Belt running state detection method

Abstract

The invention belongs to the technical field of image recognition, and particularly relates to a belt running state detection method, which comprises the following steps of: inputting a video stream, and extracting video frame angular points; eliminating wrong tracking angular points; eliminating the tracking angular points exceeding the longest distance; removing non-belt background angular points; judging the start and stop of the current frame; and eliminating abnormal results based on time sequence continuity statistics. The method is improved based on the optical flow and has the advantage of being high in speed, and the average processing speed of 1080p images is higher than 1000 FPS and far higher than that of other belt starting and stopping methods. The method has good adaptability to no-load and normal belts, respective processing for different conditions is not needed, the algorithm process is simple, and the accuracy is high. Moreover, the method does not need the support of a hardware GPU, and has the advantage of simple deployment.

Description

Technical Field

[0001] The invention belongs to the technical field of image recognition, and in particular relates to a belt running state detection method. Background Art

[0002] At present, conveyor belts play a vital role in industrial fields such as coal mines, and are used to transport raw materials, coal and other materials from mining areas to production areas or storage areas. In order to ensure the safe operation and efficient management of conveyor belts, conveyor belt start-stop detection technology has emerged. The application of conveyor belt start-stop detection technology not only improves the operating efficiency and safety of the transportation system, but also provides strong support for the intelligent management of mining enterprises.

[0003] The start and stop status of the belt includes two labels: normal operation and stop operation. This technology is usually divided into hardware-based methods and software-based methods. The hardware-based method installs sensors and monitoring equipment to monitor the start, stop status and running speed of the transport belt in real time. At the same time, it can also promptly alarm and automatically shut down the belt when an abnormal situation occurs. However, this method has disadvantages such as high hardware cost, complex installation and deployment, and difficulty in promotion. AI-based visual detection methods are applied. With the help of AI models for data analysis, the start and stop status of the belt can be detected quickly and accurately, avoiding equipment failures or safety accidents caused by failure to timely detect abnormal belt operation. Although the current mainstream deep learning-based methods based on vision have high detection accuracy, they have disadvantages such as slow running speed, high computing resource consumption, the need for hardware GPU support, and complex algorithm flow. Summary of the invention

[0004] In view of the technical problems of the above method, such as slow running speed, high computing resource consumption, requirement for hardware GPU support and complex algorithm flow, the present invention provides a belt running status detection method.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A method for detecting a belt running state comprises the following steps:

[0007] S1, input video stream and extract video frame corner points;

[0008] S2, remove the wrong tracking corner points;

[0009] S3, remove the tracking corner points that exceed the longest distance;

[0010] S4, remove the background corner points that are not belts;

[0011] S5, judging the start and stop of the current frame;

[0012] S6. Eliminate abnormal results based on time series continuity statistics.

[0013] The method for extracting the corner points of the video frame in S1 is:

[0014] For any input video stream, when the video is input for the first time, the current video frame image is extracted, and the Harris corner points of the current video frame are extracted based on the optical flow method and recorded;

[0015] When the video is not input for the first time, the Harris corner points of the previous frame are combined with the current input video frame to extract the Harris corner points of the current video frame. Then the same processing flow is reversed to extract the Harris corner points of the previous frame based on the Harris corner points of the current frame and the previous input video frame. After forward and reverse processing, the Harris corner points of the previous frame and the Harris corner points of the current frame are obtained.

[0016] The method for eliminating erroneous tracking corner points in S2 is: calculate the difference d=max(abs(xi-xj),abs(yi-yj)) between the Harris corner point coordinates (xi,yi) of the current frame and the Harris corner point coordinates (xj,yj) of the previous frame, take all points in the point set with a distance d<1 as the correctly tracked points, and discard the erroneous points directly.

[0017] The method for removing tracking corner points exceeding the longest distance in S3 is: define the longest tracking pixel distance as d track = 15, then remove the points in the previous frame and the current frame whose distance is greater than the longest tracking pixel distance d. track points to obtain a new set of points.

[0018] The method of removing the background corner points that are not belts in S4 is: define the vector distance d of the optical flow optical =5, calculate the distance between the corresponding coordinate points of the point set of the previous frame and the current frame, and set the distance less than d optical The points are removed, thereby eliminating the background corner points that are not belts.

[0019] The method for judging the start and stop of the current frame in S5 is: judging the number of Harris corner points of the current frame finally obtained, if it is greater than 1, it proves that the belt is moving and belongs to the start state, and the output belt state s=1; otherwise, it proves that the belt is in the stop state, and the output belt state s=0.

[0020] The method for eliminating abnormal results based on time series continuity statistics in S6 is: repeat the calculation process of S1-S5 for 25 adjacent frames of video images to obtain a sequence set S={s0, s1, s2...s 24}, based on the idea of ​​temporal continuity statistics, the sequences in the set are summed to obtain the start and stop state S of the belt in the final current frame video out =sum(S), which can greatly improve the robustness of start-stop detection and reduce the interference of outliers.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] The method of the present invention is improved based on optical flow and has the advantage of high speed. For 1080p images, the average processing speed is higher than 1000FPS, which is much higher than other belt start and stop methods. The method of the present invention has good adaptability to both unloaded and normal belts, and does not need to be processed separately for different situations. The algorithm flow is simple and the accuracy is high. In addition, the method of the present invention does not require the support of hardware GPU and has the advantage of simple deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0024] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0025] Figure 1 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. These descriptions are only to further illustrate the features and advantages of the present invention, rather than to limit the claims of the present invention. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0027] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0028] In this embodiment, if Figure 1 As shown, for any input video stream, when the video is input for the first time, the current video frame image is extracted, and the Harris corner points of the current video frame are extracted based on the optical flow method and recorded.

[0029] When the video is not input for the first time, the Harris corner points of the previous frame are combined with the current input video frame to extract the Harris corner points of the current video frame. Then the same processing flow is reversed, based on the Harris corner points of the current frame and the previous input video frame, the Harris corner points of the previous frame are reversely extracted. After forward and reverse processing, the Harris corner points of the previous frame and the Harris corner points of the current frame can be obtained.

[0030] Calculate the difference d = max(abs(xi-xj), abs(yi-yj)) between the Harris corner point coordinates (xi, yi) of the current frame and the Harris corner point coordinates (xj, yj) of the previous frame, and take all points in the point set with a distance d < 1 as the correctly tracked points, and discard the wrong points directly.

[0031] Define the longest tracking pixel distance as d track = 15, then remove the points in the previous frame and the current frame whose distance is greater than the longest tracking pixel distance d. track points to obtain a new set of points.

[0032] Define the vector distance d of optical flow optical =5, calculate the distance between the corresponding coordinate points of the point set of the previous frame and the current frame, and set the distance less than d optical The points are removed, thereby eliminating the background corner points that are not belts.

[0033] The number of Harris corner points of the current frame is determined. If it is greater than 1, it proves that the belt is moving and is in the start state, and the belt state s=1 is output. Otherwise, it proves that the belt is in the stop state, and the belt state s=0 is output.

[0034] Repeat the above calculation process for 25 consecutive video frames to obtain the sequence set S = {s0, s1, s2...s 24 Based on the idea of ​​temporal continuity statistics, the sequences in the set are summed to obtain the start and stop state S of the belt in the final current frame video. out =sum(S), which can greatly improve the robustness of start-stop detection and reduce the interference of outliers.

[0035] Only the preferred embodiments of the present invention are described in detail above, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention, and various changes should be included in the protection scope of the present invention.

Claims

1. A method for detecting the running state of a belt, It is characterized in that The following steps are involved: S1, input video stream and extract video frame corner points; S2, remove the wrong tracking corner points; S3, remove the tracking corner points that exceed the longest distance; S4, remove the background corner points that are not belts; S5, judging the start and stop of the current frame; S6. Eliminate abnormal results based on time series continuity statistics.

2. A belt running status detection method according to claim 1, It is characterized in that The method for extracting the corner points of the video frame in S1 is: For any input video stream, when the video is input for the first time, the current video frame image is extracted, and the Harris corner points of the current video frame are extracted based on the optical flow method and recorded; When the video is not input for the first time, the Harris corner points of the previous frame are combined with the current input video frame to extract the Harris corner points of the current video frame. Then the same processing flow is reversed to extract the Harris corner points of the previous frame based on the Harris corner points of the current frame and the previous input video frame. After forward and reverse processing, the Harris corner points of the previous frame and the Harris corner points of the current frame are obtained.

3. A belt running status detection method according to claim 1, It is characterized in that The method for eliminating erroneous tracking corner points in S2 is: calculate the difference d=max(abs(xi-xj),abs(yi-yj)) between the Harris corner point coordinates (xi,yi) of the current frame and the Harris corner point coordinates (xj,yj) of the previous frame, take all points in the point set with a distance d<1 as the correctly tracked points, and discard the erroneous points directly.

4. A belt running status detection method according to claim 1, It is characterized in that The method for removing tracking corner points exceeding the longest distance in S3 is: define the longest tracking pixel distance as d track = 15, then remove the points in the previous frame and the current frame whose distance is greater than the longest tracking pixel distance d. track points to obtain a new set of points.

5. A belt running status detection method according to claim 1, It is characterized in that The method of removing the background corner points that are not belts in S4 is: define the vector distance d of the optical flow optical =5, calculate the distance between the corresponding coordinate points of the point set of the previous frame and the current frame, and set the distance less than d optical The points are removed, thereby eliminating the background corner points that are not belts.

6. A belt running status detection method according to claim 1, It is characterized in that The method for judging the start and stop of the current frame in S5 is: judging the number of Harris corner points of the current frame finally obtained, if it is greater than 1, it proves that the belt is moving and belongs to the start state, and the output belt state s=1; otherwise, it proves that the belt is in the stop state, and the output belt state s=0.

7. A belt running status detection method according to claim 1, It is characterized in that The method for eliminating abnormal results based on time series continuity statistics in S6 is: repeat the calculation process of S1-S5 for 25 adjacent frames of video images to obtain a sequence set S={s 0 ,s 1 ,s 2 ……s 24 }, based on the idea of ​​temporal continuity statistics, the sequences in the set are summed to obtain the start and stop state S of the belt in the final current frame video out =sum(S), which can greatly improve the robustness of start-stop detection and reduce the interference of outliers.

Images