A video monitoring method and system for underground coal mine

Abstract

The present application relates to the technical field of video compression, and in particular to a video monitoring method and system in a coal mine, which solves the technical problem that in the prior art, traditional Huffman coding in video compression in a coal mine only relies on frequency distribution weight, ignores important safety information of low frequency, and causes the image of a safety hazard area to be blurred after compression, affecting abnormality identification. The method comprises: acquiring a video monitoring image in a coal mine; dividing the video monitoring image into regions, and determining the importance weight of a plurality of regions based on image features and motion features in the regions; correcting an initial weight parameter used in video coding according to the importance weight to obtain a target weight parameter after correction; the initial weight parameter is determined based on the frequency of occurrence of a gray value in the image; and the video monitoring image is Huffman coded and compressed using the target weight parameter to generate compressed video data for transmission to a ground monitoring center.

Description

Technical Field

[0001] This invention relates to the field of video compression technology, specifically to a video monitoring method and system for underground coal mines. Background Technology

[0002] In the field of coal mine safety production, video surveillance is a key technology for ensuring the safety of underground operations and preventing potential risks. Through real-time video transmission, it allows ground control to promptly grasp the underground situation. With the development of digital communication technology, in order to effectively transmit the massive amounts of surveillance video data under the limited bandwidth conditions underground, video compression methods, such as Huffman coding, are commonly used. This method constructs prefix codes based on the frequency of source symbols, reducing the overall data volume by assigning short codes to high-frequency characters, thus achieving lossless compression.

[0003] However, traditional Huffman coding methods only set weights based on character frequency, ignoring certain low-frequency details in the video that are crucial to safety, such as a small number of sharp objects stuck on a conveyor belt. These objects have low symbol frequencies but pose a high safety risk. If compressed using existing methods, the image in that area may become blurred, thus affecting the accuracy of anomaly detection and making it impossible to identify potential safety hazards in a timely manner. Summary of the Invention

[0004] To address the technical problem that traditional Huffman coding in underground coal mine video compression relies solely on frequency-based weighting, neglecting crucial low-frequency safety information, resulting in blurred images of safety hazard areas after compression and hindering anomaly identification, this invention aims to provide a video monitoring method and system for underground coal mines. The specific technical solution adopted is as follows: Firstly, a video surveillance method for underground coal mines is provided, comprising: acquiring video surveillance images from underground coal mines; dividing the video surveillance images into regions and determining the importance weights of multiple regions based on image features and motion features within the regions; correcting the initial weight parameters used in video coding according to the importance weights to obtain corrected target weight parameters; determining the initial weight parameters based on the frequency of grayscale values ​​in the image; and using the target weight parameters to perform Huffman coding compression on the video surveillance images to generate compressed video data for transmission to a ground monitoring center.

[0005] Based on the above technical solution, in the video monitoring method for underground coal mines provided by this invention, video monitoring images of underground coal mines are acquired and divided into regions. The importance weights are determined based on the image features and motion features of each region. Then, the initial weight parameters that only rely on the grayscale frequency are corrected. Finally, the corrected target weight parameters are used for Huffman coding compression. This effectively improves the image clarity of safety-critical areas (such as foreign objects or abnormally moving areas) in the video while ensuring overall compression efficiency. It avoids the blurring problem caused by traditional methods ignoring important low-frequency information and ensures that the ground monitoring center can accurately identify safety hazards.

[0006] In conjunction with the first aspect above, in one possible implementation, the aforementioned multiple regions include the entire area of ​​the video surveillance image; the method for determining the importance weights of multiple regions based on image features and motion features within the regions specifically includes: determining the foreign object degree of each region based on the image features within each region; the foreign object degree is used to characterize the probability that a region contains foreign objects; determining the safety hazard index of each region based on the motion features of each region; the safety hazard index is used to characterize the degree of safety risk of the region; and determining the importance weight of each region based on the foreign object degree and the safety hazard index.

[0007] In conjunction with the first aspect above, in one possible implementation, the aforementioned multiple regions only include foreign object regions in the video surveillance images; the method for determining the importance weights of multiple regions based on image features and motion features within the regions specifically includes: determining the foreign object degree of each region based on the image features within each region, and filtering foreign object regions based on the foreign object degree; the foreign object degree is used to characterize the probability that a region contains foreign objects; determining the safety hazard index of the foreign object region based on the motion features of the foreign object region; the safety hazard index is used to characterize the degree of safety risk of the region; and determining the importance weights of the foreign object regions based on the foreign object degree and the safety hazard index.

[0008] In conjunction with the first aspect above, in one possible implementation, the aforementioned image features include grayscale distribution and gradient changes; the method for determining the foreign object degree of each region based on the image features within each region specifically includes: determining the foreign object degree based on the grayscale value, the number of pixels, and the gradient difference in grayscale values ​​between pixels within each region.

[0009] In conjunction with the first aspect above, in one possible implementation, the motion features include the displacement of pixels in consecutive frames; the method for determining the safety hazard index based on the motion features specifically includes: determining the safety hazard index based on the displacement distance of pixels within the region in consecutive frames, and the similarity between the displacement direction and a preset direction.

[0010] In conjunction with the first aspect above, in one possible implementation, the method for correcting the initial weight parameters used in video coding based on importance weights to obtain corrected target weight parameters specifically includes: constructing a weight mapping matrix based on the grayscale value and frequency of each pixel in the video surveillance image; the elements in the weight mapping matrix are the initial weight parameters; adjusting the initial weight values ​​of pixels in the corresponding region in the weight mapping matrix according to the importance weights to obtain the corrected weight mapping matrix; normalizing the corrected weight mapping matrix to obtain the target weight mapping matrix; the elements in the target weight mapping matrix are the target weight parameters.

[0011] In conjunction with the first aspect above, in one possible implementation, the method of using target weight parameters to perform Huffman coding compression on video surveillance images to generate compressed video data specifically includes: using the weights of each element in the target weight mapping matrix as the coding weights for Huffman coding compression to encode and compress the video surveillance images; and encapsulating the encoded and compressed video data into a standardized network transmission protocol format to obtain compressed video data.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, the method for dividing a video surveillance image into regions specifically includes: using a preset region growth segmentation algorithm, selecting the pixel with the largest gray value in the video surveillance image as the growth starting point, and using the difference between the gray value of the target pixel and the growth starting point being less than a preset difference as the growth criterion, until the video surveillance image is divided into multiple independent regions.

[0013] In conjunction with the first aspect above, in one possible implementation, the method for acquiring video surveillance images from underground coal mines specifically includes: acquiring raw surveillance video from underground coal mines; preprocessing the images in the raw surveillance video to obtain video surveillance images; the preprocessing includes image enhancement and grayscale processing.

[0014] Secondly, a video surveillance system for underground coal mines is provided, comprising: an image acquisition module, a weight correction module, and a video compression module; the image acquisition module is used to acquire video surveillance images from underground coal mines and preprocess the video surveillance images; the preprocessing includes image enhancement and grayscale processing; the weight correction module is used to divide the video surveillance images into regions and determine the importance weights of multiple regions based on image features and motion features within the regions; the weight correction module is also used to correct the initial weight parameters used in video coding according to the importance weights to obtain the corrected target weight parameters; the initial weight parameters are determined based on the frequency of grayscale values ​​in the image; the video compression module is used to perform Huffman coding compression on the video surveillance images using the target weight parameters to generate compressed video data for transmission to the ground monitoring center.

[0015] Thirdly, a video surveillance device for underground coal mines is provided, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is used to execute the instructions to perform the actions described in the first aspect and any possible implementation thereof. This underground video surveillance device for coal mines can be an electronic device or a chip within an electronic device.

[0016] Fourthly, a computer-readable storage medium is provided, which stores instructions that, when executed on a video surveillance device in a coal mine, cause the video surveillance device in the coal mine to perform the actions described in the first aspect and any possible implementation thereof.

[0017] Fifthly, a computer program product containing instructions is provided, which, when run on a video monitoring device in a coal mine, causes the video monitoring device in the coal mine to perform the actions described in the first aspect and any possible implementation thereof.

[0018] The present invention has the following beneficial effects: By acquiring underground video surveillance images from coal mines and dividing them into regions, importance weights are determined based on the image and motion characteristics of each region. The initial weight parameters, which rely solely on grayscale frequency, are then corrected. Finally, the corrected target weight parameters are used for Huffman coding compression. This effectively improves the image clarity of safety-critical areas (such as areas with foreign objects or abnormal movement) in the video while ensuring overall compression efficiency. It avoids the blurring problem caused by traditional methods ignoring important low-frequency information, ensuring that the ground monitoring center can accurately identify safety hazards. Attached Figure Description

[0019] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A system structure diagram of a video surveillance system for underground coal mines, provided in one embodiment of the present invention; Figure 2 A flowchart illustrating a video surveillance method for underground coal mines, as provided in one embodiment of the present invention; Figure 3 A flowchart illustrating another video surveillance method for underground coal mines, provided in one embodiment of the present invention; Figure 4A flowchart illustrating another video surveillance method for underground coal mines, provided in one embodiment of the present invention; Figure 5 A flowchart illustrating another video surveillance method for underground coal mines, provided in one embodiment of the present invention; Figure 6 This is a schematic diagram of the hardware structure of a video monitoring device for underground coal mines, provided as an embodiment of the present invention. Detailed Implementation

[0021] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a video monitoring method and system for underground coal mines proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0023] The following description, in conjunction with the accompanying drawings, details the specific scheme of a video monitoring method and system for underground coal mines provided by this invention.

[0024] Please see Figure 1 The diagram illustrates a system structure of a video surveillance system for underground coal mines according to an embodiment of the present invention. The video surveillance system for underground coal mines includes: an image acquisition module 1, a weight correction module 2, and a video compression module 3.

[0025] Among them, the image acquisition module 1 is responsible for acquiring and preprocessing video surveillance images in the coal mine, providing high-quality processing objects for subsequent modules.

[0026] In some implementations, the image acquisition module 1 includes two sub-modules: an acquisition sub-module 11 and a preprocessing sub-module 12. The acquisition submodule 11 acquires raw surveillance video by deploying monitoring cameras at preset locations underground in the coal mine. During deployment, it is essential to ensure that the center of the camera lens is aligned with the centerline of the target to be monitored (such as a coal mine conveyor belt) to achieve full coverage of the target area. Simultaneously, the acquisition frequency of the monitoring cameras is set to a preset frequency to ensure the continuity of dynamic images, and the camera's imaging parameters must meet the requirements for clear imaging of details such as foreign objects and cracks to avoid subsequent feature recognition errors due to blurred images.

[0027] The preprocessing submodule 12 receives the raw monitoring video output by the acquisition submodule 11 and performs preprocessing operations on each frame of the image. First, a preset image enhancement algorithm is used to remove noise caused by underground dust and uneven lighting in the raw monitoring video, improving the clarity of image details and overall image quality. Then, the denoised image is converted to grayscale, and finally, a video monitoring image that meets the processing requirements of subsequent modules is output and transmitted to the weight correction module 2.

[0028] The weight correction module 2 receives the video surveillance image output by the image acquisition module 1, and generates target weight parameters adapted to Huffman coding through region division, importance weight calculation and initial weight correction, providing key coding basis for the video compression module 3.

[0029] In some implementations, the weight correction module 2 includes three sub-modules: a region partitioning sub-module 21, an importance weight calculation sub-module 22, and a weight correction sub-module 23. The region segmentation submodule 21 receives the video surveillance image output by the image acquisition module 1, and uses a preset region growth segmentation algorithm to segment the video surveillance image into regions. Specifically, the pixel with the largest gray value in the video surveillance image is used as the growth starting point, and the difference between the gray value of the target pixel and the growth starting point is less than a preset difference is used as the growth criterion, until the video surveillance image is divided into multiple independent regions. After the segmentation is completed, the region segmentation result is transmitted to the importance weight calculation submodule 22.

[0030] The importance weight calculation submodule 22 calculates the foreign object degree of each region based on the region division results output by the region division submodule 21. Then, it calculates the safety hazard index of each region based on the image features of each region. Finally, it calculates the importance weight of each region by combining the foreign object degree and the safety hazard index. After the calculation is completed, the importance weight is transmitted to the weight correction submodule 23.

[0031] The weight correction submodule 23 first obtains the frequency of occurrence of each gray value in the video surveillance image, constructs an initial weight mapping matrix based on the gray value of each pixel and its frequency of occurrence (the initial weight parameters are determined based on this matrix), and then combines the importance weight output by the importance weight calculation submodule 22 to adjust the initial weight value of the corresponding region pixels in the initial weight mapping matrix to obtain the corrected weight mapping matrix. Finally, the corrected weight mapping matrix is ​​normalized to generate the target weight parameters and transmit them to the video compression module 3.

[0032] Based on the target weight parameters output by the weight correction module 2, the video compression module 3 completes the encoding compression and data encapsulation of the video surveillance image, ensuring that the compressed video data can be stably transmitted to the ground monitoring center.

[0033] In some implementations, the video compression module 3 comprises two sub-modules: an encoding sub-module 31 and an encapsulation and transmission sub-module 32. The encoding submodule 31 receives the video surveillance image output by the image acquisition module 1 and the target weight parameters output by the weight correction module 2. It uses the weights of each element in the target weight mapping matrix corresponding to the target weight parameters as the encoding weights of Huffman coding, performs Huffman coding compression on the video surveillance image, generates preliminary compressed video data, and transmits it to the encapsulation and transmission submodule 32.

[0034] The encapsulation and transmission submodule 32 receives the preliminary compressed video data output by the encoding submodule 31, encapsulates it into a preset standardized network transmission protocol format (adapting to the data transmission needs of underground coal mines, ensuring transmission stability and compatibility), and generates the final compressed video data. Subsequently, the compressed video data is transmitted to the ground monitoring center via a preset transmission medium (such as gigabit fiber optic) in the underground coal mine, for decompression by ground personnel for safety hazard detection and analysis.

[0035] Please see Figure 2 The diagram illustrates a flowchart of a video surveillance method for underground coal mines according to an embodiment of the present invention. The video surveillance method for underground coal mines includes: S1. Acquire video surveillance images from underground coal mines.

[0036] In some implementations, methods for acquiring video surveillance images from underground coal mines may specifically include: acquiring raw surveillance video from underground coal mines; then preprocessing the images in the raw surveillance video to obtain video surveillance images. This preprocessing includes at least image enhancement and grayscale conversion.

[0037] Specifically, in the process of acquiring raw surveillance video, the installation points of the surveillance cameras must first be determined based on the actual location and size of the target to be monitored underground (such as a coal mine conveyor belt). Typically, a preset position with no obstructions is selected directly above or to the side of the target, ensuring that the center of the camera lens is precisely aligned with the centerline of the target. This achieves full coverage of the target area and avoids blind spots caused by lens misalignment. Simultaneously, the camera's acquisition frequency must be set to a preset value that meets the continuity of dynamic images (such as an acquisition frequency adapted to the operating speed of the underground conveyor belt) to prevent incomplete capture of dynamic features such as moving objects and belt misalignment due to a low frame rate. Furthermore, the camera's imaging parameters (such as resolution, focal length, and sensitivity) need to be adjusted to ensure clear imaging of details such as foreign objects and cracks on the surface of the target. This avoids deviations in subsequent foreign object detection calculations due to blurred images. Finally, this step yields raw surveillance video containing complete dynamic information of the target.

[0038] Furthermore, to address the noise issues in the original monitoring video images caused by the special environment of underground coal mines (high dust levels, uneven lighting), and the image format requirements for subsequent area division and foreign object density calculation, the processing is completed in two steps: The first step is image enhancement processing. Preset image enhancement algorithms (such as those used to remove noise and smooth lighting differences) are used to process each frame of the original surveillance video. This process enhances the grayscale contrast between foreign objects and the background (such as coal mines), suppresses the graininess caused by dust, removes noise interference with image details, and ensures that key features such as the edges and surface textures of foreign objects are clearly distinguishable.

[0039] The second step is grayscale processing, which converts the enhanced color image into a grayscale image. By simplifying the color dimension of the image, the computational workload of subsequent region segmentation, grayscale value statistics and gradient calculation is reduced. At the same time, it ensures that the grayscale difference features between foreign objects and the background in the image are preserved. For example, the low grayscale value of the coal mine area and the high grayscale value of the foreign object area will not be lost due to grayscale processing.

[0040] It should be noted that when using conveyor belts to transport coal underground, foreign objects such as gangue are often mixed in. These objects, due to their rough or sharp surfaces, may tear the conveyor belt during transport, reducing its lifespan and posing a safety hazard. Therefore, these targets should be given special attention in video surveillance. This requires that the area containing foreign objects remain clear in the compressed video image transmitted to the ground monitoring center for accurate detection. Coal is typically black with a smooth, glossy surface and relatively low grayscale values ​​with uniform gradients. In contrast, foreign objects usually exhibit significant color and brightness differences from coal. For example, gangue is typically gray with a rough surface, generally having higher grayscale values, uneven grayscale distribution, and larger gradients. Therefore, foreign objects can be identified based on the grayscale characteristics in the image.

[0041] S2. Divide the video surveillance image into regions and determine the importance weights of multiple regions based on image features and motion features within each region.

[0042] In some implementations, the method for dividing video surveillance images into regions may specifically include: employing a preset region growth segmentation algorithm, based on the characteristics of underground coal mine scenes: the grayscale value of foreign objects (such as gangue) on the monitored target (such as a conveyor belt) is significantly higher than that of the coal mine background (coal mines are black with low grayscale values), and the pixel with the highest grayscale value is likely to belong to the edge or core area of ​​the foreign object. Therefore, the pixel with the highest grayscale value in the video surveillance image is selected as the growth starting point. Then, the growth criterion is that the difference in grayscale value between the target pixel and the growth starting point is less than a preset difference (this difference needs to be adjusted based on the grayscale distribution characteristics of the underground image to ensure the consistency of grayscale features within the same area). The target pixel is then included in the current growth region, and the growth process is repeated until the video surveillance image is divided into multiple independent regions.

[0043] In some implementations, methods for determining the importance weights of multiple regions based on image and motion features within the region may specifically include: calculating a foreign object degree, representing the likelihood of a region containing foreign objects, based on image features within the region; then calculating a safety hazard index, representing the degree of safety risk in the region, based on motion features; and finally, fusing the foreign object degree and the safety hazard index to calculate the importance weight of each region.

[0044] In other implementations, candidate foreign object regions with high grayscale values ​​and sufficient area can be screened out first through global grayscale histogram analysis, excluding background regions. Then, local texture and edge features of the candidate regions are extracted to calculate the foreign object confidence. The motion anomaly degree is calculated by combining the difference between the overall motion vector of the region and the normal motion vector. The two are weighted and fused to obtain the importance weight, and the background region is directly assigned a low weight.

[0045] In other implementations, the mean and variance of the grayscale of the local neighborhood of each pixel can be calculated to form an image feature vector, and the inter-frame displacement distance and direction can be used to form a motion feature vector. Then, a preset clustering algorithm is used to cluster pixels with similar features into different risk regions. The mean of the pixel features in the region is statistically analyzed to obtain the foreign object coefficient and risk coefficient of the region. The product is fused to obtain the weight, thus skipping the pre-region division step.

[0046] In other implementations, the image can be divided into fine, medium, and coarse multi-scale regions first. The dual features of each region are calculated (fine scale focuses on edge and small object displacement, coarse scale focuses on grayscale trend and large area movement) and assigned scale weights. The weighted average of the dual features of the spatially overlapping regions is used to obtain the fusion value, and the corresponding scale features are used for non-overlapping regions. Finally, the weights are calculated based on the fusion value, taking into account both small objects and large area anomaly features.

[0047] S3. Correct the initial weight parameters used in video coding according to the importance weights to obtain the corrected target weight parameters.

[0048] The initial weight parameters are determined based on the frequency of grayscale values ​​in the image.

[0049] In some implementations, the frequency of each grayscale value in the video surveillance image can be counted first. The grayscale values ​​and their corresponding frequencies are then paired according to pixel row and column order to construct an initial weight mapping matrix (initial weight parameters). Next, based on the importance weights output by S2, the position of each region in the matrix is ​​located, and the matrix elements are adjusted. Finally, a preset normalization algorithm maps the weights in the initial correction matrix to the 0-1 interval, obtaining the target weight mapping matrix (target weight parameters) that meets the requirements of Huffman coding, thus achieving pixel-level directional correction and globally standardized weight optimization.

[0050] In other implementations, the image can be divided into four categories: foreign object core region, foreign object edge region, background transition region, and pure background region. Initial weights (corresponding to the average frequency of grayscale values) are calculated for each region type, and differentiated correction coefficients are designed for different region types (the coefficient is highest for the foreign object core region and 1 for the pure background region). The initial weights are multiplied by the corresponding coefficients to obtain the layered correction weights, which are then normalized to form the target weight parameters grouped by region type. This approach is suitable for scenarios in coal mines where the size of foreign objects (such as conveyor belts) varies greatly and precise preservation of edge details is required. It avoids the problem of insufficient or excessive correction of edge region weights caused by a single correction rule, and is particularly suitable for monitoring needs where foreign objects easily get stuck on the edge of the conveyor belt, causing safety hazards.

[0051] S4. Use the target weight parameters to perform Huffman coding compression on the video surveillance images to generate compressed video data for transmission to the ground monitoring center.

[0052] In some implementations, the weights of each element in the target weight mapping matrix are first used as the encoding weights for Huffman coding compression to encode and compress the video surveillance image. Then, the encoded and compressed video data is encapsulated into a standardized network transmission protocol format, resulting in compressed video data.

[0053] Specifically, the core of Huffman coding is to assign prefix codes according to weights. Based on the target weight mapping matrix (target weight parameters), it assigns differentiated coding lengths to pixels with different weights, ensuring the coding priority of high-importance regions. This includes: Retrieve the target weight mapping matrix. The coordinates of each element in this matrix correspond one-to-one with the pixel coordinates of the video surveillance image, and the element value is the encoding weight of the corresponding pixel (pixels in high-importance areas have higher weights than those in low-importance areas). Associate the weights of all pixels in the matrix with their corresponding grayscale values ​​to obtain data pairs of grayscale values ​​and weights, clarifying the weight differences of each grayscale value in different regions (for example, the same grayscale value has a higher weight in foreign object areas than in background areas).

[0054] Then, all grayscale value and weight data pairs are sorted from largest to smallest weight, with higher-weight data pairs participating in the coding tree construction first. During coding tree construction, shorter binary prefix codes are assigned to high-weight data pairs, and longer prefix codes are assigned to low-weight data pairs. For example, the grayscale value corresponding to a high-weight pixel in the foreign object region may be assigned a 4-bit prefix code, while the grayscale value corresponding to a low-weight pixel in the background may be assigned an 8-bit prefix code. This difference in encoding length ensures lower data redundancy in high-importance regions and more complete preservation of details after decompression.

[0055] Next, each pixel of each frame of the video surveillance image is encoded according to the frame order (from the first frame to the last frame) to form a single frame image encoded data stream. This process is repeated until all frames are encoded to obtain a continuous preliminary compressed video data stream. This data stream has achieved the goal of accurately preserving high-importance areas and efficiently compressing low-importance areas.

[0056] Finally, the initially compressed data stream is encapsulated into a standardized protocol format (such as real-time transport protocol (RTP) or real-time transport control protocol (RTCP)) to ensure stability and compatibility in the underground coal mine transmission environment and to ensure that the ground monitoring center can decompress it normally.

[0057] Based on the above technical solution, by acquiring underground video surveillance images of coal mines and dividing them into regions, the importance weights are determined based on the image features and motion features of each region. Then, the initial weight parameters that only rely on the grayscale frequency are corrected. Finally, the corrected target weight parameters are used for Huffman coding compression. This effectively improves the image clarity of safety-critical areas (such as foreign objects or abnormally moving areas) in the video while ensuring overall compression efficiency. It avoids the blurring problem caused by traditional methods ignoring important low-frequency information, ensuring that the ground monitoring center can accurately identify safety hazards.

[0058] In one possible implementation, combining Figure 2 ,like Figure 3 As shown, if multiple regions encompass the entire area of ​​the video surveillance image, the method in S2 above for determining the importance weights of multiple regions based on image features and motion features within the regions can be specifically implemented through the following S21 to S23, which are explained in detail below: S21. Determine the foreign object degree of each region based on the image features within each region.

[0059] Foreign matter content is used to characterize the likelihood that a region contains foreign matter.

[0060] In some implementations, image features include grayscale distribution and gradient changes. In this case, the method for determining the foreign object degree of each region based on the image features within each region can specifically include: determining the foreign object degree based on the grayscale value, the number of pixels, and the gradient difference in grayscale values ​​between pixels within each region, denoted as: In the formula, denoted as the standard deviation of gray values ​​in the i-th region, reflecting the degree of dispersion of gray value distribution within the region. The larger the value, the more drastic the fluctuation of gray values ​​within the region (e.g., rough surface of foreign objects, uneven reflection), and the higher the probability of foreign objects.

[0061] This represents the average grayscale value within the i-th region, reflecting the overall grayscale level of the region. Since a region is composed of actual pixels, there are no completely black regions with no pixels. It must not be 0.

[0062] This reflects the relative gray-level dispersion within the i-th region. If the standard deviation of gray levels in two regions... Same, average gray level Smaller areas (such as high-grayscale foreign objects appearing in a low-grayscale background) have a higher degree of dispersion and a greater likelihood of foreign objects.

[0063] This represents the total number of pixels contained in the i-th region. A region generated by the region growing segmentation algorithm must contain at least a certain number of pixels; otherwise, it cannot form a valid region. .

[0064] , represents the grayscale gradient magnitude of the k-th and (k+1)-th pixels within the region, respectively. The grayscale gradient magnitude can be calculated using the Sobel operator. The pixel indices are assigned according to the order in which pixels were added during the region growing process; that is, the order of pixels k and k+1 is based on the pixel addition order generated by the region growing segmentation algorithm.

[0065] This represents the sum of gradient magnitude differences, indicating the total intensity of gradient changes and reflecting the gradient changes at the edges or in the texture of a pixel. The more uneven the grayscale within a region and the greater the difference in gradient magnitude, the larger the sum, indicating that the edges and textures within the region are more complex (such as the irregular shape or rough surface of a foreign object), and the higher the likelihood of a foreign object.

[0066] Divide by This can eliminate the influence of region size on the gradient sum. If two regions have the same gradient characteristics, the gradient sum of the larger region will be larger due to the greater number of pixels. Dividing by... Normalization ensures that the gradient contributions of regions of different sizes are comparable.

[0067] Finally, by combining the relative dispersion of gray levels and the total intensity of gradient changes, the probability of a region containing foreign objects is quantified, and the foreign object degree of the i-th region is obtained. The higher the value, the less it conforms to the characteristics of a smooth and shiny coal mine surface, indicating a greater likelihood that the area contains foreign objects.

[0068] In other implementations, texture templates of common foreign objects (such as gangue and metal fragments) in coal mines can be pre-stored, the texture features of the area to be detected can be extracted and compared with the template library, and the Euclidean distance can be calculated to obtain the matching similarity. If the similarity is greater than or equal to the preset threshold, the similarity is used as the foreign object degree; otherwise, it is 0. This is suitable for scenarios where the foreign object type is fixed and the texture features are significant.

[0069] In other implementations, a foreign object recognition model based on a lightweight convolutional neural network can be built and trained. The model is input into a sub-block of the region image, and it automatically extracts deep features and outputs the probability that the region contains foreign objects. This probability is directly used as the foreign object degree, which is suitable for complex scenarios where foreign objects have diverse shapes and are difficult to cover with manual features.

[0070] S22. Determine the safety hazard index for each area based on the movement characteristics of each area.

[0071] Among them, the safety hazard index is used to characterize the degree of safety risk in a region.

[0072] In some implementations, motion features include the displacement of pixels in consecutive frames. This can be achieved by obtaining the coordinate set of pixels in each region of the image, taking the current frame image, the coordinate set of pixels in the region, and the previous frame image as input, and using optical flow to obtain the displacement distance and direction of the pixels.

[0073] At this point, the method for determining the safety hazard index based on motion characteristics may specifically include: determining the safety hazard index based on the displacement distance of pixels within a region in consecutive frames, and the similarity between the displacement direction and a preset direction, denoted as: In the formula, This indicates the displacement direction of the k-th pixel within the j-th region. This indicates the pre-calibrated direction of belt travel for the conveyor belt.

[0074] The cosine similarity is used to measure the degree of deviation between the pixel displacement direction and the normal operating direction of the belt. The closer the value is to -1 (the direction is completely opposite), the more serious the offset is; the closer the value is to 1, the more the direction matches the normal operating trajectory.

[0075] This represents the displacement distance of the k-th pixel within the j-th region.

[0076] This is a fusion term of displacement amplitude and direction. The total risk contribution of the region is obtained by summing the amplitude and orientation fusion terms of all pixels within the region.

[0077] This represents the number of pixels in the j-th region.

[0078] Total risk contribution divided by The total risk contribution of the region can be averaged to each pixel, eliminating the impact of region size on risk assessment and making foreign object regions of different sizes comparable.

[0079] Finally, by representing the hyperbolic tangent function By compressing the average risk contribution range to (-1, 1), the safety hazard index for the j-th region is obtained. .when When the value is close to 1, it indicates that the displacement distance of the foreign object is large and its direction is highly consistent with the normal operating direction of the belt (such as a foreign object moving at high speed along the belt's running direction), posing an extremely high risk of potential scratches and jamming; when When the value is close to -1, it indicates that the displacement distance of the foreign object is large and its direction deviates significantly from the normal operating direction of the conveyor belt (such as foreign objects moving in the opposite direction or shifting laterally), posing an extremely high potential risk of falling or impact; when When the value is close to 0, it indicates that the displacement distance or directional deviation of the foreign object is small, and the potential safety hazard is low.

[0080] In other implementations, the average motion speed (converted to actual speed) and average acceleration of pixels within a region can be calculated from consecutive frame images. Weights (e.g., 0.6) for speed and (e.g., 0.4) for acceleration are assigned and superimposed to obtain a basic risk value. This value is then normalized to [0, 1] by combining it with a safety threshold as a safety hazard index. The faster the speed and the more drastic the acceleration change, the closer the index is to 1. This is suitable for assessing the risk of foreign object impact on high-speed belt conveyors.

[0081] In other implementations, a normal trajectory model of the belt conveyor can be preset, the center pixel trajectory of the tracking area is continuously framed, the average deviation distance and the deviation rate (the proportion of times the deviation exceeds the allowable range) from the trajectory point to the normal trajectory can be calculated, and the safety hazard index can be obtained by weighting the normalized average deviation distance and the deviation rate. The larger the deviation and the more frequent the deviation, the higher the index. This is suitable for scenarios where foreign objects at the turning point of the belt conveyor are deviating from the normal transmission path.

[0082] S23. Determine the importance weight of each area based on the foreign object degree and safety hazard index.

[0083] Areas with foreign objects and significant safety hazards are key areas of concern. To avoid neglecting the importance of these areas during video compression, resulting in lower image quality and affecting the detection results of conveyor belt anomalies, and causing huge losses due to the inability to detect areas with significant safety hazards in a timely manner, these areas need to be given greater weight.

[0084] In some implementations, the importance weight of each area is determined based on the foreign object degree and the safety hazard index, denoted as: In the formula, This represents the weight adjustment factor, which limits the overall increase in weight. Its value range is (0, 1). No special restrictions are imposed here, but 0.4 is preferred to prevent excessive importance weights from causing other areas to become blurred (such as the coal mine area becoming blurred due to excessive weight compression).

[0085] This represents the normalized value of the foreign object degree in the j-th region. The larger the value, the higher the probability that the region is a foreign object, and the higher the weight. The foreign object degree is normalized to the interval [0, 1] for example by linear min-max normalization.

[0086] This represents the safety hazard index of the j-th region. The closer the value is to 1, the greater the security risk.

[0087] The absolute value of the safety hazard index is non-linearly amplified, and through the exponential growth characteristic, the weight of foreign object areas with high safety hazards is increased exponentially.

[0088] This represents the importance weight of the j-th region. The three factors are multiplied together to give the weight of regions with high foreign object degree and high safety hazard. Significantly increased, enabling precise weighting of high-risk foreign object areas.

[0089] Based on the above technical solution, the method first determines the "foreign object degree" (representing the likelihood of a region containing foreign objects) based on image features within the region. Then, it combines the region's motion characteristics to determine the "safety hazard index" (representing the degree of safety risk in the region). Finally, it integrates the two to determine the importance weight of each region. This method can accurately identify key areas in underground coal mine video surveillance that contain both foreign objects and pose high safety risks. It provides a basis for subsequently adjusting video encoding weights to ensure clear details in compressed images of such areas. This effectively solves the problem of traditional encoding relying solely on grayscale frequency to allocate weights while ignoring important low-frequency safety information, thus improving the accuracy and timeliness of identifying safety hazards in underground coal mines.

[0090] In one possible implementation, combining Figure 2,like Figure 4 As shown, if multiple regions only include foreign object regions in the video surveillance image, the method in S2 above for determining the importance weights of multiple regions based on image features and motion features within the regions can be specifically implemented through the following S24 to S26, which are explained in detail below: S24. Determine the foreign object degree of each region based on the image features within each region, and filter foreign object regions based on the foreign object degree.

[0091] Foreign matter content is used to characterize the likelihood that a region contains foreign matter.

[0092] In some implementations, the method for determining the foreign object degree of each region based on the image features within each region can be found in the description in S21, and will not be repeated here.

[0093] Furthermore, coal mine conveyor belts typically vibrate during operation. Within a certain range, the belt will not shift. However, if the coal transported on the belt is unevenly distributed or has an uneven weight, it may cause uneven stress on the belt, leading to increased belt vibration and belt shift. Conversely, the belt shift will further increase the uneven stress, exacerbating belt vibration and potentially causing coal and foreign objects on the belt to vibrate or even slip and move, increasing the likelihood of belt scratches. In this case, it is even more important to ensure that high-risk foreign object areas are clearly visible during video compression so that the ground monitoring center can accurately identify foreign objects.

[0094] In addition, it can reduce the computation of non-foreign object areas, save computing resources, and improve image processing efficiency.

[0095] In some implementations, the Otsu's threshold can be calculated using the Otsu's inter-class variance algorithm. This involves iterating through possible candidate threshold values, dividing the dataset into two classes for each candidate value: "out-of-class variance ≥ candidate value" and "out-of-class variance < candidate value," and calculating the variance between the two classes (inter-class variance). When the inter-class variance reaches its maximum value, the corresponding candidate value is the out-of-class threshold.

[0096] Finally, based on the threshold, areas with a foreign object degree greater than or equal to the threshold are identified as foreign object areas (these areas are highly likely to contain foreign objects and are the focus of subsequent safety hazard analysis), while areas with a foreign object degree less than the threshold are identified as non-foreign object areas (these areas are mainly coal mines with low safety risks).

[0097] S25. Determine the safety hazard index of the foreign object area based on the movement characteristics of the foreign object area.

[0098] Among them, the safety hazard index is used to characterize the degree of safety risk in a region.

[0099] In some implementations, the method for determining the safety hazard index based on motion characteristics can be found in the description in S22, which will not be repeated here.

[0100] S26. Determine the importance weight of the foreign object area based on the foreign object degree and safety hazard index.

[0101] In some implementations, the method for determining the importance weight of the foreign object area based on the foreign object degree and the safety hazard index can be found in the description in S23, which will not be repeated here.

[0102] In addition, the importance weight of non-foreign object regions is 0.

[0103] Based on the above technical solution, the foreign object degree, representing the probability of a region containing foreign objects, is determined by analyzing image features within the region, and foreign object regions are screened out. Then, a safety hazard index, representing the degree of safety risk, is determined for each foreign object region based on its motion characteristics. Finally, the importance weight of the foreign object region is determined by merging the foreign object degree and the safety hazard index. This method can accurately identify key areas in underground coal mine video monitoring scenarios that contain both foreign objects and pose high safety risks. It provides a precise basis for adjusting video encoding weights to ensure clear details in compressed images of such areas, effectively solving the problem of traditional encoding relying solely on grayscale frequency to allocate weights while ignoring important low-frequency safety information. This improves the accuracy of identifying safety hazards in underground coal mines and contributes to efficient monitoring of safe production in underground coal mines.

[0104] In one possible implementation, combining Figure 2 ,like Figure 5 As shown, the method in S3 above can be specifically implemented through the following steps S31 to S33, which are explained in detail below: S31. Construct a weight mapping matrix based on the grayscale value of each pixel in the video surveillance image and its frequency of occurrence.

[0105] Specifically, iterate through all pixels of the image, record the grayscale value (0-255) of each pixel, and record the data in the m-th row as follows: ,in This represents the second grayscale value in the m-th row.

[0106] Count the number of times each grayscale value appears, and record it as . Where q represents the total number of different gray values ​​appearing in the image, that is, each gray value in the grayscale image corresponds to a frequency, for example, the gray value in the m-th row and n-th column. The frequency of occurrence is .

[0107] Will and Matching is used to form a data pair, denoted as Similarly, all grayscale values ​​are matched with their corresponding frequencies to obtain data pairs. .

[0108] The frequency in each data pair is used as the initial weight for each grayscale value. A weight mapping matrix B is constructed for all data pairs according to their rows and columns. It is located at the position in the m-th row and n-th column of the matrix.

[0109] S32. Adjust the initial weight values ​​of the pixels in the corresponding region in the weight mapping matrix according to the importance weight to obtain the corrected weight mapping matrix.

[0110] In some implementations, the initial weight values ​​can be corrected as follows: In the formula, This represents the importance weight of the j-th region.

[0111] This represents the frequency of the b-th data pair corresponding to the j-th region in matrix B, i.e., the initial weight value.

[0112] This represents the corrected weight, or target weight value, of the b-th data pair within the j-th region of matrix B.

[0113] Update the weights in matrix B according to the target weight values ​​to obtain the weight mapping matrix. .

[0114] S33. Normalize the corrected weight mapping matrix to obtain the target weight mapping matrix.

[0115] Since the sum of the weights of all grayscale values ​​after the update is greater than 1, the weights need to be normalized. Specifically, the statistical matrix... The data pairs are normalized by dividing each weight value by the sum of all weight values. The normalized weight values ​​are then reassigned to the matrix. From the corresponding data pairs, the target weight mapping matrix is ​​obtained. The weight of the foreign object area was originally small, but after the weight was adjusted, the weight of the foreign object area increased.

[0116] Based on the above technical solution, a weight mapping matrix is ​​first constructed according to the grayscale value and frequency of each pixel in the video surveillance image. Then, the initial weight values ​​of the corresponding pixels in the region are adjusted by combining the weights of regional importance to obtain the corrected weight mapping matrix. Finally, the matrix is ​​normalized to obtain the target weight mapping matrix. This effectively solves the problem that traditional Huffman coding only relies on the frequency of grayscale values ​​to assign weights and ignores important low-frequency safety information. This allows key areas containing foreign objects and high safety risks in underground coal mine video surveillance to receive higher weights during encoding and compression to preserve details, while also taking into account video compression efficiency. As a result, the ground monitoring center can clearly identify areas with safety hazards after receiving and decompressing the video, providing strong support for underground coal mine safety production monitoring.

[0117] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0118] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0119] In this embodiment of the invention, the video monitoring device in an underground coal mine can be divided into functional units according to the above-described method example. For example, each function can be divided into its own functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this embodiment is illustrative and represents only one logical functional division; in actual implementation, other division methods may be used.

[0120] This invention also provides a hardware structure diagram of a video surveillance device for underground coal mines, see below. Figure 6 The underground video monitoring device 600 in the coal mine includes a processor 601, and optionally, a memory 602 connected to the processor 601.

[0121] In the first possible implementation, see Figure 6The underground video surveillance device 600 also includes a transceiver 603. The processor 601, memory 602, and transceiver 603 are connected via a bus. The transceiver 603 is used to communicate with other devices or communication networks. Optionally, the transceiver 603 may include a transmitter and a receiver. The device in the transceiver 603 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of the present invention. The device in the transceiver 603 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of the present invention.

[0122] Based on the first possible implementation method Figure 6 The structural diagram shown can be used to illustrate the structure of the video monitoring device for underground coal mines involved in the above embodiments.

[0123] in, Figure 6 This can also be illustrated by the system chip in the video surveillance device located underground in a coal mine. In this case, the actions performed by the aforementioned video surveillance device underground in a coal mine can be implemented by this system chip. The specific actions performed can be found above and will not be repeated here.

[0124] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in this embodiment can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0125] The processor in this invention may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., which are various computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a standalone semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may be integrated with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a System-on-a-Chip (SoC), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), programmable logic devices (PLDs), or logic circuits that implement dedicated logic operations.

[0126] The memory in the embodiments of the present invention may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable read-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0127] This invention also provides a computer-readable storage medium including instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0128] This invention also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0129] This invention also provides a chip, which includes a processor and an interface circuit. The interface circuit is coupled to the processor. The processor is used to run computer programs or instructions to implement the above-described method. The interface circuit is used to communicate with other modules outside the chip.

[0130] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0131] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings and the disclosure, will understand and implement other variations of the disclosed embodiments in carrying out the claimed invention. In this invention, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several of the functions listed in this invention.

[0132] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the invention and its equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A video surveillance method for underground coal mines, characterized in that, include: Acquire video surveillance images from underground coal mines; The video surveillance image is divided into regions, and importance weights for multiple regions are determined based on image features and motion features within each region. Specifically, this includes: determining the foreign object level of each region based on its grayscale value, number of pixels, and grayscale gradient differences between pixels; determining a security hazard index for each region based on the displacement distance of pixels within the region in consecutive frames and the similarity of the displacement direction to a preset direction; and determining the importance weight of each region based on the foreign object level and the security hazard index. The multiple regions include all regions of the video surveillance image. The foreign object level characterizes the probability of a region containing foreign objects, and the security hazard index characterizes the degree of security risk of the region. The quantification formula for determining the foreign matter content of each region is as follows: In the formula, Let the foreign object degree be the degree of the i-th region. Let be the standard deviation of the gray values ​​in the i-th region. This represents the average gray value within the i-th region. This represents the total number of pixels contained in the i-th region. , These represent the grayscale gradient magnitudes of the k-th and (k+1)-th pixels within the region, respectively. The formula for determining the safety hazard index for each region is as follows: In the formula, Let the safety hazard index be the value of the j-th region. This indicates the displacement direction of the k-th pixel within the j-th region. This indicates the pre-calibrated belt running direction of the conveyor belt. Represents cosine similarity. This represents the displacement distance of the k-th pixel within the j-th region. This represents the number of pixels in the j-th region; It is the hyperbolic tangent function; The quantification formula for determining the importance weight of each region is as follows: In the formula, This represents the importance weight of the j-th region. Indicates the weight adjustment factor. This represents the normalized value of the foreign matter degree in the j-th region; The initial weight parameters used in video coding are corrected according to the importance weights to obtain the corrected target weight parameters; the initial weight parameters are determined based on the frequency of grayscale values ​​in the image. The target weight parameters are used to perform Huffman coding compression on the video surveillance image to generate compressed video data for transmission to the ground monitoring center.

2. The video surveillance method according to claim 1, characterized in that, The multiple regions only include foreign object regions in the video surveillance image; the determination of importance weights for the multiple regions based on image features and motion features within the regions includes: The foreign object degree of each region is determined based on the image features within each region, and foreign object regions are filtered based on the foreign object degree; the foreign object degree is used to characterize the probability that a region contains foreign objects. The safety hazard index of the foreign object region is determined based on its motion characteristics; the safety hazard index is used to characterize the degree of safety risk in the region. The importance weight of the foreign object area is determined based on the foreign object degree and the safety hazard index.

3. The video surveillance method according to claim 1, characterized in that, The step of correcting the initial weight parameters used in video coding according to the importance weights to obtain the corrected target weight parameters includes: A weight mapping matrix is ​​constructed based on the grayscale value and frequency of each pixel in the video surveillance image; the elements in the weight mapping matrix are the initial weight parameters. The initial weight values ​​of the pixels in the corresponding region in the weight mapping matrix are adjusted according to the importance weights to obtain the corrected weight mapping matrix; The modified weight mapping matrix is ​​normalized to obtain the target weight mapping matrix; the elements in the target weight mapping matrix are the target weight parameters.

4. The video surveillance method according to claim 3, characterized in that, The step of using the target weight parameters to perform Huffman coding compression on the video surveillance image to generate compressed video data includes: The weights of each element in the target weight mapping matrix are used as the encoding weights for Huffman coding compression to encode and compress the video surveillance image. The encoded and compressed video data is encapsulated into a standardized network transmission protocol format, and the resulting compressed video data is obtained after encapsulation.

5. The video surveillance method according to claim 1, characterized in that, The process of dividing video surveillance images into regions includes: A preset region growth segmentation algorithm is used to select the pixel with the largest gray value in the video surveillance image as the growth starting point. The growth criterion is that the difference between the gray value of the target pixel and the growth starting point is less than a preset difference, until the video surveillance image is divided into multiple independent regions.

6. The video surveillance method according to claim 1, characterized in that, The acquisition of video surveillance images from underground coal mines includes: Collect raw surveillance videos from underground coal mines; The images in the original surveillance video are preprocessed to obtain the video surveillance images; the preprocessing includes image enhancement and grayscale processing.

7. A video surveillance system for underground coal mines, characterized in that, include: Image acquisition module, weight correction module, and video compression module; The image acquisition module is used to acquire video surveillance images from underground coal mines and to preprocess the video surveillance images; the preprocessing includes image enhancement and grayscale processing. The weight correction module is used to divide the video surveillance image into regions and determine the importance weights of multiple regions based on image features and motion features within each region. Specifically, this includes: determining the foreign object degree of each region based on its grayscale value, number of pixels, and grayscale gradient differences between pixels; determining the safety hazard index of each region based on the displacement distance of pixels within the region in consecutive frames and the similarity of the displacement direction to a preset direction; and determining the importance weight of each region based on the foreign object degree and the safety hazard index. The multiple regions include all regions of the video surveillance image; the foreign object degree characterizes the probability of a region containing foreign objects; and the safety hazard index characterizes the degree of safety risk of a region. The quantification formula for determining the foreign matter content of each region is as follows: In the formula, Let the foreign object degree be the degree of the i-th region. Let be the standard deviation of the gray values ​​in the i-th region. This represents the average gray value within the i-th region. This represents the total number of pixels contained in the i-th region. , These represent the grayscale gradient magnitudes of the k-th and (k+1)-th pixels within the region, respectively. The formula for determining the safety hazard index for each region is as follows: In the formula, Let the safety hazard index be the value of the j-th region. This indicates the displacement direction of the k-th pixel within the j-th region. This indicates the pre-calibrated belt running direction of the conveyor belt. Represents cosine similarity. This represents the displacement distance of the k-th pixel within the j-th region. This represents the number of pixels in the j-th region; It is the hyperbolic tangent function; The quantification formula for determining the importance weight of each region is as follows: In the formula, This represents the importance weight of the j-th region. Indicates the weight adjustment factor. This represents the normalized value of the foreign matter degree in the j-th region; The weight correction module is further configured to correct the initial weight parameters used in video encoding according to the importance weights to obtain the corrected target weight parameters; the initial weight parameters are determined based on the frequency of grayscale values ​​in the image. The video compression module is used to perform Huffman coding compression on the video surveillance image using the target weight parameters, generating compressed video data for transmission to the ground monitoring center.

Images