Belt state monitoring method and device fusing video coding motion vector and target detection, equipment and medium

Abstract

The application discloses a belt state monitoring method and device fusing video coding motion vectors and target detection, equipment and medium, the method comprises the following steps: acquiring the monitoring video stream of the belt and decoding to obtain the motion vector of the motion macro block in each video frame; input the video frame into the YOLO model trained in advance to identify the belt boundary and load state in the video frame; according to the motion vector, the motion intensity of the belt in the belt boundary area is calculated, and the running state of the belt is judged according to the motion intensity. The above-mentioned method realizes real-time, non-contact and high-precision judgment of the running state of the belt, and has strong anti-interference ability to light change and dust interference, simultaneously outputs the running state and load state of the belt, and supports multiple abnormality detection.

Description

Technical Field

[0001] This application relates to the field of target detection technology, and in particular to a belt condition monitoring method, apparatus, equipment and medium that integrates video-coded motion vectors and target detection. Background Technology

[0002] Belt conveyors are widely used in industrial settings such as mines, power plants, and ports, and their operating status directly affects production safety and efficiency. Traditional belt conveyor operation status detection relies heavily on contact sensors (such as speed sensors and photoelectric switches), which suffer from problems such as complex installation, easy wear, and significant susceptibility to environmental factors.

[0003] In recent years, vision-based non-contact detection methods have gradually emerged, such as using background modeling, optical flow, or deep learning techniques to determine the operating status of conveyor belts.

[0004] For example, Chinese patent application 202311802201.9 discloses a belt unload detection method based on support vector machine, including the following steps: 1) acquiring belt images and performing preprocessing; 2) selecting features from the preprocessed images and acquiring image data; 3) labeling the acquired belt images as unloaded (0) and loaded (1), and dividing the dataset after feature acquisition and labeling into training set and test set; 4) importing the training set into support vector machines based on different kernel functions for training to obtain the support vector machine classification model; 5) classifying the test set using the support vector machine classification model to obtain the accuracy of the test set.

[0005] For example, Chinese patent application 202410690891.1 discloses an intelligent monitoring method and device for belt conveyors based on image recognition. This includes a multi-task, multi-branch U-shaped network. The inputs to the U-shaped network are color images and optical flow images. The backbone of the U-shaped network includes multiple upsampling modules and multiple downsampling modules. The branches of the U-shaped network include multiple split modules, reshape modules, merge modules, BILSTM modules, down5 modules, and multiple FC modules. The outputs of the U-shaped network include start / stop status 2-classification, personnel segmentation feature maps, belt idler segmentation feature maps, belt segmentation feature maps, foreign object segmentation feature maps, coal segmentation feature maps, and coal pile segmentation feature maps. Figure 7 Each output is processed to obtain the detection results of different defects in the belt.

[0006] These methods are typically computationally complex, have poor real-time performance, and are easily affected by environmental factors such as changes in lighting and dust interference. Summary of the Invention

[0007] This application provides a belt state monitoring method, apparatus, device, and medium that integrates video-coded motion vectors and target detection, to solve the technical problems of existing vision-based non-contact detection methods, such as computational complexity, poor real-time performance, and susceptibility to environmental factors such as changes in lighting and dust interference.

[0008] The first aspect of this application provides a belt state monitoring method that integrates video-coded motion vectors and target detection, comprising:

[0009] Acquire the monitoring video stream of the belt and decode it to obtain the motion vector of the motion macroblock in each video frame;

[0010] The video frames are input into a pre-trained YOLO model to identify the belt boundaries and load states in the video frames;

[0011] The motion intensity of the belt within the belt boundary area is calculated based on the motion vector, and the operating state of the belt is determined based on the motion intensity.

[0012] Preferably, calculating the motion intensity of the belt within the belt boundary region based on the motion vector data includes:

[0013] The motion vectors are filtered according to the belt boundary, and the motion vectors whose starting coordinates are located within the boundary area are retained as the first motion vectors. The motion intensity of the belt within the belt boundary area is calculated based on the first motion vector.

[0014] Preferably, calculating the motion intensity of the belt within the belt boundary region based on the first motion vector includes:

[0015] Based on the amplitude of the first motion vector, the first motion vector with a motion amplitude within a preset amplitude threshold range is retained as the second motion vector, and the motion intensity of the belt within the belt boundary area is calculated based on the second motion vector.

[0016] Preferably, the step of calculating the motion intensity of the belt within the belt boundary region based on the second motion vector further includes:

[0017] Based on the consistency of the direction of the motion vectors, the second motion vector whose deviation from the main direction is within a preset threshold range is retained as the third motion vector, and the motion intensity of the belt within the belt boundary area is calculated based on the third motion vector.

[0018] Preferably, the second motion vector whose deviation from the main direction is within a preset threshold range is the third motion vector, including:

[0019] The angle corresponding to the average unit vector of the direction angles of all the second motion vectors is the principal direction.

[0020] Preferably, calculating the motion intensity of the belt within the belt boundary region based on the motion vector data includes:

[0021] The average magnitude of all the third motion vectors is calculated as the motion intensity.

[0022] Preferably, the method for determining the operating state of the belt based on the motion intensity includes:

[0023] When the motion intensity exceeds a preset high threshold in at least M frames out of a consecutive T frames, it is determined to be in a running state.

[0024] When the motion intensity is lower than a preset low threshold for at least N frames in a series of T frames, the motion is determined to be in a stopped state.

[0025] Where M > T - N.

[0026] A second aspect of this application provides a belt state monitoring device that integrates video-coded motion vectors and target detection, comprising:

[0027] The decoding module is used to acquire the monitoring video stream of the belt and decode it to obtain the motion vector of the motion macroblock in each video frame;

[0028] The boundary and load recognition module is used to input the video frame into a pre-trained YOLO model to identify the belt boundary and load status in the video frame;

[0029] The status determination module is used to calculate the motion intensity of the belt within the belt boundary area based on the motion vector, and to determine the operating status of the belt based on the motion intensity.

[0030] A third aspect of this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it causes the electronic device to perform the method described in the first aspect of this application.

[0031] A fourth aspect of this application provides a computer-readable storage medium for storing a computer program that, when run on a computer, causes the computer to perform the method described in the first aspect of this application.

[0032] Beneficial effects

[0033] This invention proposes a belt status monitoring method that integrates video-encoded motion vectors and target detection. By collecting motion vectors generated during video encoding, the belt's operating status is calculated, avoiding redundant calculations and significantly reducing computational complexity. Combined with a target detection algorithm, the belt area is accurately located, constraint areas are provided, and background interference is eliminated. The above method achieves real-time, non-contact, and high-precision judgment of the belt's operating status, and has strong anti-interference capabilities against changes in lighting and dust interference. It also outputs the belt's operating status and load status, and supports various anomaly detection methods. Attached Figure Description

[0034] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0035] Figure 1 This is a flowchart of a belt state monitoring method that integrates video-coded motion vectors and target detection according to an embodiment of this application.

[0036] Figure 2 These are video frames containing the original motion vectors obtained according to embodiments of this application;

[0037] Figure 3 This is a schematic diagram of the belt boundary according to an embodiment of this application;

[0038] Figure 4 This is a schematic diagram illustrating the selection of motion amplitude according to an embodiment of this application;

[0039] Figure 5 This is a schematic diagram of the filtering direction according to an embodiment of this application;

[0040] Figure 6 This is a schematic diagram of a belt state monitoring device that integrates video-coded motion vectors and target detection according to an embodiment of this application.

[0041] Figure 7 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0042] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0044] In existing technologies, vision-based non-contact detection methods require complete decoding after acquiring the video stream to obtain a complete pixel image (YUV or RGB format) for each frame. Then, complex computer vision algorithms (such as optical flow) are run to re-estimate the motion of pixels or blocks, a process that involves a huge amount of computation.

[0045] To reduce the large amount of computation in existing technologies, embodiments of this application provide a belt state monitoring method that integrates video-coded motion vectors and target detection, such as... Figure 1 The diagram shown is a schematic representation of a belt state monitoring method that integrates video-coded motion vectors and target detection according to an embodiment of this application. Figure 1 It can be seen that the method includes the following steps:

[0046] Step S102: Obtain the monitoring video stream of the belt and decode it to obtain the motion vector of the motion macroblock in each video frame.

[0047] In video coding, motion macroblocks are the basic units used in standards such as H.264 to describe motion information within a video frame. A macroblock is typically a 16x16 pixel luma block and a corresponding 8x8 pixel chroma block (U and V components). The encoder divides each frame into these non-overlapping blocks for independent processing.

[0048] Motion estimation techniques analyze changes between adjacent frames to find the most similar matching block in the reference frame for each macroblock, and record their relative displacements (i.e., motion vectors) and residual data (difference information). This way, the encoder only needs to store motion vectors and residuals, rather than the pixel values ​​of the entire macroblock, thus effectively compressing the data. For example, during motion compensation, the encoder uses motion vectors to predict the current macroblock from the reference frame and improves prediction accuracy and reduces residuals through interpolation (e.g., 1 / 4 pixel precision).

[0049] Therefore, motion macroblocks and their motion vectors are "byproducts" calculated by standard video encoders (such as H.264 / HEVC) during the "encoding" process for video compression. The method of this invention does not recalculate motion using traditional visual algorithms (such as optical flow), but rather partially decodes the video stream at the receiving end, utilizing motion information already included in the encoding standard, such as the position, size, and motion vector values ​​of motion macroblocks. The decoder directly extracts this existing motion information when parsing the stream. This reuses encoding computing power and avoids redundant calculations, which is key to reducing computational complexity.

[0050] Specifically, in the early stages of the decoding process (after entropy decoding), this motion information is extracted as metadata and used directly for analysis, omitting the complete pixel reconstruction and complex motion estimation algorithms found in existing technologies. By utilizing the front end of the standard video decoding process, pre-calculated motion information can be obtained quickly and conveniently, and then used as the core data source for subsequent analysis, thereby achieving an order-of-magnitude improvement in computational efficiency at the source.

[0051] The above steps specifically include the following processes:

[0052] Start the decoder: Use a library such as FFmpeg, OpenCV VideoCapture, or a hardware decoding library to open the video stream.

[0053] Configure decoding mode: Set the decoder to motion vector extraction mode.

[0054] Decoding and Extraction: The decoder parses the network abstraction layer units of the bitstream. It performs entropy decoding on the slices to obtain macroblock-level syntax elements. During the parsing of these syntax elements, information such as motion vectors, reference frame indices, macroblock types, and macroblock locations are directly read and stored in the memory structure. At this point, the decoder typically has not yet performed the steps required for pixel reconstruction, such as motion compensation, inverse transform, and inverse quantization.

[0055] Data Output: This information is read directly from the metadata interface provided by the decoder to form motion vectors. Each motion vector is represented as [i, block_w, block_h, src_x, src_y, dst_x, dst_y], where i represents the current frame ID, block_w and block_h represent the width and height of the motion macroblock, respectively, and the motion macroblock is the basic unit of motion information. src_x and src_y represent the top-left pixel coordinates of the motion macroblock in the current frame, indicating the starting point of the motion, and dst_x and dst_y represent the top-left pixel coordinates of the block in the reference frame that best matches the current motion macroblock (usually the previous or next frame), indicating the ending point of the motion.

[0056] Step S104: Input the video frame into the pre-trained YOLO model to identify the belt boundary and load state in the video frame.

[0057] like Figure 2 In the embodiment shown, the belts in the video frame are arranged from far to near. Therefore, in order to further reduce the amount of computation and reduce the impact of distant interference in the image, the video frame image is cropped during this step. For example, only the area at the bottom center of the video frame image is retained, which accounts for about 1 / 12 of the entire video frame image.

[0058] Furthermore, the video stream is sampled at a low frame rate (e.g., 1fps), and the sampled frames are input into a pre-trained YOLO object detection model. The model outputs the bounding box of the conveyor belt region and its corresponding state (coal present / no coal) in each frame. If multiple conveyor belts are detected, one or all of them can be selected for monitoring to meet different application requirements.

[0059] Step S106: Calculate the motion intensity of the belt within the belt boundary area based on the motion vector, and determine the operating state of the belt based on the motion intensity.

[0060] The embodiments of this invention cleverly utilize the characteristic that motion vectors generated during the encoding process of video compression standards can directly reflect the motion information of objects in the video. By analyzing the motion vector information in the bitstream, complex calculations on pixel-level data can be avoided, thereby significantly improving detection efficiency. On the other hand, the target detection algorithm can quickly and accurately locate the belt area and determine whether it is loaded, while providing spatial constraints for motion analysis.

[0061] The above method makes full use of the motion vectors generated during video encoding, avoiding the complex pixel-level calculations or massive neural network inferences in existing technologies, greatly reducing computational overhead and enabling real-time edge computing.

[0062] The above method uses the precise spatial mask provided by YOLO to eliminate motion interference in non-belt areas. It is completely insensitive to changes in lighting, dust, lens stains, and background movement, and has strong robustness.

[0063] The above method can output two key types of information, namely "running status" and "load status", providing more comprehensive data support for intelligent operation and maintenance.

[0064] The above method can be seamlessly integrated into existing monitoring systems without the need to install contact sensors or deploy multiple dedicated models. Analysis can be completed using existing monitoring video streams, and low frame rate target detection models can use the CPU for inference, significantly reducing implementation costs.

[0065] In some preferred embodiments, in order to reduce the computational load and perform efficient computation, after extracting the motion vectors of all motion macroblocks in the current frame, step S106 specifically includes the following steps:

[0066] Step S1061: Filter the motion vectors according to the belt boundary, and retain the motion vectors whose starting coordinates are located within the boundary area as the first motion vector; in this way, all interference factors outside the boundary area (such as personnel movement, equipment vibration, background changes) can be deleted.

[0067] Specifically, such as Figure 3As shown, the belt boundary is selected based on the belt region bounding box R(x_min, y_min, x_max, y_max) provided by YOLO, filtering out motion vectors that are completely within this region, i.e., retaining data where src_x and src_y are within the range of R. Specifically, the starting point (src_x, src_y) of the motion macroblock must be greater than or equal to the upper left corner (x_min, y_min) of the bounding box, and the lower right corner of the motion macroblock must be less than or equal to the lower right corner of the bounding box. Since the size of the motion macroblock is known, the coordinates of the lower right corner of the motion macroblock are (src_x + block_w, src_y + block_h). Therefore, it is necessary to satisfy...

[0068] x_min≦src_x≦x_max-block_w

[0069] y_min≦src_y≦y_max-block_h.

[0070] Step S1062: Filter according to the amplitude of the first motion vector, and retain the first motion vector whose motion amplitude is within the preset amplitude threshold range as the second motion vector.

[0071] Specifically, such as Figure 4 As shown, the upper limit of the amplitude threshold range is set to... The maximum limit filters out abnormally large movements caused by sudden lens shake, scene transitions, or large objects suddenly appearing in the road. These are abnormal events and do not represent the normal or stopped state of the conveyor belt; the lower limit is... The min filter can filter out tiny, meaningless jitters caused by image noise and compression errors, which do not represent the movement of real objects.

[0072] By filtering out the first motion vectors that satisfy the following requirement, both small and large motion changes can be filtered out. ,in:

[0073] The magnitude of the motion vector represents the displacement of the macroblock from the current frame to the reference frame;

[0074] m x Let m be the component of the motion in the x-direction. x =dst_x-src_x;

[0075] m y Let m be the component of the motion in the y-direction. y =dst_y-src_y.

[0076] Step S1063: Based on the consistency of the direction of the motion vectors, filter and retain the second motion vector whose motion direction deviates from the main direction within a preset threshold range as the third motion vector. The main direction is obtained by calculating the angle corresponding to the average unit vector of the direction angles of all the second motion vectors. The motion intensity of the belt in the belt boundary area is calculated based on the average amplitude of the third motion vector.

[0077] After directional consistency screening, it can be ensured that the motion vectors used for analysis are consistent in direction, thus more accurately representing the unidirectional movement of the belt and filtering out outlier vectors with opposite directions (e.g., local reverse movement caused by falling coal blocks, or other disturbances).

[0078] Specifically, such as Figure 5 As shown, the process includes the following steps:

[0079] Step S10631: To solve the problem of angular circumference, for example, 359° and 1° differ greatly in numerical value but are almost the same in direction, the direction angle of each motion vector is... Convert to a point on the unit circle to form a unit vector. .

[0080] Step S10632: Summate all unit vectors and take the average to obtain the average unit vector. The mean cosine value Mean sine value N represents the total number of motion macroblocks.

[0081] The direction this average unit vector points to It is the principal direction of all motion vectors, that is, the dominant direction of belt motion.

[0082] Step S10633: Calculate the deviation between the direction of each motion vector and the principal direction. Set angle threshold ,if Perform outlier motion vector filtering.

[0083] The above process further purified the motion data, and by calculating θ avg It automatically adapts to the installation angle of the belt, eliminating the need for manual pre-setting of the movement direction, thus enhancing the method's adaptability and versatility.

[0084] In some preferred embodiments, the method for determining the operating state of the belt based on the motion intensity includes:

[0085] A dynamic threshold is set, for example, the dynamic threshold is adaptively adjusted based on historical data. For each frame, the motion intensity is calculated to determine whether it exceeds the high threshold. If it is greater than or equal to the threshold, the belt running status is marked as 1; if it is less than the threshold, the belt running status is marked as 0.

[0086] To prevent the state from frequently fluctuating near the threshold, the following method is used to achieve a hysteresis effect by implementing asymmetric judgments for switching from motion to stop and from stop to motion:

[0087] When the motion intensity exceeds a preset high threshold in at least M frames out of a consecutive T frames, it is determined to be in a running state.

[0088] When the motion intensity is lower than a preset low threshold for at least N frames in a series of T frames, the motion is determined to be in a stopped state.

[0089] M is the minimum number of frames required to switch from stop to motion and that are considered motion candidates, while N is the minimum number of frames required to switch from motion to stop and that are considered stop candidates.

[0090] To make the state transition lag-dependent, the conditions for starting should be more stringent than those for stopping, and vice versa; therefore, M > T - N. For example, if T = 5, and we want to achieve "difficult to start, easy to stop," we can set M = 4 and N = 2. Then 4 > 5 - 2, satisfying the condition that at least 4 out of 5 consecutive frames show motion for the belt to be considered truly started, while at least 2 out of 5 consecutive frames show stopping for the belt to be considered truly stopped. This effectively avoids misjudgments caused by instantaneous fluctuations.

[0091] Through the entire process described above, the current overall status of the conveyor belt (running / stopping, coal present / no coal) is output; if an abnormal status is detected (such as the conveyor belt stopping but coal is detected, which may indicate a blockage; or the conveyor belt running unloaded for a long time), an alarm mechanism is triggered.

[0092] Based on the same inventive concept as the above-described method embodiments, this application also provides a belt state monitoring device that integrates video-coded motion vectors and target detection, such as... Figure 6 The diagram shown is a schematic of a belt state monitoring device that integrates video-coded motion vectors and target detection. The device includes:

[0093] The decoding module is used to acquire the monitoring video stream of the belt and decode it to obtain video frames;

[0094] The extraction module is used to extract the motion vectors of motion macroblocks in each video frame;

[0095] The boundary and load recognition module is used to input the video frame into a pre-trained YOLO model to identify the belt boundary and load status in the video frame;

[0096] The status determination module is used to calculate the motion intensity of the belt within the belt boundary area based on the motion vector, and to determine the operating status of the belt based on the motion intensity.

[0097] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0098] Based on the same inventive concept as the above method embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it enables the electronic device to implement the control method described in the above embodiments.

[0099] In one embodiment, the electronic device may be a server, and in this embodiment, the structure of the electronic device may be as follows: Figure 7 As shown, it includes a memory 2001, a communication module 2003, and one or more processors 2002.

[0100] The memory 2001 is used to store computer programs executed by the processor 2002. The memory 2001 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required to run instant messaging functions, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0101] Memory 2001 may be volatile memory, such as random access memory (RAM); memory 2001 may also be nonvolatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 2001 may be any other medium capable of carrying or storing a desired computer program having the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 2001 may be a combination of the above-mentioned memories.

[0102] Processor 2002 may include one or more central processing units (CPUs) or digital processing units, etc. Processor 2002 is used to implement the above-mentioned audio data processing method when calling computer programs stored in memory 2001.

[0103] The communication module 2003 is used to communicate with terminal devices and other servers.

[0104] This application embodiment does not limit the specific connection medium between the memory 2001, communication module 2003, and processor 2002. This application embodiment... Figure 7 The memory 2001 and the processor 2002 are connected via a bus 2004, which is in... Figure 7 The connections between other components are illustrated with arrows and are for illustrative purposes only, not as limiting information. The Bus 2004 can be divided into address bus, data bus, control bus, etc. For ease of description, Figure 7 The text uses only one arrow to describe it, but does not indicate that there is only one bus or one type of bus.

[0105] Based on the same inventive concept as the above-described method embodiments, embodiments of the present invention also provide a computer-readable storage medium for storing a computer program. When the computer program is run on a computer, it enables the electronic device to implement the control method described in the above embodiments. The computer-readable storage medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CDROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0106] Based on the same inventive concept as the above-described method embodiments, embodiments of the present invention also provide a computer program product, which includes a computer program that, when run on an electronic device, causes the electronic device to perform the steps of the control methods described above according to various exemplary embodiments of this application. The program product may take the form of any combination of one or more readable media. These computer program commands can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the commands executed by the processor of the computer or other programmable data processing device generate a process for implementing... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

Claims

1. A belt state monitoring method integrating video-coded motion vectors and target detection, characterized in that, include: Acquire the monitoring video stream of the belt and decode it to obtain the motion vector of the motion macroblock in each video frame; The video frames are input into a pre-trained YOLO model to identify the belt boundaries and load states in the video frames; The motion intensity of the belt within the belt boundary area is calculated based on the motion vector, and the operating state of the belt is determined based on the motion intensity.

2. The method according to claim 1, characterized in that, The step of calculating the motion intensity of the belt within the belt boundary area based on the motion vector data includes: The motion vectors are filtered according to the belt boundary, and the motion vectors whose starting coordinates are located within the boundary area are retained as the first motion vectors. The motion intensity of the belt within the belt boundary area is calculated based on the first motion vector.

3. The method according to claim 2, characterized in that, The step of calculating the motion intensity of the belt within the belt boundary region based on the first motion vector includes: Based on the amplitude of the first motion vector, the first motion vector with a motion amplitude within a preset amplitude threshold range is retained as the second motion vector, and the motion intensity of the belt within the belt boundary area is calculated based on the second motion vector.

4. The method according to claim 3, characterized in that, The step of calculating the motion intensity of the belt within the belt boundary region based on the second motion vector further includes: Based on the consistency of the direction of the motion vectors, the second motion vector whose deviation from the main direction is within a preset threshold range is retained as the third motion vector, and the motion intensity of the belt within the belt boundary area is calculated based on the third motion vector.

5. The method according to claim 3, characterized in that, The second motion vector, whose deviation from the main direction is within a preset threshold range, is the third motion vector, including: The angle corresponding to the average unit vector of the direction angles of all the second motion vectors is the principal direction.

6. The method according to claim 5, characterized in that, The step of calculating the motion intensity of the belt within the belt boundary area based on the motion vector data includes: The average magnitude of all the third motion vectors is calculated as the motion intensity.

7. The method according to claim 1, characterized in that, The method for determining the operating state of the belt based on the motion intensity includes: When the motion intensity exceeds a preset high threshold in at least M frames out of a consecutive T frames, it is determined to be in a running state. When the motion intensity is lower than a preset low threshold for at least N frames in a series of T frames, the motion is determined to be in a stopped state. Where M > T - N.

8. A belt state monitoring device integrating video-coded motion vectors and target detection, characterized in that, include: The decoding module is used to acquire the monitoring video stream of the belt and decode it to obtain the motion vector of the motion macroblock in each video frame; The boundary and load recognition module is used to input the video frame into a pre-trained YOLO model to identify the belt boundary and load status in the video frame; The status determination module is used to calculate the motion intensity of the belt within the belt boundary area based on the motion vector, and to determine the operating status of the belt based on the motion intensity.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it causes the electronic device to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when run on a computer, causes the computer to perform the method as described in any one of claims 1 to 7.

Images