CN-122023438-A - Protection device mobile monitoring method, device and medium based on color segmentation

CN122023438ACN 122023438 ACN122023438 ACN 122023438ACN-122023438-A

Abstract

The invention provides a color segmentation-based protection device mobile monitoring method, device, equipment and medium, wherein the method comprises the steps of collecting an image which is in a normal and good state and has good light, taking the image as a front background image, acquiring the image of a target area where the protection device is located in real time, performing scene filtering on the image acquired in real time by adopting black-and-white image judgment and moving target or pedestrian detection to obtain a screened real-time image, respectively performing semantic segmentation on the front background image and the screened real-time image by adopting a BiSeNet depth learning segmentation network, extracting at least one protection device area with preset color, generating a background segmentation mask and a real-time segmentation mask, calculating the difference between the real-time segmentation mask and the background segmentation mask, and judging whether the protection device area is in a missing or displacement state so as to solve the problems of poor real-time performance, high omission rate and dependence on manpower in the existing protection device monitoring mode.

Inventors

TAN HENG
TIAN LEI
WANG FEI

Assignees

中建材信息技术股份有限公司
中建材信云智联科技有限公司
中建材信息科技有限公司
中建材信云智联科技有限公司北京分公司
中建材信云智联科技(北京)有限公司

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (10)

1. The color segmentation-based protection device mobile monitoring method is characterized by comprising the following steps of: collecting an image with good light of the protective device in a normal and sound state, and taking the image as a front background image; Acquiring an image of a target area where the protective device is positioned in real time, and performing scene filtering on the image acquired in real time by adopting black-and-white image judgment and moving target or pedestrian detection to obtain a screened real-time image; Performing semantic segmentation on the front background image and the screened real-time image by adopting BiSeNet deep learning segmentation network, extracting at least one protection device area with preset color, and generating a background segmentation mask and a real-time segmentation mask; and calculating the difference between the real-time segmentation mask and the background segmentation mask, and judging whether the protective device area is missing or displaced.
2. The color segmentation based guard movement monitoring method as set forth in claim 1, further comprising the steps of: and when the difference is larger than a preset minimum threshold value and the comparison of the real-time segmentation mask and the background segmentation mask meets the alarm condition, determining an alarm object and pushing alarm information.
3. The method for mobile monitoring of a color segmentation-based guard according to claim 1, wherein the BiSeNet deep learning segmentation network comprises a spatial path, a context path, and a feature fusion module; a space path for reserving space detail information of the image and generating a high-resolution characteristic image with 1/8 size of the input image through 3 convolution layers with the step length of 2; The context path is used for extracting context semantic information of the image, takes a lightweight network as a backbone network and is additionally provided with a global average pooling layer; And the feature fusion module fuses the features of the space path and the context path through feature splicing, batch normalization and weighting.
4. The method for monitoring the movement of the guard device based on the color segmentation as set forth in claim 1, wherein the scene filtering is performed on the image acquired in real time by adopting black-and-white image judgment and moving object or pedestrian detection to obtain a filtered real-time image, specifically comprising the following steps: Firstly, judging whether an image acquired in real time is a black-and-white image, if so, ending the detection, and if not, entering the next step; Then, whether a moving object or a pedestrian exists in the image acquired in real time is detected, if so, the moving object or the pedestrian is judged to be a working scene, the detection is ended, and if not, the next step is carried out until the screened real-time image is obtained.
5. The method for monitoring movement of a guard based on color segmentation according to claim 1, wherein the calculating the difference between the real-time segmentation mask and the background segmentation mask comprises the steps of: And counting the pixel difference quantity and the position deviation of the target color area in the real-time segmentation mask and the background segmentation mask by adopting a pixel level comparison mode.
6. The color segmentation based guard movement monitoring method as set forth in claim 1, further comprising the steps of: And collecting the images of the protective device under different scenes and different states, and carrying out pixel level division mask labeling on the characteristic color areas in the collected images for training BiSeNet the deep learning division network.
7. Protection device removes monitoring devices based on colour segmentation, characterized by, include: The background image acquisition module is used for acquiring images with good light rays of the protection device in a normal and intact state and taking the images as a front background image; the real-time image acquisition module is used for acquiring an image of a target area where the protective device is positioned in real time, and performing scene filtering on the image acquired in real time by adopting black-and-white image judgment and moving target or pedestrian detection to obtain a screened real-time image; The semantic segmentation module is used for carrying out semantic segmentation on the front background image and the screened real-time image by adopting BiSeNet deep learning segmentation network respectively, extracting at least one protection device area with preset color, and generating a background segmentation mask and a real-time segmentation mask; And the monitoring module is used for calculating the difference between the real-time segmentation mask and the background segmentation mask and judging whether the protective device area is missing or displaced.
8. The color segmentation based guard mobile monitoring device of claim 7, further comprising an alarm module configured to determine an alarm object and push alarm information when the difference is greater than a preset minimum threshold and the real-time segmentation mask and the background segmentation mask both satisfy an alarm condition.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the color segmentation based guard movement monitoring method according to any of claims 1 to 6 when executing the computer program.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the steps of the color segmentation based guard movement monitoring method according to any one of claims 1 to 6.

Description

Protection device mobile monitoring method, device and medium based on color segmentation Technical Field The invention relates to the technical field of intelligent safety monitoring of industrial/construction sites based on computer vision, in particular to a method, a device, equipment and a medium for monitoring the movement of a protective device based on color segmentation. Background In industrial working scenarios such as factories, construction sites, etc., the protection device is a core infrastructure for guaranteeing personnel safety and equipment stable operation, and covers various types such as physical isolation devices (e.g. metal/mesh fences, movable guard fences), equipment protection structures (e.g. machine housings, rotating part protecting covers), safety identification systems (e.g. warning signs, reflective warning tapes), etc. The device can effectively avoid safety accidents such as cutting injury and crush injury by blocking dangerous mechanical parts from contacting, isolating harmful substances from leaking, warning dangerous areas and the like, reduce misoperation and external interference of equipment, reduce maintenance cost, and simultaneously help enterprises to meet the requirements of safety production regulations and maintain good operation order and enterprise reputation. Because of the key function of the protection function, the protection device is always in an intact and in-place state, which is a precondition for guaranteeing the safety of the operation environment. However, the state monitoring of the existing protection device is mainly dependent on traditional modes such as manual regular inspection and employee visual inspection, and the like, so that on one hand, the manual inspection efficiency is low, the real-time monitoring requirements of large-area operation scenes such as factories and construction sites are difficult to adapt, and the abnormal conditions such as movement and damage of the protection device are easily caused due to the influence of factors such as responsibility and fatigue degree of personnel, on the other hand, the abnormal instant detection and early warning cannot be realized in the traditional monitoring mode, if the protection device is removed and is not recovered in time or is not maintained in time after damaged, safety accidents can be possibly caused, and enterprises are further caused to face risks such as legal responsibility, economic reimbursement and reputation damage. In addition, the simple image monitoring adopted by part of scenes lacks intelligent analysis capability, still needs manual review judgment, and cannot fundamentally solve the problem of insufficient instantaneity and accuracy. Therefore, a method for monitoring the movement of the protection device based on color segmentation is needed to solve the technical problems of poor real-time performance, high omission factor and dependence on manpower in the existing protection device monitoring method. Disclosure of Invention In order to overcome the problems in the related art, the present disclosure provides a method, a device, equipment and a medium for monitoring the movement of a protection device based on color segmentation, so as to solve the technical problems of poor real-time performance, high omission factor and dependence on manpower in the existing protection device monitoring mode in the related art. One or more embodiments of the present disclosure provide a method for monitoring movement of a protection device based on color segmentation, including the following steps: collecting an image with good light of the protective device in a normal and sound state, and taking the image as a front background image; Acquiring an image of a target area where the protective device is positioned in real time, and performing scene filtering on the image acquired in real time by adopting black-and-white image judgment and moving target or pedestrian detection to obtain a screened real-time image; Performing semantic segmentation on the front background image and the screened real-time image by adopting BiSeNet deep learning segmentation network, extracting at least one protection device area with preset color, and generating a background segmentation mask and a real-time segmentation mask; and calculating the difference between the real-time segmentation mask and the background segmentation mask, and judging whether the protective device area is missing or displaced. Preferably, the method further comprises the following steps: and when the difference is larger than a preset minimum threshold value and the comparison of the real-time segmentation mask and the background segmentation mask meets the alarm condition, determining an alarm object and pushing alarm information. Preferably, the BiSeNet deep learning segmentation network comprises a spatial path, a context path and a feature fusion module; a space path for reserving space detail information of the i