A security monitoring video optimization transmission method and video monitoring terminal

Abstract

The present invention discloses a security monitoring video optimized transmission method and a video monitoring terminal, which relates to the field of video monitoring technology, and includes the following steps: extracting the current frame and the previous frame from a continuous monitoring video frame sequence, and then extracting two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis; pre-processing the extracted two adjacent frames of images, using image difference technology to calculate the difference between the two adjacent frames of images, and generating a differential image through pixel-level comparison. The present invention extracts features such as motion intensity and change density by pre-processing and differential calculation of two adjacent frames of images, and inputs a deep learning model to analyze dynamic changes, and divides the scene into fast and normal dynamic changes. An initial compression algorithm is used for normal scenes, and the bit rate is dynamically adjusted for fast dynamic scenes to ensure the capture of details and balance video quality and compression efficiency, thereby improving the performance and practicality of the monitoring system.

Description

Technical Field

[0001] The present invention relates to the technical field of video monitoring, and in particular to a security monitoring video optimization transmission method and a video monitoring terminal. Background Art

[0002] Security surveillance video transmission optimization refers to improving the transmission efficiency and quality of video data through technical means under the premise of ensuring the reliability and security of the video surveillance system. Its purpose is to ensure that video data can be efficiently and losslessly transmitted to the monitoring center or storage device under limited bandwidth, so as to achieve real-time monitoring and event playback.

[0003] This optimized transmission usually involves a variety of technical means, such as video compression technology, network bandwidth management, data encryption, and error correction. Video compression technology can reduce the size of video data, thereby saving bandwidth; network bandwidth management ensures smooth data transmission; data encryption technology ensures the security of video data during transmission; error correction technology improves the reliability of data transmission and reduces the possibility of data loss and transmission errors.

[0004] When optimizing the transmission of security surveillance videos, video compression technology is used to compress the video in order to optimize the transmission. Video compression is usually divided into two types: lossy compression and lossless compression. Lossless compression reduces the amount of video data without losing any original video information, but the compression rate is low; lossy compression achieves a higher compression rate by removing some information that is not easily perceived by the human eye. Although it will have a certain impact on the video quality, this impact is usually within an acceptable range. Commonly used video compression standards include H.264, H.265, MPEG-4, etc. They can effectively achieve video compression and improve transmission efficiency and storage efficiency through advanced coding technology and algorithms.

[0005] The prior art has the following deficiencies:

[0006] When dealing with complex scenes with fast motion, for lossy compression, the compression algorithm may have difficulty accurately distinguishing between noise and important information. If the algorithm does not handle these situations properly, important visual information may be over-compressed. In security monitoring scenarios, video recordings are an important source of evidence. If over-compression causes loss of picture details, key information may be missed.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0008] The purpose of the present invention is to provide a security surveillance video optimization transmission method and a video surveillance terminal, which extracts features such as motion intensity and change density by preprocessing and differential calculation of two adjacent frames of images, and inputs a deep learning model to analyze dynamic changes, and divides the picture scene into fast dynamic change and normal dynamic change scenes. For normal scenes, an initial compression algorithm is used; for fast dynamic scenes, the video encoding bit rate is dynamically adjusted to ensure the capture of details and fast-moving targets. This method balances video quality and compression efficiency, ensures the provision of high-quality images at critical moments, helps case investigation and evidence collection, and improves the performance and practicality of the monitoring system to solve the problems in the above-mentioned background technology.

[0009] In order to achieve the above object, the present invention provides the following technical solution: a security monitoring video optimization transmission method, comprising the following steps:

[0010] Extract the current frame and the previous frame from the continuous surveillance video frame sequence, and then extract two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis;

[0011] Preprocess the extracted two adjacent frames of images, use image difference technology to calculate the difference between the two adjacent frames of images, and generate a differential image through pixel-level comparison to highlight the changed area between frames;

[0012] Extracting dynamic change feature information of the differential image, analyzing the dynamic change feature information, and inputting the analyzed dynamic change feature information into the trained deep learning model to analyze the dynamic change of the picture scene;

[0013] According to the analysis results of the deep learning model, the dynamic changes of the scene are divided into fast dynamic change scenes and normal dynamic change scenes;

[0014] For scenes with normal dynamic changes, the initial compression algorithm is used to continue compressing the video stream. For scenes with rapid dynamic changes, the compression algorithm is used to dynamically adjust the bit rate of video encoding to ensure that fast-moving targets and complex details in the video image are captured.

[0015] Preferably, the image difference technology is used to calculate the difference between two adjacent frames of images, and a difference image is generated by pixel-level comparison. The specific steps are as follows:

[0016] Convert two adjacent frames of images into grayscale images;

[0017] Through image registration technology, feature point matching method is used to adjust images so that images can be compared in the same coordinate system to ensure that two adjacent frames of images are aligned in space;

[0018] Smooth the grayscale image to remove noise and small fluctuations;

[0019] The difference between two adjacent grayscale images is calculated pixel by pixel. For each pixel position, the difference between the corresponding pixel values ​​of the current frame and the previous frame is calculated to generate a differential image. The calculation formula for the differential image is: ,in Indicates the difference image The difference value of the pixel position (specifically, the brightness difference between two adjacent frames of images at the pixel position), and denote the images at time t and time t−1 respectively, is the pixel coordinate, and The two frames of images are at pixel positions The pixel value of

[0020] The differential image is binarized and the differential value is converted into a binary image. The specific process is: the differential value is compared and analyzed with the preset differential reference threshold, and the pixels whose differential value is greater than or equal to the differential reference threshold are marked as white, and the pixels whose differential value is less than the differential reference threshold are marked as black.

[0021] Preferably, the dynamic change feature information of the extracted differential image includes motion intensity and change density, where motion intensity refers to the overall motion intensity in the image, and change density refers to the proportion of pixels that have changed in the image to the total number of pixels. After analyzing the motion intensity and change density of the differential image, a motion intensity index and a change density index are generated, and the motion intensity index and change density index obtained by the analysis are input into a trained deep learning model to generate a dynamic change quantization coefficient, and the dynamic changes of the picture scene are analyzed through the dynamic change quantization coefficient.

[0022] Preferably, the dynamic change quantization coefficient generated after analyzing the differential image is compared and analyzed with a preset dynamic change quantization coefficient reference threshold, and the dynamic change of the picture scene is divided, and the division process is as follows:

[0023] If the dynamic change quantization coefficient is greater than or equal to the dynamic change quantization coefficient reference threshold, the dynamic change of the picture scene is classified as a fast dynamic change scene;

[0024] If the dynamic change quantization coefficient is less than the dynamic change quantization coefficient reference threshold, the dynamic change of the picture scene is classified as a normal dynamic change scene.

[0025] Preferably, for fast dynamic changing scenes, the specific steps of dynamically adjusting the bit rate of video encoding through a compression algorithm are as follows:

[0026] The calculated dynamic change quantization coefficient is compared with the preset dynamic change quantization coefficient reference threshold, and the dynamic change coefficient deviation value is calculated. The calculation expression is: , where Indicates the deviation value of the dynamic change coefficient, Represents the dynamically changing quantization coefficient, represents the dynamically changing quantization coefficient reference threshold, and ;

[0027] Deviation value according to dynamic change coefficient , dynamically adjust the video encoding bit rate, set the initial video encoding bit rate to The actual bit rate after adjustment is , the adjustment factor is K , calculated by the following formula: , where is the maximum change ratio of the adjustment coefficient, is the sensitivity coefficient of the change, is the hyperbolic tangent function, which is used to smoothly adjust the coefficient changes;

[0028] The actual bit rate after dynamic adjustment Optimizes the video encoding process to ensure that fast-moving objects and complex details in video images are captured.

[0029] Preferably, after analyzing the motion intensity of the differential image, the steps of generating the motion intensity index are as follows:

[0030] Calculate the motion intensity for each pixel in the difference image and introduce a weight function To enhance the exercise intensity of certain areas, calculate the weighted exercise intensity, the calculation expression is: , where represents weighted exercise intensity;

[0031] The motion intensity of the entire differential image is summed to obtain the overall motion intensity, and the calculation expression is: , where, W and H Respectively represent the width and height of the image, Indicates overall exercise intensity;

[0032] The overall intensity of exercise Normalization is used for comparison between different images. The normalized expression is: , where represents the normalized exercise intensity;

[0033] Considering the change of motion intensity of multiple consecutive frames, the time-weighted average is introduced to calculate the motion intensity index. The calculation expression is: , where represents the exercise intensity index, Represents the time weighting coefficient, ranging from 0 to 1, which controls the weight ratio of the motion intensity of the current frame and the previous frame. is the motion intensity index of the previous frame.

[0034] Preferably, after analyzing the change density in the differential image, the steps of generating the change density index are as follows:

[0035] Apply threshold to the difference image To determine the changed pixels, define the changed pixels , ;

[0036] Count the total number of changed pixels in the difference image , ;

[0037] Calculate the change density, the calculation expression is: ,in, is the total number of pixels in the image, Indicates the density of change;

[0038] The change density index is calculated based on the change density. The calculation expression is: , where represents the change density index, Indicates the expected change density threshold.

[0039] A video monitoring terminal includes a data extraction module, an image difference module, a feature extraction and analysis module, a scene classification module and a compression strategy adjustment module;

[0040] The data extraction module extracts the current frame and the previous frame from the continuous surveillance video frame sequence, and then extracts two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis;

[0041] The image difference module pre-processes the extracted two adjacent frames of images, uses the image difference technology to calculate the difference between the two adjacent frames of images, and generates a difference image through pixel-level comparison to highlight the changed area between the frames;

[0042] The feature extraction and analysis module extracts the dynamic change feature information of the differential image, analyzes the dynamic change feature information, and then inputs the analyzed dynamic change feature information into the trained deep learning model to analyze the dynamic change of the picture scene;

[0043] The scene classification module divides the dynamic changes of the scene into fast dynamic change scenes and normal dynamic change scenes according to the analysis results of the deep learning model;

[0044] The compression strategy adjustment module uses the initial compression algorithm to continue compressing the video stream for normal dynamic change scenes. For fast dynamic change scenes, it dynamically adjusts the video encoding bit rate through the compression algorithm to ensure that fast-moving targets and complex details in the video image are captured.

[0045] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0046] The present invention extracts dynamic change feature information such as motion intensity and change density by extracting two adjacent frames of images for preprocessing and differential calculation, and inputs these feature information into the deep learning model to accurately analyze the dynamic changes of the picture scene. According to the analysis results of the deep learning model, the picture scene is divided into a rapid dynamic change scene and a normal dynamic change scene. In the normal dynamic change scene, the initial compression algorithm is used to continue video stream compression, and in the rapid dynamic change scene, the bit rate of video encoding is dynamically adjusted to ensure that fast-moving targets and complex details are captured. Not only does it achieve a balance between video quality and compression efficiency, it also ensures that the monitoring video can provide high-quality image details at critical moments, thereby ensuring the reliability and effectiveness of security monitoring, facilitating case investigation and evidence collection, and ultimately improving the overall performance and practicality of the monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0048] Figure 1 The present invention is a method flow chart of a security monitoring video optimization transmission method.

[0049] Figure 2 The present invention is a module schematic diagram of a video monitoring terminal. DETAILED DESCRIPTION

[0050] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0051] The present invention provides Figure 1 A security surveillance video optimization transmission method shown includes the following steps:

[0052] Extract the current frame and the previous frame from the continuous surveillance video frame sequence, and then extract two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis;

[0053] The current frame and the previous frame are selected from the continuous surveillance video frame sequence for subsequent dynamic change analysis. The two frames are adjacent and can reflect the changes in the video within a short period of time, providing basic data for analyzing changes in video content.

[0054] The main purpose of extracting two adjacent frames is to compare and analyze the changes between video frames. By comparing adjacent frames, it is possible to identify which parts of the video have changed, such as the trajectory, speed, and direction of moving objects. This is very important for the analysis of surveillance scenes because it can help identify and track dynamic events, such as intrusions, moving targets, etc., thereby improving the efficiency and accuracy of the surveillance system.

[0055] When extracting two adjacent frames of images, the following points need to be noted:

[0056] Inter-frame interval: Make sure the time interval between the two extracted frames is short enough to accurately capture dynamic changes in a short period of time. If the interval is too long, some subtle but important changes may be missed.

[0057] Image quality: The extracted frame images need to maintain high quality to ensure the accuracy of subsequent analysis. Problems such as noise and blur in the image will affect the detection and analysis results of dynamic changes.

[0058] Synchronization issues: When extracting frame images, ensure that the timestamps and order of the images are accurate to avoid analysis errors caused by asynchronous frame sequences.

[0059] Preprocess the extracted two adjacent frames of images, use image difference technology to calculate the difference between the two adjacent frames of images, and generate a differential image through pixel-level comparison to highlight the changed area between frames;

[0060] Preprocessing of the extracted two adjacent frames of images usually includes grayscale processing, noise filtering, edge detection, etc. The role of these preprocessing is to reduce redundant information and noise in the image, highlight the main features of the image, and provide clearer and more accurate data input for subsequent dynamic change analysis, thereby improving the effectiveness and reliability of the analysis.

[0061] Use image difference technology to calculate the difference between two adjacent frames of images, and generate a differential image through pixel-level comparison. The specific steps are as follows:

[0062] Convert two adjacent frames of images into grayscale images;

[0063] Grayscale processing is to convert a color image into a grayscale image, retaining only the brightness information and removing the color information. This can simplify subsequent processing steps, reduce the amount of calculation, and make the difference operation more direct and effective. Grayscale processing can be achieved by weighted averaging or directly taking a single color channel.

[0064] Through image registration technology, feature point matching method is used to adjust images so that images can be compared in the same coordinate system to ensure that two adjacent frames of images are aligned in space;

[0065] If the camera moves or shakes slightly during shooting, image alignment can correct this error and ensure the accuracy of the differential result.

[0066] Smooth the grayscale image to remove noise and small fluctuations;

[0067] Smoothing can be achieved through convolution operations, using a Gaussian kernel to smooth the image, thereby reducing the impact of high-frequency noise. Smoothing can make subsequent differential images smoother and more stable, and reduce the situation where noise is misjudged as a changing area.

[0068] The difference between two adjacent grayscale images is calculated pixel by pixel. For each pixel position, the difference between the corresponding pixel values ​​of the current frame and the previous frame is calculated to generate a differential image. The calculation formula for the differential image is: ,in Indicates the difference image The difference value of the pixel position (specifically, the brightness difference between two adjacent frames of images at that pixel position), and They represent the images at time t (i.e., the current frame) and time t−1 (i.e., the previous frame), respectively. is the pixel coordinate, and The two frames of images are at pixel positions The pixel value of

[0069] Binarization is performed on the difference image to convert the difference value into a binary image. The specific process is: the difference value is compared and analyzed with the preset difference reference threshold, and the pixel points with a difference value greater than or equal to the difference reference threshold are marked as white (changing area), and the pixel points with a difference value less than the difference reference threshold are marked as black (static area);

[0070] Binarization can clearly separate the changing area and the static area, making the subsequent dynamic analysis more intuitive and effective. The appropriate threshold can be adjusted according to the actual scenario to ensure that important change information can be accurately captured.

[0071] Extracting dynamic change feature information of the differential image, analyzing the dynamic change feature information, and inputting the analyzed dynamic change feature information into the trained deep learning model to analyze the dynamic change of the picture scene;

[0072] The dynamic change feature information of the extracted differential image includes motion intensity and change density. Motion intensity refers to the overall motion intensity in the image, and change density refers to the proportion of pixels that have changed in the image to the total number of pixels. After analyzing the motion intensity and change density of the differential image, a motion intensity index and a change density index are generated. The motion intensity index and change density index obtained by the analysis are input into the trained deep learning model to generate a dynamic change quantization coefficient, and the dynamic changes of the scene are analyzed through the dynamic change quantization coefficient.

[0073] In the differential image, the overall motion intensity in the image is large, indicating that the dynamic changes in the scene are large. This is because the motion intensity measures the average value of the brightness difference of each pixel in two adjacent frames of the image, reflecting the degree of pixel change in the entire image. A larger motion intensity means that more pixels have changed significantly, which is usually caused by significant motion or dynamic events in the scene. Specifically, when there are a large number of objects moving in a scene, or the objects move very fast, the pixel values ​​will change greatly, resulting in an increase in motion intensity. For example, in a surveillance video, if there is a group of people running fast or multiple cars driving at high speed, the motion intensity in the differential image will increase significantly, reflecting that the dynamic changes in the picture are very significant. This situation not only indicates that the amplitude of the movement is large, but also indicates that the range of the movement may be wide, covering a large image area. Therefore, by analyzing the motion intensity in the differential image, the dynamic changes of the scene can be effectively identified and evaluated. A larger overall motion intensity in the image usually indicates that the dynamic changes of the scene are large, because this reflects the presence of significant and extensive motion in the scene, which is crucial for monitoring and analyzing dynamic events.

[0074] After analyzing the motion intensity of the differential image, the steps to generate the motion intensity index are as follows:

[0075] Calculate the motion intensity for each pixel in the difference image and introduce a weight function To enhance the exercise intensity of certain areas, calculate the weighted exercise intensity, the calculation expression is: , where represents weighted exercise intensity;

[0076] It should be noted that It is a weight function that can be customized according to the application scenario. For example, different weights can be assigned according to the importance of different regions in the image.

[0077] The motion intensity of the entire differential image is summed to obtain the overall motion intensity, and the calculation expression is: , where, W and H Respectively represent the width and height of the image, Indicates overall exercise intensity;

[0078] The overall intensity of exercise Normalization is used for comparison between different images. The normalized expression is: , where represents the normalized exercise intensity;

[0079] Considering the change of motion intensity of multiple consecutive frames, the time-weighted average is introduced to calculate the motion intensity index. The calculation expression is: , where represents the exercise intensity index, Represents the time weighting coefficient, ranging from 0 to 1, which controls the weight ratio of the motion intensity of the current frame and the previous frame. is the motion intensity index of the previous frame.

[0080] It can be seen from the motion intensity index that the larger the performance value of the motion intensity index generated after analyzing the motion intensity of the differential image, the greater the dynamic change of the picture scene. This means that there are more pixel value changes between two adjacent frames, which usually reflects significant movement or changes, such as fast-moving objects or drastic scene changes. On the contrary, the smaller the performance value of the motion intensity index, the smaller the dynamic change of the picture scene, indicating that there are fewer changes between two adjacent frames, indicating that the scene is relatively static or has only slight changes. The motion intensity index can effectively help the monitoring system identify and distinguish different types of dynamic scenes.

[0081] In the differential image, a large change density in the image indicates that the dynamic changes of the scene are large. Change density refers to the ratio of pixels that have changed in the differential image to the total number of pixels. A high change density means that a large number of pixels have changed between two adjacent frames, which is usually caused by more or larger moving objects in the scene. For example, in a busy street surveillance image, the movement of pedestrians, vehicles, etc. will lead to a high change density, because the movement of these objects will affect the position and color changes of a large number of pixels. In addition, fast-moving objects in the scene, such as running people, speeding vehicles, or sudden objects (such as the opening and closing of a door or the falling of an object), will also cause a significant increase in the change density. These changes reflect the frequency and intensity of dynamic events occurring in the scene, so the dynamic changes of the scene can be judged by analyzing the change density. When the change density is large, it means that there are many movements or changes in the scene. In summary, change density is an important indicator for evaluating the dynamic changes of the scene. A high change density means that there are significant dynamic changes in the scene, which poses higher challenges and requirements for monitoring and video compression.

[0082] After analyzing the change density in the difference image, the steps to generate the change density index are as follows:

[0083] Apply threshold to the difference image To determine the changed pixels, define the changed pixels ;

[0084] It should be noted that is a pre-set threshold that With threshold Perform comparative analysis to distinguish significant changes from background noise or minor changes.

[0085] Count the total number of changed pixels in the difference image , ;

[0086] This step counts all pixels marked as 1, that is, pixels that meet the threshold condition.

[0087] Calculate the change density, the expression is: ,in, is the total number of pixels in the image, Indicates the density of change;

[0088] The change density index is calculated based on the change density. The calculation expression is: , where represents the change density index, represents the expected change density threshold;

[0089] The expected change density threshold is a preset parameter that defines the midpoint that is considered important when calculating the change density index. When the change density (the ratio of pixels that have changed in the image to the total number of pixels) is close to this threshold, the response of the change density index will be more significant. It helps the model better adapt to specific application scenarios, making it more sensitive to change density within a specific range, thereby accurately reflecting the dynamic changes between video frames.

[0090] The larger the change density index value generated after analyzing the change density of the differential image, the greater the dynamic change of the scene. This is because the change density index is calculated based on the proportion of changed pixels in the differential image. The higher the change density means that more pixels have changed significantly between adjacent frames, indicating that there are more or larger moving objects in the scene. These significant changes usually represent important dynamic events in the monitoring scene. On the contrary, the smaller the change density index value, the smaller the dynamic change of the scene, which means that fewer pixels have changed between adjacent frames, and the scene is basically static or has only slight changes.

[0091] The machine learning model is not specifically limited here, and can realize the exercise intensity index and the change density index Perform comprehensive analysis to generate dynamic change quantitative coefficients In order to realize the technical solution of the present invention, the present invention provides a specific implementation method;

[0092] Dynamically changing quantization coefficients The resulting calculation formula is: , where , Exercise Intensity Index and the change density index The preset scaling factor of , Both are greater than 0.

[0093] It can be seen from the dynamic change quantization coefficient that the greater the expression value of the motion intensity index generated after analyzing the motion intensity of the differential image, the greater the expression value of the change density index generated after analyzing the change density of the differential image. In other words, the greater the expression value of the dynamic change quantization coefficient generated after analyzing the differential image, the greater the dynamic change of the picture scene, and vice versa.

[0094] According to the analysis results of the deep learning model, the dynamic changes of the scene are divided into fast dynamic change scenes and normal dynamic change scenes;

[0095] The dynamic change quantization coefficient generated after analyzing the differential image is compared and analyzed with the preset dynamic change quantization coefficient reference threshold, and the dynamic change of the picture scene is divided. The division process is as follows:

[0096] If the dynamic change quantization coefficient is greater than or equal to the dynamic change quantization coefficient reference threshold, the dynamic change of the picture scene is classified as a fast dynamic change scene;

[0097] If the dynamic change quantization coefficient is less than the dynamic change quantization coefficient reference threshold, the dynamic change of the picture scene is classified as a normal dynamic change scene.

[0098] For scenes with normal dynamic changes, the initial compression algorithm is used to continue compressing the video stream. For scenes with rapid dynamic changes, the compression algorithm is used to dynamically adjust the bit rate of video encoding to ensure that fast-moving targets and complex details in the video image are captured.

[0099] In surveillance video processing, if the differential image formed by two adjacent frames is analyzed to determine that the scene is a normal dynamic change scene (i.e., a scene with small or few changes), then in this case, the originally set compression algorithm can continue to be used to compress the video stream. Specifically, this means that there is no need to adjust or optimize the compression algorithm, because the changes in normal dynamic change scenes are small, and the original compression algorithm is sufficient to cope with these changes, and can effectively compress video data while ensuring video quality.

[0100] For fast dynamic changing scenes, the specific steps to dynamically adjust the bit rate of video encoding through compression algorithms are as follows:

[0101] The calculated dynamic change quantization coefficient is compared with the preset dynamic change quantization coefficient reference threshold, and the dynamic change coefficient deviation value is calculated. The calculation expression is: , where Indicates the deviation value of the dynamic change coefficient, Represents the dynamically changing quantization coefficient, represents the dynamically changing quantization coefficient reference threshold, and ;

[0102] Deviation value according to dynamic change coefficient , dynamically adjust the video encoding bit rate, set the initial video encoding bit rate to The actual bit rate after adjustment is , the adjustment factor is K , calculated by the following formula: , where is the maximum change ratio of the adjustment coefficient, is the sensitivity coefficient of the change, is the hyperbolic tangent function, which is used to smoothly adjust the coefficient changes;

[0103] The maximum change ratio of the adjustment coefficient is , determines the maximum bit rate adjustment. Specifically, It is used to control the maximum ratio by which the video encoding bit rate can be increased or decreased. When the dynamic change quantization coefficient of the scene is significantly higher than the reference threshold, the bit rate needs to be significantly increased to capture fast-moving targets and complex details, and vice versa. If the setting is too high, the video bit rate will be adjusted too drastically, which may lead to a waste of network bandwidth and storage resources; if the setting is too low, the improvement or reduction in video quality may not be enough to cope with the dynamic changes of the scene. Careful adjustments need to be made based on specific application scenarios and system performance to find the best balance between video quality and resource usage.

[0104] The sensitivity coefficient of the change is , controls the sensitivity of the bitrate adjustment to the dynamic change of the quantization coefficient difference. It determines the sensitivity of the bitrate adjustment, that is, how the bitrate responds to the change of the dynamically changing quantization coefficient. A higher A value of 0 means that the system will be very sensitive to small dynamic changes in quantization coefficients, resulting in rapid changes in the bit rate, which may be necessary in highly dynamic and complex scenes. However, too high a value This may cause frequent bitrate fluctuations, which is not conducive to bandwidth stability and smooth video playback. The value will make the system respond more slowly to the difference in the quantization coefficient, which is suitable for relatively static or small-change scenes, but may not fully capture the details of fast dynamic changes. The setting needs to comprehensively consider multiple factors such as scene dynamic characteristics, system response speed and bandwidth stability.

[0105] The actual bit rate after dynamic adjustment Optimizes the video encoding process to ensure that fast-moving objects and complex details in video images are captured.

[0106] The present invention extracts dynamic change feature information such as motion intensity and change density by extracting two adjacent frames of images for preprocessing and differential calculation, and inputs these feature information into the deep learning model to accurately analyze the dynamic changes of the picture scene. According to the analysis results of the deep learning model, the picture scene is divided into a rapid dynamic change scene and a normal dynamic change scene. In the normal dynamic change scene, the initial compression algorithm is used to continue video stream compression, and in the rapid dynamic change scene, the bit rate of video encoding is dynamically adjusted to ensure that fast-moving targets and complex details are captured. Not only does it achieve a balance between video quality and compression efficiency, it also ensures that the monitoring video can provide high-quality image details at critical moments, thereby ensuring the reliability and effectiveness of security monitoring, facilitating case investigation and evidence collection, and ultimately improving the overall performance and practicality of the monitoring system.

[0107] The present invention provides Figure 2 A video surveillance terminal shown includes a data extraction module, an image difference module, a feature extraction and analysis module, a scene classification module and a compression strategy adjustment module;

[0108] The data extraction module extracts the current frame and the previous frame from the continuous surveillance video frame sequence, and then extracts two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis;

[0109] The image difference module pre-processes the extracted two adjacent frames of images, uses the image difference technology to calculate the difference between the two adjacent frames of images, and generates a difference image through pixel-level comparison to highlight the changed area between the frames;

[0110] The feature extraction and analysis module extracts the dynamic change feature information of the differential image, analyzes the dynamic change feature information, and then inputs the analyzed dynamic change feature information into the trained deep learning model to analyze the dynamic change of the picture scene;

[0111] The scene classification module divides the dynamic changes of the scene into fast dynamic change scenes and normal dynamic change scenes according to the analysis results of the deep learning model;

[0112] The compression strategy adjustment module uses the initial compression algorithm to continue video stream compression for normal dynamic change scenes. For fast dynamic change scenes, it dynamically adjusts the video encoding bit rate through the compression algorithm to ensure that fast-moving targets and complex details in the video image are captured.

[0113] An optimized transmission method for security surveillance video provided in an embodiment of the present invention is implemented by the above-mentioned video surveillance terminal. The specific method and process of a video surveillance terminal are detailed in the embodiment of the optimized transmission method for security surveillance video, which will not be repeated here.

[0114] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0115] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0116] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A security surveillance video optimization transmission method, characterized in that: The following steps are involved: Extract the current frame and the previous frame from the continuous surveillance video frame sequence, and then extract two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis; Preprocess the extracted two adjacent frames of images, use image difference technology to calculate the difference between the two adjacent frames of images, and generate a differential image through pixel-level comparison to highlight the changed area between frames; Extracting dynamic change feature information of the differential image, analyzing the dynamic change feature information, and inputting the analyzed dynamic change feature information into the trained deep learning model to analyze the dynamic change of the picture scene; According to the analysis results of the deep learning model, the dynamic changes of the scene are divided into fast dynamic change scenes and normal dynamic change scenes; For scenes with normal dynamic changes, the initial compression algorithm is used to continue compressing the video stream. For scenes with rapid dynamic changes, the compression algorithm is used to dynamically adjust the bit rate of video encoding to ensure that fast-moving targets and complex details in the video image are captured. The extracted dynamic change feature information of the differential image includes motion intensity and change density. Motion intensity refers to the overall motion intensity in the image, and change density refers to the proportion of pixels that have changed in the image to the total number of pixels. After analyzing the motion intensity and change density of the differential image, a motion intensity index and a change density index are generated. The motion intensity index and change density index obtained by the analysis are input into the trained deep learning model to generate a dynamic change quantization coefficient. The dynamic change of the scene is analyzed through the dynamic change quantization coefficient. The dynamic change quantization coefficient generated after analyzing the differential image is compared and analyzed with the preset dynamic change quantization coefficient reference threshold, and the dynamic change of the picture scene is divided. The division process is as follows: If the dynamic change quantization coefficient is greater than or equal to the dynamic change quantization coefficient reference threshold, the dynamic change of the picture scene is classified as a fast dynamic change scene; If the dynamic change quantization coefficient is less than the dynamic change quantization coefficient reference threshold, the dynamic change of the picture scene is classified as a normal dynamic change scene; For fast dynamic changing scenes, the specific steps to dynamically adjust the bit rate of video encoding through compression algorithms are as follows: The calculated dynamic change quantization coefficient is compared with the preset dynamic change quantization coefficient reference threshold, and the dynamic change coefficient deviation value is calculated. The calculation expression is: , where Indicates the deviation value of the dynamic change coefficient, Represents the dynamically changing quantization coefficient, represents the dynamically changing quantization coefficient reference threshold, and ; Deviation value according to dynamic change coefficient , dynamically adjust the video encoding bit rate, set the initial video encoding bit rate to The actual bit rate after adjustment is , the adjustment factor is K , calculated by the following formula: , where is the maximum change ratio of the adjustment coefficient, is the sensitivity coefficient of the change, is the hyperbolic tangent function, which is used to smoothly adjust the coefficient changes; The actual bit rate after dynamic adjustment Optimizes the video encoding process to ensure that fast-moving objects and complex details in video images are captured.

2. A security surveillance video optimization transmission method according to claim 1, characterized in that: Use image difference technology to calculate the difference between two adjacent frames of images, and generate a differential image through pixel-level comparison. The specific steps are as follows: Convert two adjacent frames of images into grayscale images; Through image registration technology, feature point matching method is used to adjust images so that images can be compared in the same coordinate system to ensure that two adjacent frames of images are aligned in space; Smooth the grayscale image to remove noise and small fluctuations; The difference between two adjacent grayscale images is calculated pixel by pixel. For each pixel position, the difference between the corresponding pixel values ​​of the current frame and the previous frame is calculated to generate a differential image. The calculation formula for the differential image is: ,in Indicates the difference image The difference value of the pixel position, and denote the images at time t and time t−1 respectively, is the pixel coordinate, and The two frames of images are at pixel positions The pixel value of The differential image is binarized and the differential value is converted into a binary image. The specific process is: the differential value is compared and analyzed with the preset differential reference threshold, and the pixels whose differential value is greater than or equal to the differential reference threshold are marked as white, and the pixels whose differential value is less than the differential reference threshold are marked as black.

3. The security surveillance video optimization transmission method according to claim 1 is characterized in that: After analyzing the motion intensity of the differential image, the steps to generate the motion intensity index are as follows: Calculate the motion intensity for each pixel in the difference image and introduce a weight function To enhance the exercise intensity of certain areas, calculate the weighted exercise intensity, the calculation expression is: , where represents weighted exercise intensity; The motion intensity of the entire differential image is summed to obtain the overall motion intensity, and the calculation expression is: , where, W and H Respectively represent the width and height of the image, Indicates overall exercise intensity; The overall intensity of exercise Normalization is used for comparison between different images. The normalized expression is: , where represents the normalized exercise intensity; Considering the change of motion intensity of multiple consecutive frames, the time-weighted average is introduced to calculate the motion intensity index. The calculation expression is: , where represents the exercise intensity index, Represents the time weighting coefficient, ranging from 0 to 1, which controls the weight ratio of the motion intensity of the current frame and the previous frame. is the motion intensity index of the previous frame.

4. The method for optimizing transmission of security surveillance video according to claim 1, characterized in that: After analyzing the change density in the difference image, the steps to generate the change density index are as follows: Apply threshold to the difference image To determine the changed pixels, define the changed pixels , ; Count the total number of changed pixels in the difference image , ; Calculate the change density, the calculation expression is: ,in, is the total number of pixels in the image, Indicates the density of change; The change density index is calculated based on the change density. The calculation expression is: , where represents the change density index, Indicates the expected change density threshold.

5. A video surveillance terminal, used to implement the security surveillance video optimization transmission method described in any one of claims 1 to 4, characterized in that: It includes data extraction module, image difference module, feature extraction and analysis module, scene classification module and compression strategy adjustment module; The data extraction module extracts the current frame and the previous frame from the continuous surveillance video frame sequence, and then extracts two adjacent frames of images from the video stream to provide basic data for subsequent dynamic change analysis; The image difference module pre-processes the extracted two adjacent frames of images, uses the image difference technology to calculate the difference between the two adjacent frames of images, and generates a difference image through pixel-level comparison to highlight the changed area between the frames; The feature extraction and analysis module extracts the dynamic change feature information of the differential image, analyzes the dynamic change feature information, and then inputs the analyzed dynamic change feature information into the trained deep learning model to analyze the dynamic change of the picture scene; The scene classification module divides the dynamic changes of the scene into fast dynamic change scenes and normal dynamic change scenes according to the analysis results of the deep learning model; The compression strategy adjustment module uses the initial compression algorithm to continue compressing the video stream for normal dynamic change scenes. For fast dynamic change scenes, it dynamically adjusts the video encoding bit rate through the compression algorithm to ensure that fast-moving targets and complex details in the video image are captured.

Images