Monitoring video quality detection and evaluation method

Abstract

The invention discloses a monitoring video quality detection and evaluation method. The method comprises the following steps: S1, acquiring a monitoring video and detecting the whole process of the monitoring video; s2, calculating a whole stack quality index of the monitoring video; s3, carrying out dimension reduction on the whole stack quality index and forming six dimension spaces for monitoring the video quality; and S4, respectively calculating an average value of the six dimension spaces and a quality score of the monitoring video, and carrying out quality evaluation according to the quality score. According to the invention, comprehensive and complete evaluation of the monitoring video quality is realized, and evaluation objectivity and data integrity are ensured. According to video quality evaluation, the data index is used as a unique source, the quality of the video can be truly represented, a user can carry out datamation evaluation on the video monitoring system, a comprehensive and real data basis is provided for video operation and maintenance work, and long-term stable operation of the video monitoring system is helped to be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of video quality detection, and in particular to a monitoring video quality detection and evaluation method. Background Art

[0002] The quality of surveillance video directly affects the user experience of the video surveillance system and the effectiveness of intelligent applications based on surveillance video. Therefore, the ability to accurately detect abnormal surveillance video quality is crucial to the stable and efficient operation of the video surveillance system.

[0003] At present, the solutions for detecting the quality of surveillance videos mainly adopt the following means: 1. Regularly collect image frames of surveillance video, analyze the image to determine whether the image is too dark, too bright, color cast, blurry, blocked, etc., and generate an alarm based on this, thereby characterizing the abnormal quality of the surveillance video. 2. Regularly collect image frames of surveillance video, load and run the image classification model, classify the image frames, and predict whether the image frame belongs to normal or abnormal type, as well as the probability.

[0004] However, through the analysis of surveillance video, we can find that the complete process of surveillance video from the issuance of video request signaling to the successful completion of video transmission is as follows: 1. The video requester (client) uses the RTSP protocol to send a signal to the video provider (server) to request the video; 2. The video provider performs permission authentication and responds to the video request; 3. After receiving the response from the video provider, the video requester establishes a video transmission channel with the video provider; 4. The video provider transmits the video to the video requester, and the video requester receives the video and decodes and plays it; 5. The video requester sends an end signal to end the video transmission process, and the video provider closes the video transmission channel.

[0005] It can be seen that the good operation of surveillance video requires the following conditions: normal signaling interaction, normal video stream transmission, and normal image screen. Therefore, in order to accurately evaluate the quality of surveillance video and determine whether it is in an abnormal state, the video signaling interaction, video stream transmission process, and image screen should be fully detected and analyzed. However, the current methods for monitoring video quality detection only detect and analyze image screens, which obviously makes it difficult to accurately detect the quality of surveillance video.

[0006] In addition, the current surveillance video quality detection and evaluation methods, in terms of the means of judging whether there are abnormalities, mainly include: 1. Using human experience to determine the threshold, and judging whether there is some kind of picture abnormality by comparing with the threshold; 2. Using a black box form of deep learning-based image classification or detection network for prediction.

[0007] The existing problems include: threshold judgment relies too much on experience to achieve a relatively accurate judgment result. At the same time, due to the unexplainability of convolutional neural networks, in order to achieve a high degree of accuracy in prediction results, a large amount of abnormal data needs to be prepared for model training, and in the case of inaccurate prediction results, the reasons cannot be explained. Summary of the invention

[0008] In order to solve the above problems, the present invention provides a monitoring video quality detection and evaluation method, comprising the following steps: S1. Acquire the monitoring video and detect the monitoring video throughout the whole process; S2. Calculate the full-stack quality index of the monitoring video; S3. Reduce the dimension of the full-stack quality index and form a 6-dimensional space of the monitoring video quality; S4. Calculate the average value of the 6-dimensional space and the quality score of the monitoring video respectively, and perform quality evaluation based on the quality score.

[0009] Furthermore, the S1 step specifically includes the following sub-steps: S11. Constructing a signaling feature library and a video feature library; S12. Pulling a monitoring video stream, capturing network data packets and disassembling and analyzing the network data packets; S13. Matching and identifying signaling data packets and video data packets in the network data packets according to the signaling feature library and the video feature library; S14. Extracting features from the identified signaling data packets and video data packets.

[0010] Furthermore, the features extracted from the identified signaling data packet in step S14 include: the number of key frames, the number of non-key frames, and the video data sequence number; the features extracted from the identified video data packet include: brightness, chroma, two-dimensional Fourier spectrum intensity, and texture complexity.

[0011] Furthermore, the full-stack quality indicators calculated in the S2 step specifically include: signaling delay, key frame loss rate, non-key frame loss rate, video key frame delay, brightness, chroma, two-dimensional Fourier spectrum intensity, and texture complexity.

[0012] Furthermore, the six dimensional spaces specifically include: signaling completion rate, smoothness rate, occlusion rate, blur rate, brightness rate, and color cast rate.

[0013] Furthermore, the calculation formulas for the dimensional characteristic values ​​of the six dimensional spaces are respectively: signaling completion rate = signaling sending rate*0.5+signaling receiving rate*0.5; smoothness rate = key frame loss rate*0.8+non-key frame loss rate*0.2; occlusion rate = (1-texture complexity)*0.6+brightness*0.4; blur rate = two-dimensional Fourier spectrum intensity*0.7+brightness*0.4; brightness and darkness rate = (255-brightness) / 255.0; color cast rate = (255-chroma) / 255.0.

[0014] Furthermore, the S4 step specifically includes the following sub-steps: S41. Repeat steps S1-S3 to obtain multiple groups of dimensional feature values; S42. Calculate the legal feature values ​​of each dimension according to the least squares method; S43. Use the algorithm to average and obtain the feature mean of each dimension; S44. Take the geometric average of the feature means of the six dimensions to obtain the quality evaluation coefficient; S45. Judge the quality of the surveillance video based on the quality evaluation coefficient.

[0015] Furthermore, the standard for judging the quality of the monitoring video according to the quality evaluation coefficient in step S45 is specifically: the higher the quality evaluation coefficient, the better the quality of the monitoring video; conversely, the lower the quality evaluation coefficient, the worse the quality of the monitoring video.

[0016] The present invention provides a monitoring video quality detection and evaluation method, which has the following beneficial effects: The present invention realizes a comprehensive and complete evaluation of the quality of surveillance videos, ensuring the objectivity of the evaluation and the integrity of the data. Video quality evaluation uses data indicators as the only source, which can truly characterize the quality of the video, allowing users to conduct a digital evaluation of the video surveillance system, providing a comprehensive and real data basis for video operation and maintenance work, and helping to achieve long-term stable operation of the video surveillance system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0018] Figure 1 The present invention provides a flow chart of the method. DETAILED DESCRIPTION

[0019] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0020] The following is a detailed description of the implementation method of the present invention in conjunction with the accompanying drawings. Only some embodiments are described, not all embodiments. For the purpose of clarity, representations and descriptions that are not related to the present invention are omitted in the drawings and descriptions.

[0021] In order to have a clearer understanding of the technical features, purposes and beneficial effects of the present invention, the technical solution of the present invention is now described in detail below. Obviously, the implementation cases described are part of the embodiments of the present invention, not all of the embodiments, and cannot be understood as limiting the scope of the implementation of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection scope of the present invention.

[0022] like Figure 1 As shown, the present invention provides a monitoring video quality detection and evaluation method, comprising the following steps: S1. Acquire monitoring video and detect the monitoring video throughout the entire process; S2. Calculate the full-stack quality index of the monitoring video; S3. Reduce the dimension of the full-stack quality index and form a 6-dimensional space of the monitoring video quality; S4. Calculate the average value of the 6-dimensional space and the quality score of the monitoring video respectively, and perform quality evaluation based on the quality score.

[0023] Step S1 specifically includes the following sub-steps: S11. Construct a signaling feature library and a video feature library; S12. Pull the monitoring video stream, capture the network data packet and disassemble and analyze the network data packet; S13. Match and identify the signaling data packet and the video data packet in the network data packet according to the signaling feature library and the video feature library; S14. Extract features from the identified signaling data packet and the video data packet. The features extracted from the identified signaling data packet in step S14 include: the number of key frames, the number of non-key frames, and the video data sequence number; the features extracted from the identified video data packet include: brightness, chroma, two-dimensional Fourier spectrum intensity, and texture complexity.

[0024] The full-stack quality indicators calculated in step S2 specifically include: signaling delay, key frame loss rate, non-key frame loss rate, video key frame delay, brightness, chroma, two-dimensional Fourier spectrum intensity, and texture complexity.

[0025] The six dimensional spaces specifically include: signaling completion rate, smoothness rate, occlusion rate, blur rate, brightness rate, and color cast rate. The calculation formulas for the dimensional characteristic values ​​of the six dimensional spaces are: signaling completion rate = signaling sending rate * 0.5 + signaling receiving rate * 0.5; smoothness rate = key frame loss rate * 0.8 + non-key frame loss rate * 0.2; occlusion rate = (1-texture complexity) * 0.6 + brightness * 0.4; blur rate = two-dimensional Fourier spectrum intensity * 0.7 + brightness * 0.4; brightness rate = (255-brightness) / 255.0; color cast rate = (255-chroma) / 255.0.

[0026] Step S4 specifically includes the following sub-steps: S41. Repeat steps S1-S3 to obtain multiple groups of dimensional feature values; S42. Calculate the legal feature values ​​of each dimension according to the least squares method; S43. Use the algorithm to average and obtain the feature mean of each dimension; S44. Perform geometric averaging on the feature means of the six dimensions to obtain the quality evaluation coefficient; S45. Judge the quality of the surveillance video according to the quality evaluation coefficient: the higher the quality evaluation coefficient, the better the surveillance video quality; conversely, the lower the quality evaluation coefficient, the worse the surveillance video quality. .

[0027] Embodiment: A monitoring video quality detection and evaluation method starts with detecting and collecting video signaling, and collects and analyzes video signaling quality, video transmission quality, and image quality. The entire process of monitoring video transmission is collected and analyzed. In the case of full-process collection, characteristic values ​​that affect the quality of the monitoring video are further extracted, and abnormal value detection is performed on the characteristic values. This method detects and collects indicators of the entire process of monitoring video transmission, further extracts characteristic values ​​of the entire process indicators, and performs multiple characteristic value data extractions to reduce the overfitting of characteristic values, and finally evaluates the quality of the monitoring video based on the characteristic values.

[0028] In order to detect and evaluate the quality of surveillance video, the video receiver needs to use RTSP (Streaming Media Transfer Protocol) to obtain the video stream from the video sender. The video quality detection and evaluation party is the video receiver. The complete video request transmission interaction process is as follows: the acquisition of the video stream starts with sending a video request signaling to the video sender. After receiving the video request signaling, the video sender returns a signaling that requires authentication information; the video receiver again sends a request to create a video session with authentication information, and the video sender confirms it and returns a confirmation signaling; the video receiver initiates a video playback request signaling, and the video sender receives the request signaling and starts transmitting the video; after a period of video transmission, the video receiver initiates a request signaling to close the video session link, and the video sender stops transmitting the video stream and returns a successful stop signaling.

[0029] When initiating the video request signaling, start the task of collecting network packet data. Capture the data packets received and sent by the network card in real time. The network data packets are placed in the cache queue for subsequent tasks to use. From the cache queue, read the network data packets in the first-in-first-out order, parse and extract the packet header and data payload of the network packet. Build the signaling data feature library of the video transmission protocol, extract the byte combination of the beginning part of the network packet data payload, and query the corresponding signaling in the signaling data feature library to achieve the purpose of identifying the signaling in the network packet; build the video data feature library, extract the beginning part of the network packet data payload by itself, use it as the query primary key, query the corresponding value, and if the query hits, the network packet is identified as a video packet.

[0030] After the signaling packet is identified, the signaling delay is calculated as one of the basic quality feature values. After the video packet is identified, the sequence number, key frame, and non-key frame identification information in the video data payload are further extracted. The key frame loss rate and non-key frame loss rate are further calculated as one of the basic quality feature values. The key frame video data is extracted from the video packet, and the video is decoded to obtain the original video data.

[0031] Read the original video data and use the following method to calculate the image feature value as one of the quality basic feature values: (1) Brightness: The average brightness of all pixels after the image is grayed out, with a value distribution of 0-255.

[0032] (2) Chroma: Calculate the color histogram of the image to obtain the number of pixels corresponding to each color, and then calculate the mean of the histogram as the chroma of the image, with a value distribution of 0-255.

[0033] (3) Two-dimensional Fourier spectrum intensity: Perform two-dimensional Fourier calculation on the image to obtain the spectrum distribution of the image. Then normalize it according to the spectrum distribution to obtain the spectrum intensity distributed between 0 and 1.

[0034] (4) Texture complexity: Through edge detection, the number of edge lines in the image is calculated, and the ratio of the number of lines to the number of pixels is calculated; the value is distributed between 0 and 1.

[0035] After completing the calculation of the image feature value, the indicators calculated during the execution of all the above steps are used to calculate the impact value according to the impact factor of the indicator, and the corresponding 6-dimensional feature values ​​are obtained, namely: signaling completion rate, smoothness rate, occlusion rate, blur rate, brightness rate, and color cast rate.

[0036] Among them: signaling completion rate = signaling sending rate * 0.5 + signaling receiving rate * 0.5; smoothness rate = key frame loss rate * 0.8 + non-key frame loss rate * 0.2; occlusion rate = (1-texture complexity) * 0.6 + brightness * 0.4; blur rate = two-dimensional Fourier spectrum intensity * 0.7 + brightness * 0.4; brightness and darkness rate = (255-brightness) / 255.0; color cast rate = (255-chroma) / 255.0.

[0037] Repeat steps 1-10 to get 9 more groups, a total of 10 groups of dimension eigenvalues. For each dimension, the 10 eigenvalues ​​are removed according to the principle of least squares, and the remaining eigenvalues ​​are used as the legal eigenvalues ​​of the dimension. The legal eigenvalues ​​are averaged to get the eigenvalue mean of the dimension.

[0038] The feature means of the six dimensions are then geometrically averaged to obtain the quality evaluation coefficient for video quality detection and evaluation, which is used to characterize the quality of the surveillance video.

[0039] The present invention realizes a comprehensive and complete evaluation of the quality of surveillance videos, ensuring the objectivity of the evaluation and the integrity of the data. Video quality evaluation uses data indicators as the only source, which can truly characterize the quality of the video, allowing users to conduct a digital evaluation of the video surveillance system, providing a comprehensive and real data basis for video operation and maintenance work, and helping to achieve long-term stable operation of the video surveillance system.

[0040] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A monitoring video quality detection and evaluation method, characterized in that: The following steps are involved: S1. Obtain surveillance video and perform full-scale detection of the surveillance video; S2. Calculate the full-stack quality indicators of the surveillance video; S3. Reduce the dimension of the full-stack quality indicators and form a 6-dimensional space for monitoring video quality; S4. Calculate the average values ​​of the six dimensional spaces and the quality scores of the surveillance videos respectively, and perform quality evaluation based on the quality scores.

2. The monitoring video quality detection and evaluation method according to claim 1 is characterized in that: The S1 step specifically includes the following sub-steps: S11. Build a signaling feature library and a video feature library; S12. Pull the monitoring video stream, capture the network data packets and disassemble and analyze the network data packets; S13. According to the signaling feature library and the video feature library, the signaling data packet and the video data packet are matched and identified in the network data packet; S14. Perform feature extraction on the identified signaling data packets and video data packets.

3. The monitoring video quality detection and evaluation method according to claim 2 is characterized in that: The features extracted from the identified signaling data packet in step S14 include: the number of key frames, the number of non-key frames, and the video data sequence number; the features extracted from the identified video data packet include: brightness, chroma, two-dimensional Fourier spectrum intensity, and texture complexity.

4. The monitoring video quality detection and evaluation method according to claim 1 is characterized in that: The full-stack quality indicators calculated in the S2 step specifically include: signaling delay, key frame loss rate, non-key frame loss rate, video key frame delay, brightness, chroma, two-dimensional Fourier spectrum intensity, and texture complexity.

5. The monitoring video quality detection and evaluation method according to claim 1 is characterized in that: The six dimensional spaces specifically include: signaling completion rate, smoothness rate, occlusion rate, blur rate, brightness rate, and color cast rate.

6. The monitoring video quality detection and evaluation method according to claim 5 is characterized in that: The calculation formulas for the dimensional eigenvalues ​​of the six dimensional spaces are: Signaling completion rate = signaling sending rate * 0.5 + signaling receiving rate * 0.5; Fluency rate = key frame loss rate * 0.8 + non-key frame loss rate * 0.2; Occlusion rate = (1-texture complexity) * 0.6 + brightness * 0.4; Blur rate = 2D Fourier spectrum intensity * 0.7 + brightness * 0.4; Brightness ratio = (255-brightness) / 255.0; Color cast rate = (255-chroma) / 255.

0.

7. The monitoring video quality detection and evaluation method according to claim 1 is characterized in that: The S4 step specifically includes the following sub-steps: S41. Repeat steps S1-S3 to obtain multiple groups of dimensional feature values; S42. Calculate the legal feature values ​​of each dimension according to the least squares method; S43. Use the algorithm to average and obtain the feature mean of each dimension; S44. Take the geometric average of the feature means of the six dimensions to obtain the quality evaluation coefficient; S45. Judge the quality of the surveillance video according to the quality evaluation coefficient.

8. The monitoring video quality detection and evaluation method according to claim 7 is characterized in that: The standard for judging the quality of the monitoring video according to the quality evaluation coefficient in the step S45 is specifically: the higher the quality evaluation coefficient, the better the quality of the monitoring video; conversely, the lower the quality evaluation coefficient, the worse the quality of the monitoring video.

Images