval of the matched stationary point as the first reference value of the operation time;
[0012] S500, taking the time point when the grain transport vehicle at the stationary point completes the operation as the second reference value of the operation time, and if the difference between the time interval of the second reference value of the operation time and the time interval of the first reference value of the operation time exceeds a preset time difference threshold, issuing an alarm prompt;
[0013] S600, grouping the grain transport vehicle trajectory data and performing grain transport vehicle trajectory anomaly detection and static point anomaly detection, and issuing an alarm prompt when an abnormality is detected.
[0014] As an improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0015] The process of obtaining the grain transport vehicle trajectory is as follows:
[0016] The surveillance video stream data of the same surveillance camera is input into the target detection algorithm model, and the vehicles in the surveillance video stream data are detected to obtain the grain truck target detection frame and confidence. The grain truck target detection frame is passed through the multi-target tracking algorithm model to obtain the grain truck trajectory data under a single camera; then the cross-camera trajectory association algorithm is used to obtain the associated grain truck trajectory data.
[0017] As a further improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0018] The cross-camera trajectory association algorithm is specifically:
[0019] The grain truck trajectory data under a single camera is converted into relative coordinates in the grain warehouse scene. Then, the trajectory data of the same grain truck under multiple single cameras is traversed based on the time sequence.
[0020] (1) When there are overlapped coordinate points in different trajectory data at the same time, the overlapping multiple coordinate points are merged into one coordinate point;
[0021] (2) When there are inconsistent coordinate points in different trajectory data at the same time, the coordinate points with high confidence are spliced and the coordinate points with low confidence are discarded;
[0022] (3) In addition to the above two cases, the coordinate points are directly spliced to obtain the associated grain truck trajectory data.
[0023] As a further improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0024] The target detection algorithm model is the YOLOv8 convolutional neural network, and the multi-target tracking algorithm model is the BoT-SORT algorithm.
[0025] As a further improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0026] The stationary point and stationary time interval of the grain transport vehicle are obtained from the associated trajectory of the grain transport vehicle through the Traj-DBSCAN algorithm.
[0027] As a further improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0028] The process of obtaining the first reference value of the operation time is as follows:
[0029] The grain depot operation point location area includes the tare operation point, gross weight measurement operation point, pre-inspection sampling operation point and grain loading and unloading operation point area. The static point is compared with the location area of each grain depot operation point in turn:
[0030] (1) When the stationary point is a single coordinate point, when the minimum Euclidean distance between the stationary point coordinates and the boundary of the grain depot operation point location area is less than the preset distance threshold, the stationary time interval of the current operation point is used as the first reference value of the operation time, and the grain depot operation point location area matching the stationary point is recorded;
[0031] (2) When the stationary point is a coordinate point cluster, when the coordinates of the stationary point exceeding the ratio threshold are in the grain depot operation point location area, the stationary time interval of the current operation point is used as the first reference value of the operation time, and the grain depot operation point location area matching the stationary point is recorded;
[0032] (3) If no grain storage operation point matches the static point, repeat steps S200, S300 and S400 until the first reference value of the operation time is obtained.
[0033] As a further improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0034] The time point at which the grain transport vehicle at the stationary point completes the operation is when the grain transport vehicle uses the IC card to report the operation data when performing the operation, including the operation type and the time point at which the operation is completed.
[0035] As a further improvement of the method for supervising the operation of a grain transport vehicle in a grain depot of the present invention:
[0036] The process of trajectory anomaly detection and stationary point anomaly detection is as follows:
[0037] The grain depot map is gridded, and the relative coordinates of the associated grain truck trajectory and the grain truck's stationary point in the grain depot scene are converted into grid coordinates; then the grain truck trajectory and stationary point converted into network coordinates are grouped according to the business type performed by the grain truck and the granary number, and then the grouped grain truck trajectory and stationary point are respectively subjected to anomaly detection using an isolation-based trajectory anomaly detection algorithm.
[0038] The beneficial effects of the present invention are mainly reflected in:
[0039] 1. The present invention can ensure a one-to-one correspondence between the IC card and the grain transport vehicle by binding the IC card with the grain transport vehicle track when issuing the IC card;
[0040] 2. The present invention extracts the stationary time interval of the grain transport vehicle in the operation area through the grain transport vehicle trajectory, reports the operation time of the grain transport vehicle in each operation area through the IC card, and monitors the corresponding relationship between the IC card and the grain transport vehicle by comparing the operation time and the stationary time interval of the grain transport vehicle in real time, ensuring that the entire operation process of the grain transport vehicle is supervised, thereby forming supervision on cheating behaviors of the "changing vehicle and changing card" type during the operation of the grain transport vehicle;
[0041] 3. The present invention can effectively and fully utilize the trajectory data of the grain transport vehicle through abnormal detection of the grain transport vehicle trajectory and the abnormal detection of the grain transport vehicle stationary point, and supervise the abnormal situations such as detours and abnormal stationary situations of the grain transport vehicle during the operation process, thereby realizing systematic, intelligent and information-based management of grain depots, which is conducive to the construction of smart grain depots. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings.
[0043] Figure 1 The present invention is a flow chart of a method for supervising grain transport vehicles in a grain depot.
[0044] Figure 2 It is a flow chart for obtaining track data of a grain transport vehicle in real time in a method for supervising grain transport vehicles in a grain depot of the present invention. DETAILED DESCRIPTION
[0045] The present invention is further described below in conjunction with specific embodiments, but the protection scope of the present invention is not limited thereto:
[0046] Embodiment 1: A method for supervising grain transport vehicles in a grain depot, such as Figure 1-2 As shown, the following steps are included:
[0047] S100, issue an IC card, create a grain transport vehicle track field, and bind the IC card number to the track ID.
[0048] When a grain truck enters a grain warehouse, it is necessary to register the grain truck, write the grain truck operation serial number into the IC card and issue it, and report the time of IC card issuance at the same time. Then, create the grain truck track field, bind the IC card number with the grain truck track ID. Optionally, binding means that the corresponding grain truck track ID can be directly queried through the IC card number, and the corresponding IC card number can also be directly queried through the grain truck track ID. After the IC card is issued, the starting point of the grain truck track is created. The coordinates and time of the starting point are the coordinates of the IC card issuance point and the time of IC card issuance, respectively.
[0049] S200, obtains grain truck trajectory data in real time.
[0050] The process of obtaining the grain truck trajectory is as follows Figure 2 As shown, specifically including:
[0051] S201, obtaining monitoring video stream data of the grain storage area.
[0052] Surveillance cameras are set up within the grain depot to obtain surveillance video stream data. It should be noted that the surveillance in the grain depot should be fully covered, and there is overlap in viewing angles between surveillance cameras.
[0053] S202, using a target detection algorithm and a multi-target tracking algorithm to extract the trajectory data of the grain transport vehicle under a single camera.
[0054] In the host computer, the surveillance video stream data of the same surveillance camera is input into the target detection algorithm model, and the vehicles in the surveillance video stream data are detected to obtain the grain truck target detection frame and confidence. Then, the grain truck target detection frame is passed through the multi-target tracking algorithm model to obtain the grain truck trajectory data.
[0055] As an example, the target detection algorithm model can be a YOLOv8 convolutional neural network, and the multi-target tracking algorithm model can be a BoT-SORT algorithm, wherein the core of the BoT-SORT algorithm is the grain truck ReID algorithm, the Kalman filter algorithm, and the Hungarian algorithm. The BoT-SORT algorithm uses the Kalman filter algorithm to predict the grain truck target detection frame of the current frame based on the grain truck target detection frame in the previous frame, and obtains the grain truck target detection frame features by using the grain truck ReID algorithm. Based on the grain truck target detection frame features and confidence, the real target detection frame of the current frame obtained by the target detection algorithm is matched with the Kalman filter predicted detection frame, and the successfully matched grain truck detection frame is used as the track segment of the current frame. The track segments of multiple frames constitute the grain truck track segment, and the grain truck track data is finally obtained by continuously extracting the grain truck track segments. The grain truck ReID algorithm is a grain truck feature re-identification algorithm, and the grain truck ReID algorithm can be a ResNest50 convolutional neural network.
[0056] As an implementation method of a target detection algorithm model generation method, an initial model and a data set of the target detection algorithm are constructed, wherein the data set includes pictures of grain trucks in a grain storage scenario; the data set is labeled to obtain a training set; and the initial model of the target detection algorithm is trained based on the training set to obtain a grain truck target detection model.
[0057] As an implementation method of a method for generating a grain truck ReID algorithm model, an initial model and a data set of the grain truck ReID algorithm are obtained, wherein the data set includes grain truck pictures; the data set is labeled to obtain a training set; and the initial model of the grain truck ReID algorithm is trained according to the training set to obtain a grain truck ReID model.
[0058] In this embodiment, the YOLOv8 convolutional neural network and the ResNest50 convolutional neural network are trained according to the training set and the test set. Among them, the training process of the convolutional neural network is divided into two stages. The first stage is the stage where data is propagated from low level to high level, that is, the forward propagation stage. The other stage is when the result obtained by the forward propagation does not meet the expectation, the error is propagated from the high level to the low level, that is, the back propagation stage. Specifically, the training process is:
[0059] The first step is to initialize the network weights;
[0060] In the second step, the input data is forward propagated through the convolution layer, downsampling layer, and fully connected layer to obtain the output value;
[0061] The input data are classified and labeled images of loaded vehicles and empty vehicles;
[0062] The third step is to find the error between the network output value and the target value;
[0063] Step 4: When the error is greater than the expected value, the error is passed back to the network, and the errors of the fully connected layer, downsampling layer, and convolutional layer are obtained in turn; the error of each layer can be understood as how much the network should bear for the total error of the network; when the error is equal to or less than the expected value, the training ends;
[0064] Step 5: Update the weights based on the obtained error, and then proceed to step 2.
[0065] S203, using a cross-camera trajectory association algorithm to obtain grain truck trajectory data.
[0066] The target detection algorithm model and the multi-target tracking algorithm model obtain the grain truck trajectory data under a single camera by inputting video data. The monitoring in the grain warehouse is fully covered, and there is overlap between the monitoring images. Therefore, there may be overlap between the multiple trajectory data extracted from multiple cameras, and the target detection algorithm model and the multi-target tracking algorithm model may also obtain the wrong trajectory of the grain truck, that is, the coordinate points in multiple trajectories of the same grain truck at the same time may be inconsistent. Therefore, it is necessary to use the cross-camera trajectory association algorithm to perform trajectory association operations on the grain truck trajectories extracted from multiple cameras, and merge or select multiple trajectories. The input of the cross-camera trajectory association algorithm is the grain truck trajectory data extracted from multiple cameras for the same grain truck, and the output is the grain truck trajectory data after association.
[0067] As an example, the cross-camera trajectory association algorithm is specifically as follows: first, the grain truck trajectory data obtained under a single camera in step S202 is converted into relative coordinates in the grain warehouse scene. When there are multiple grain truck trajectory data under multiple single cameras for the same grain truck, trajectory association is required. During the association, the coordinate points of multiple trajectories are traversed based on the time sequence:
[0068] (1) When there are overlapped coordinate points between multiple trajectory data of the grain transport vehicle at the same time, the overlapping multiple coordinate points are merged into one coordinate point;
[0069] (2) When there are inconsistent coordinate points in multiple trajectory data of the grain truck at the same time, the coordinate points are selected based on the confidence of the trajectory segment corresponding to the coordinate point. The confidence of the trajectory segment is the confidence of the target detection frame of the grain truck corresponding to the trajectory segment. The coordinate points with high confidence are spliced, and the coordinate points with low confidence are discarded;
[0070] (3) In addition to the above two cases (1) and (2), the coordinate points are directly spliced to finally obtain the associated grain transport vehicle trajectory data.
[0071] S300, extracting the stationary point and stationary time interval of the grain transport vehicle from the associated trajectory of the grain transport vehicle.
[0072] The stationary point of the grain truck is the location where the grain truck stops to perform business operations in the grain storage area, such as the tare weight measurement operation point and gross weight measurement operation point where the grain truck needs to be weighed, as well as the pre-inspection sampling operation point and grain loading and unloading operation point. The stationary time interval of the grain truck is the time interval of the grain truck at each stop point. If the coordinates of the stationary point of the grain truck are (x, y), the corresponding stationary time interval is [x t ,y t ], that is, the grain truck starts at time x t At time y t is stationary at coordinates (x,y).
[0073] As an example, the algorithm for extracting the stationary point and stationary time interval of the grain transport vehicle can be the Traj-DBSCAN algorithm (Robust and Hierarchical Stop Discovery in Sparse and Diverse Trajectories, Tran, Le Hung; Nguyen, Quoc Viet Hung; Do, Ngoc Hoan; Yan, Zhixian; 2011). In the Traj-DBSCAN algorithm, the stationary point in the trajectory is defined as a trajectory point cluster whose maximum distance between trajectory points is less than the maximum distance threshold ∈ and whose trajectory point time interval is greater than the minimum time threshold minTime. The trajectory point cluster refers to the set of trajectory points, and the trajectory point time interval refers to the difference between the maximum time and the minimum time of the trajectory points in the trajectory point cluster.
[0074] As an example, the maximum distance threshold ∈ is set to 10 m, and the minimum time threshold minTime is set to 30 s.
[0075] The pseudo code of the Traj-DBSCAN algorithm is shown in Table 1. The process is:
[0076] Table 1. Pseudo code of Traj-DBSCAN algorithm
[0077]
[0078] S400, matching the stationary point with the grain depot operation point area, and obtaining a first reference value of the operation time according to the stationary time interval.
[0079] S401, obtaining location area information of each operation point of the grain depot;
[0080] The grain warehouse operation points include tare weight measurement operation points, gross weight measurement operation points, pre-inspection sampling operation points and grain loading and unloading operation points. The grain warehouse operation point location area is a square parking area when the grain transport vehicle is operating at the tare weight measurement operation point, gross weight measurement operation point, pre-inspection sampling operation point and grain loading and unloading operation point, which includes four boundaries: top, bottom, left and right.
[0081] As an example, the grain depot operation point location area information includes the coordinate values of each operation point location area.
[0082] S402, comparing the location information of the stationary point with the location area information of the grain depot operation point in sequence. When the stationary point is a single coordinate point, when the minimum Euclidean distance between the coordinates of the stationary point and the boundary of the grain depot operation point location area is less than a preset distance threshold, the grain depot operation point location area matching the stationary point is recorded, and the stationary time interval of the current operation point is used as the first reference value of the operation time, and the comparison is stopped;
[0083] When the stationary point is a coordinate point cluster, when the coordinates of the stationary point exceeding the ratio threshold are in the location area, the location area of the grain storage operation point matching the stationary point is recorded, and the stationary time interval of the current operation point is used as the first reference value of the operation time, and the comparison is stopped.
[0084] As an example, when the stationary point is a single coordinate point, the preset distance threshold is 1m; when the stationary point is a coordinate point cluster, the ratio threshold is 60%, that is, when more than 60% of the coordinates of the stationary point are in the working point location area.
[0085] S403, the grain transport vehicle has not completed all operations (exiting the card and leaving the vehicle), and continues to generate operation trajectory data. It is necessary to continue to perform trajectory extraction, static point extraction and abnormality detection steps. Therefore, if there is no static point that matches the location information in the grain depot operation point location area information, repeat steps S200, S300 and S400 until the first reference value of the operation time and the grain depot operation point location area matching the static point are obtained.
[0086] S500, query the operation data reported by the IC card of the corresponding operation point matching the stationary point, obtain the second reference value of the operation time of the grain transport vehicle, compare the first reference value of the operation time of the grain transport vehicle with the second reference value, and issue an alarm when the comparison is abnormal.
[0087] The grain transport vehicle uses the IC card to report the operation data when performing the operation. Whenever the grain transport vehicle completes an operation at an operation point, the IC card is used to report the operation data. The operation data reported by the IC card includes the operation type (tare weight measurement operation, gross weight measurement operation, pre-inspection sampling operation and grain loading and unloading operation) and the operation time. The operation type corresponding to the stationary point is obtained according to the grain depot operation point area matched with the stationary point obtained in step S402. The time point of completing the operation reported by the IC card is queried according to the operation type as the second reference value of the operation time of the grain transport vehicle.
[0088] Specifically, the first reference value of the grain transport vehicle operation time is a time interval, and the second reference value is a moment. It is determined whether the second reference value of the grain transport vehicle operation time is within the time interval of the first reference value of the grain transport vehicle operation time; if not, it is determined whether the difference between the second reference value of the grain transport vehicle operation time and the upper or lower limit of the first reference value of the grain transport vehicle operation time exceeds a preset time difference threshold; if yes, an alarm is issued for abnormal comparison. The present invention obtains the second reference value of the grain transport vehicle operation time through data reported by an IC card, compares the first reference value of the grain transport vehicle operation time with the second reference value of the operation time, and issues an alarm when the comparison is abnormal, so that the corresponding relationship between the IC card and the grain transport vehicle can be monitored in real time.
[0089] As an example, the preset time difference threshold may be 1 minute.
[0090] S600, the grain transport vehicle trajectory data is grouped, and the grain transport vehicle trajectory anomaly detection and static point anomaly detection are performed. When an abnormality is detected, an alarm prompt is issued.
[0091] S601, obtaining the business type being performed by the grain transport vehicle; optionally, the business type includes grain storage-in business and grain storage-out business.
[0092] S602, obtaining the grain warehouse number where the grain transport vehicle is performing the business;
[0093] S603, grouping the grain transport vehicle trajectory data according to the grain depot business type performed by the grain transport vehicle and the grain depot number involved, and grouping the trajectories with the same grain depot business type and the same grain depot number involved in the same group.
[0094] S604, perform abnormal detection of grain truck trajectories and static points, and issue an alarm when abnormal detection is detected. Specifically, obtain a grain depot map and grid the grain depot map; gridding is to divide the grain depot map into grids. As an example, the grid size can be 5m×5m. According to the gridded grain depot map, the grain truck trajectory obtained in step 200 and the static points of the grain truck extracted in step 300 are transformed into coordinates of the grain truck trajectory. Coordinate transformation refers to converting the relative coordinates in the grain depot scene into grid coordinates; then, the grain truck trajectory and static points converted into grid coordinates are grouped according to steps S601-S603, and then anomaly detection is performed using the isolation-based trajectory anomaly detection algorithm (iBOAT: Isolation-Based Online Anomalous Trajectory Detection, Chao Chen; Daqing Zhang; Pablo Samuel Castro; Nan Li; Lin Sun; Shijian Li; Zonghui Wang, 2013).
[0095] The core idea of the isolation-based trajectory anomaly detection algorithm is based on the isolation forest algorithm. The pseudo code of the algorithm is shown in Table 2. In the grain warehouse scenario, the normal trajectories of grain trucks are numerous and similar, while the abnormal trajectories are few and special. Compared with normal trajectories, abnormal trajectories are easier to isolate. The core basis of the algorithm is that coordinate points that do not exist in normal trajectories often exist in abnormal trajectories. The anomaly score of a trajectory depends on the average number of times the trajectory is isolated from the trajectory set.
[0096] Table 2. Pseudocode example of the isolation-based trajectory anomaly detection algorithm
[0097]
[0098] The calculation formula of the anomaly score s of trajectory t is shown in formula (1), where N is the number of all trajectories in a trajectory group. c(N) is the average number of trajectory isolations after a given number of trajectories N, which is used to regularize the anomaly detection score. The calculation method of c(N) is shown in formula (2), where H(N-1) is the harmonic number, which can be estimated as shown in formula (3).
[0099]
[0100] H(N-1)≈ln(N-1)+0.57721566 (3)
[0101] When the average number of trajectory isolations approaches c(N), the trajectory anomaly score approaches 0.5; when the average number of trajectory isolations approaches 0, the trajectory anomaly score approaches 1; when the average number of trajectory isolations approaches N-1, the trajectory anomaly score approaches 0; when the trajectory anomaly score obtained by the grain truck trajectory anomaly detection exceeds the trajectory anomaly threshold, the trajectory is considered abnormal and an anomaly alarm is issued.
[0102] As an example, the number of runs m is set to 20, the subsample size ψ is set to 80% of the number of trajectories in the trajectory set T, and the maximum number of isolations is set to 50. The anomaly threshold in stationary point anomaly detection is set to 0.5, and the anomaly threshold in trajectory anomaly detection is set to 0.4.
[0103] S700: Store the data in a database.
[0104] The data includes trajectory data, trajectory point data, IC reported business flow data, extracted trajectory static point data and static time interval data, and grain transport vehicle operation abnormality detection result data.
[0105] The trajectory data includes a trajectory ID and a trajectory point sequence, and the trajectory point data includes coordinate data and coordinate point time data.
[0106] The implementation principle of the embodiment of the present invention is:
[0107] By binding the IC card with the grain truck trajectory when issuing the IC card, the correspondence between the IC card and the grain truck can be guaranteed; by acquiring the grain truck trajectory in real time and extracting the grain truck's stationary point and stationary time interval data, the first reference value of the grain truck's operating time can be obtained; the second reference value of the grain truck's operating time can be obtained through the IC card reporting data; by comparing the first and second reference values of the grain truck's operating time and issuing an alarm when the comparison is abnormal, the correspondence between the IC card and the grain truck can be monitored in real time, and the above process can be repeated before the grain truck completes all operations to ensure supervision of the entire operation process of the grain truck; through the grain truck trajectory anomaly detection and grain truck stationary point anomaly detection, the grain truck trajectory data can be effectively and fully utilized, so that the grain depot can be managed in a systematic, intelligent, and information-based manner, which is conducive to the construction of smart grain depots.
[0108] Finally, it should be noted that the above examples are only some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and there are many variations. All variations that can be directly derived or associated with the content disclosed by a person skilled in the art should be considered as the protection scope of the present invention.
Claims
1. A method for supervising grain transport vehicles in a grain depot, characterized in that: The following steps are involved: S100, binding the card number of the IC card with the track ID of the grain transport vehicle; S200, obtaining the grain transport vehicle trajectory in real time from the surveillance video data in the grain storage area; S300, extracting the stationary point and stationary time interval of the grain transport vehicle from the trajectory of the grain transport vehicle; S400, matching the stationary point with the grain depot operation point area, and taking the stationary time interval of the matched stationary point as the first reference value of the operation time; S500, taking the time point when the grain transport vehicle at the stationary point completes the operation as the second reference value of the operation time, and if the difference between the time interval of the second reference value of the operation time and the time interval of the first reference value of the operation time exceeds a preset time difference threshold, issuing an alarm prompt; S600, grouping the grain transport vehicle trajectory data and performing grain transport vehicle trajectory anomaly detection and static point anomaly detection, and issuing an alarm prompt when an abnormality is detected.
2. A method for supervising grain transport vehicle operations in a grain depot according to claim 1, characterized in that: The process of obtaining the grain transport vehicle trajectory is as follows: The surveillance video stream data of the same surveillance camera is input into the target detection algorithm model, and the vehicles in the surveillance video stream data are detected to obtain the target detection frame and confidence of the grain truck. The target detection frame of the grain truck is passed through the multi-target tracking algorithm model to obtain the trajectory data of the grain truck under a single camera; Then, the cross-camera trajectory association algorithm is used to obtain the associated grain truck trajectory data.
3. A method for supervising grain transport vehicle operations in a grain depot according to claim 2, characterized in that: The cross-camera trajectory association algorithm is specifically: The grain truck trajectory data under a single camera is converted into relative coordinates in the grain warehouse scene. Then, the trajectory data of the same grain truck under multiple single cameras is traversed based on the time sequence. (1) When there are overlapped coordinate points in different trajectory data at the same time, the overlapping multiple coordinate points are merged into one coordinate point; (2) When there are inconsistent coordinate points in different trajectory data at the same time, the coordinate points with high confidence are spliced and the coordinate points with low confidence are discarded; (3) In addition to the above two cases, the coordinate points are directly spliced to obtain the associated grain truck trajectory data.
4. A method for supervising grain transport vehicle operations in a grain depot according to claim 3, characterized in that: The target detection algorithm model is the YOLOv8 convolutional neural network, and the multi-target tracking algorithm model is the BoT-SORT algorithm.
5. A method for supervising grain transport vehicle operations in a grain depot according to claim 4, characterized in that: The stationary point and stationary time interval of the grain transport vehicle are obtained from the associated trajectory of the grain transport vehicle through the Traj-DBSCAN algorithm.
6. A method for supervising grain transport vehicle operations in a grain depot according to claim 5, characterized in that: The process of obtaining the first reference value of the operation time is as follows: The grain depot operation point location area includes the tare operation point, gross weight measurement operation point, pre-inspection sampling operation point and grain loading and unloading operation point area. The static point is compared with the location area of each grain depot operation point in turn: (1) When the stationary point is a single coordinate point, when the minimum Euclidean distance between the stationary point coordinates and the boundary of the grain depot operation point location area is less than the preset distance threshold, the stationary time interval of the current operation point is used as the first reference value of the operation time, and the grain depot operation point location area matching the stationary point is recorded; (2) When the stationary point is a coordinate point cluster, when the coordinates of the stationary point exceeding the ratio threshold are in the grain depot operation point location area, the stationary time interval of the current operation point is used as the first reference value of the operation time, and the grain depot operation point location area matching the stationary point is recorded; (3) If no grain storage operation point matches the static point, repeat steps S200, S300 and S400 until the first reference value of the operation time is obtained.
7. A method for supervising grain transport vehicle operations in a grain depot according to claim 6, characterized in that: The time point at which the grain transport vehicle at the stationary point completes the operation is when the grain transport vehicle uses the IC card to report the operation data when performing the operation, including the operation type and the time point at which the operation is completed.
8. A method for supervising grain transport vehicle operations in a grain depot according to claim 7, characterized in that: The process of trajectory anomaly detection and stationary point anomaly detection is as follows: The grain depot map is gridded, and the relative coordinates of the associated grain truck trajectory and the grain truck's stationary point in the grain depot scene are converted into grid coordinates; then the grain truck trajectory and stationary point converted into network coordinates are grouped according to the business type performed by the grain truck and the granary number, and then the grouped grain truck trajectory and stationary point are respectively subjected to anomaly detection using an isolation-based trajectory anomaly detection algorithm.
Citation Information
Patent Citations
Coaling branch station and coaling system
CN107194297A
Granary inbound and outbound management method and device, and computer readable storage medium
CN108460542A
Grain depot vehicle tracking method based on YOLOv5 and improved StrongSORT
CN116580333A
Multi-feature fusion special operation vehicle working state automatic identification system and method
CN117237692A
Analytical apparatus and analytical method
JP2019028979A
Cited By
Grain logistics supervision method and system considering sensitive information
CN120509802A
Warehouse in-out abnormal behavior detection method based on multi-dimensional feature trajectory
CN120599510A
