Safety helmet real-time detection method in electric power scene

Abstract

The invention provides a safety helmet real-time detection method in an electric power scene. The method comprises the following steps of: 1, acquiring an electric power monitoring video, and carryingout preprocessing and normalization of the electric power monitoring video; 2, inputting a video frame processed in step 1 into a PoseNet model to obtain skeleton information of a worker in the videoframe; 3, calculating the position of the safety helmet according to the skeleton information of the worker; 4, detecting the safety helmet wearing state of the worker by using a trained VGG16 model;5, judging whether the detection frame number reaches a set threshold value or not, if not, repeatedly executing the steps 1 to 4, and if yes, executing step 6; and 6, comprehensively analyzing the classification result of each frame by detecting a plurality of continuous video frames to obtain a safety helmet wearing detection result. According to the invention, the safety helmet wearing state of the worker can be precisely and quickly detected, and the method can adapt to various electric power working environments.

Description

technical field [0001] The invention relates to technical fields such as deep learning, image processing, and video analysis, and specifically relates to a method for real-time detection of safety helmets in an electric power scene. Background technique [0002] In the electric power working environment, the wearing of safety helmets is an important guarantee for safe work and production, which can effectively reduce the damage caused by accidents such as falls, falls from high altitudes, and falling objects. Therefore, in the electric work environment, the wearing of safety helmets has always been the iron rule of safe production. However, due to factors such as hot weather and affecting work, there are often staff who do not wear safety helmets for work regardless of their own safety. This behavior has huge potential safety hazards. The regulatory authorities have introduced relevant regulatory measures and dispatched specialized safety supervisors to inspect the work sit...

AI Technical Summary

Problems solved by technology

In the first method, when pedestrians are blocked, pedestrians cannot be detected directly, and subsequent helmet detection cannot be completed. At the same time, the upper body of pedestrians is directly intercepted for classification, which not only cannot be used for many special working postures, such as squatting, Lying down and waiting, and directly intercepting the upper body image will have many useless features, which will affect the classification accuracy

The second is to locate the head directly through face detection. The biggest problem is that when the person faces the camera, the head information of the pedestrian cannot be detected, so it is impossible to judge whether to wear a helmet.

Finally, both of these methods have one of the biggest problems in industrial applications, that is, they cannot achieve real-time performance.

Images