CN-121985097-A - Safety production monitoring method and system based on video data

CN121985097ACN 121985097 ACN121985097 ACN 121985097ACN-121985097-A

Abstract

The invention relates to the technical field of video compression, in particular to a safety production monitoring method and system based on video data, comprising the steps of collecting video frame data of a coal mine site; the method comprises the steps of detecting video frame data by using a target detection network to obtain the confidence coefficient of a target, obtaining the confidence coefficient of the target by using the confidence coefficient of the target, determining a high-confidence target and a low-confidence target based on the confidence coefficient of the target, obtaining the judgment quantity of each video frame as a key frame according to the high-confidence target, screening out a first key frame based on the judgment quantity of each video frame as the key frame, obtaining the judgment quantity of the low-confidence target as the key target by using the low-confidence target in the rest video frames, obtaining a second key frame based on the judgment quantity of the low-confidence target as the key target, and compressing and transmitting the first key frame data and the second key frame data. The invention can improve the transmission efficiency and reliability of the video monitoring system.

Inventors

ZHAO HU
ZHAO YADONG
LIU JIE
MA TENGFEI
HE MAOWEI
Si Wangdou
LI YONG

Assignees

国能神东煤炭集团有限责任公司

Dates

Publication Date: 20260505
Application Date: 20260311

Claims (10)

1. A method for monitoring the safe production based on video data, characterized in that the method comprises the following steps: Collecting video frame data of a coal mine site; Detecting the video frame data by adopting a target detection network, and acquiring the confidence coefficient of each target in each video frame; acquiring the confidence level of each target in each video frame by using the confidence level of each target in each video frame; Determining a high confidence target and a low confidence target in each video frame based on the confidence level of each target in each video frame; Acquiring the judgment quantity of each video frame as a key frame according to a high confidence target in each video frame; Screening a first key frame from all video frames based on the judgment amount that each video frame is a key frame, wherein the first key frame is the key frame screened for the first time from all video frames; acquiring a judgment quantity of which the low confidence coefficient target is a key target by utilizing the low confidence coefficient targets in the rest video frames except the first key frame in all the video frames; Acquiring a second key frame based on the judgment amount of which the low confidence level target is a key target, wherein the second key frame is a key frame screened out of all video frames again; The first key frame data and the second key frame data are compressed and transmitted.
2. The method for monitoring the safe production based on the video data according to claim 1, wherein the step of obtaining the confidence level of each target in each video frame by using the confidence level of each target in each video frame comprises the following specific steps: for a current target in a current video frame, acquiring a confidence entropy value and a confidence mean value of the target in a first preset number of video frames before the current frame; Subtracting a preset confidence coefficient threshold value from the confidence coefficient mean value to obtain a difference value, multiplying the difference value by the reciprocal of the confidence coefficient entropy value, and calculating to obtain the confidence coefficient of the current target in the current video frame; And obtaining the confidence level of each target in each video frame.
3. The method for monitoring the safe production based on the video data according to claim 2, wherein the step of obtaining the confidence entropy value of the target in the first preset number of video frames before the current frame comprises the following specific steps: acquiring the confidence coefficient of the target in a first preset number of video frames before the current frame, and sequencing according to a time sequence to obtain a confidence coefficient sequence; Dividing the confidence coefficient range into a plurality of equal-width intervals, counting the distribution frequency of the confidence coefficient sequence in each interval, and then calculating the confidence coefficient entropy value of the target in a first preset number of video frames before the current frame according to an information entropy formula.
4. The method for monitoring the safe production based on the video data according to claim 1, wherein the step of determining the high confidence level target and the low confidence level target in each video frame based on the confidence level of each target in each video frame comprises the following specific steps: Comparing a normalized value of the confidence level of the current target in the current video frame with a preset high-low threshold value: if the normalized value of the confidence coefficient of the target is larger than a preset high-low threshold value, judging that the target is a high-confidence coefficient target; if the normalized value of the confidence level of the target is smaller than or equal to a preset high-low threshold value, judging that the target is a low-confidence target.
5. The method for monitoring the safe production based on the video data according to claim 1, wherein the step of obtaining the determination amount of each video frame as the key frame according to the high confidence target in each video frame comprises the following specific steps: taking any high-confidence target in the current video frame as a first target; Calculating the ratio of the frame length to the frame length of the identification frame of the first target in the current video frame, and taking the ratio as the current length ratio of the first target; acquiring the length ratio of the first target in a second preset number of video frames before the current frame, and taking each length ratio as the historical length ratio of the first target; calculating the difference value of the current length ratio of the first target and each historical length ratio, and summing to obtain the length ratio difference value sum of the first target; summing the sum of the length ratio differences of all the first targets to obtain a first sum value; summing the confidence degrees of all the first targets to obtain a second sum value; The second sum value is multiplied by the first sum value to determine the current video frame as the key frame.
6. The method for monitoring the safe production based on the video data according to claim 1, wherein the step of screening the first key frame from all the video frames based on the determination that each video frame is a key frame comprises the following specific steps: Comparing the judgment quantity of each video frame as a key frame with a preset first judgment quantity threshold value: If the judgment quantity of each video frame as the key frame is larger than a preset first judgment quantity threshold value, judging the video frame as the first key frame, and determining the video frames with the third preset quantity after the first key frame as the first key frame.
7. The method for monitoring the safe production based on the video data according to claim 1, wherein the step of obtaining the decision of the low confidence target as the key target by using the low confidence targets in the remaining video frames except the first key frame in all the video frames comprises the following specific steps: determining each video frame except the first key frame in all video frames as a first video frame; determining any low confidence level target in the current first video frame as a second target; Acquiring a gray density value of a connected domain of the second target based on the identification frame connected domain of the second target; Acquiring a moving direction angle value of the identification frame connected domain of the second target by adopting an optical flow algorithm, calculating a difference absolute value between the angle value and 90 degrees, and recording the difference absolute value as a first absolute value; Dividing the gray density value of the connected domain of the second target by the first absolute value, and calculating to obtain the judgment quantity of the second target as the key target.
8. The method for monitoring the safe production based on the video data according to claim 7, wherein the step of obtaining the gray density value of the connected domain of the second object based on the connected domain of the identification frame of the second object comprises the following specific steps: Acquiring an identification frame connected domain of a second target, and gray values of all pixel points and total number of the pixel points in the connected domain; Acquiring the gray scale range of the connected domain based on the gray scale values of all pixel points in the connected domain; dividing the gray scale range into a plurality of intervals uniformly, and counting the number of pixel points of which gray scale values fall into each interval; and determining the ratio of the number of the pixels corresponding to the interval with the largest number of the pixels to the total number of the pixels in the connected domain as the gray density value of the connected domain of the second target.
9. The method for monitoring the safe production based on the video data according to claim 1, wherein the step of obtaining the second key frame based on the determination amount that the low confidence target is the key target comprises the following specific steps: if the judgment amount of at least one second target which is a key target in the current first video frame is larger than a preset second judgment amount threshold, judging the first video frame as a second key frame, and determining a fourth preset number of video frames after the first video frame as the second key frame.
10. A video data based safety production monitoring system, comprising the following modules: The acquisition module is used for acquiring video frame data of a coal mine site; The analysis module is used for detecting the video frame data by adopting a target detection network and acquiring the confidence coefficient of each target in each video frame; acquiring the confidence level of each target in each video frame by using the confidence level of each target in each video frame; Determining a high confidence target and a low confidence target in each video frame based on the confidence level of each target in each video frame; Acquiring the judgment quantity of each video frame as a key frame according to a high confidence target in each video frame; Screening a first key frame from all video frames based on the judgment amount that each video frame is a key frame, wherein the first key frame is the key frame screened for the first time from all video frames; acquiring a judgment quantity of which the low confidence coefficient target is a key target by utilizing the low confidence coefficient targets in the rest video frames except the first key frame in all the video frames; Acquiring a second key frame based on the judgment amount of which the low confidence level target is a key target, wherein the second key frame is a key frame screened out of all video frames again; and the transmission module is used for compressing and transmitting the first key frame data and the second key frame data.

Description

Safety production monitoring method and system based on video data Technical Field The invention relates to the technical field of video compression, in particular to a safety production monitoring method and system based on video data. Background In the field of coal mine safety production monitoring, a video monitoring system is an important means for guaranteeing underground operation safety. However, the bandwidth of the downhole network is often severely limited, and it is difficult to achieve real-time stable transmission of full-high-definition video. To solve this problem, selective video transmission is a necessary choice. At present, a video transmission method based on deep learning target identification is mainly adopted, namely, a target detection network (such as YOLO) is deployed, a video stream is analyzed in real time, and only when an interested target (such as a person or a vehicle) is identified, the uploading of video data is triggered, so that the bandwidth utilization efficiency is improved. However, such methods still have significant drawbacks in a practical complex coal mine monitoring scenario. The transmission decision is excessively dependent on a single confidence index output by the target detection network, and the index mainly reflects the probability of the existence of the target and does not fully consider the problem of identification reliability in the underground environment. Especially when facing the high risk targets with fuzzy characteristics and uncertain morphology caused by smoke, the detection confidence is low and the fluctuation is large. If filtering is performed only according to a fixed threshold, the targets with low confidence but high risk are easily misjudged as background noise and discarded, so that serious safety warning is missed, and the problems of low monitoring reliability and low transmission efficiency of the existing method are caused. Disclosure of Invention The invention provides a video data-based safety production monitoring method and a video data-based safety production monitoring system, which are used for solving the existing problems. The invention relates to a safety production monitoring method and a system based on video data, which adopts the following technical scheme: One embodiment of the present invention provides a video data-based safety production monitoring method, which includes the steps of: Collecting video frame data of a coal mine site; Detecting the video frame data by adopting a target detection network, and acquiring the confidence coefficient of each target in each video frame; acquiring the confidence level of each target in each video frame by using the confidence level of each target in each video frame; Determining a high confidence target and a low confidence target in each video frame based on the confidence level of each target in each video frame; Acquiring the judgment quantity of each video frame as a key frame according to a high confidence target in each video frame; Screening a first key frame from all video frames based on the judgment amount that each video frame is a key frame, wherein the first key frame is the key frame screened for the first time from all video frames; acquiring a judgment quantity of which the low confidence coefficient target is a key target by utilizing the low confidence coefficient targets in the rest video frames except the first key frame in all the video frames; Acquiring a second key frame based on the judgment amount of which the low confidence level target is a key target, wherein the second key frame is a key frame screened out of all video frames again; The first key frame data and the second key frame data are compressed and transmitted. Further, the step of obtaining the confidence level of each target in each video frame by using the confidence level of each target in each video frame comprises the following specific steps: for a current target in a current video frame, acquiring a confidence entropy value and a confidence mean value of the target in a first preset number of video frames before the current frame; Subtracting a preset confidence coefficient threshold value from the confidence coefficient mean value to obtain a difference value, multiplying the difference value by the reciprocal of the confidence coefficient entropy value, and calculating to obtain the confidence coefficient of the current target in the current video frame; And obtaining the confidence level of each target in each video frame. Further, the obtaining the confidence entropy value of the target in the first preset number of video frames before the current frame includes the following specific steps: acquiring the confidence coefficient of the target in a first preset number of video frames before the current frame, and sequencing according to a time sequence to obtain a confidence coefficient sequence; Dividing the confidence coefficient range into a plurality of equal-width interva