Climbing unmanned escalator identification system based on portable deployment and control ball

Abstract

The invention relates to the technical field of computer vision, in particular to a climbing unmanned escalator recognition system based on a portable deployment and control ball, and the specific operation process is as follows: S1, panoramic snapshot; s2, positioning a person and a ladder target; s3, judging a position relation; s4, performing local snapshot; s5, human body posture recognition; s6, excluding abnormal personnel; according to the system, intelligent snapshot of an actual operation scene is achieved by controlling movement of the camera and zoom control, the recognition effect on the climbing behavior and the escalator behavior of an operator can be improved, and the problem of low accuracy of a target detection and behavior recognition algorithm at a long distance is solved; according to the system, a human body posture estimation algorithm and a target detection algorithm based on RGB images are combined, an existing action recognition scheme is replaced, the recognition effect on climbing behaviors and escalator behaviors of operators can be improved, and the equipment dependence degree of the algorithms is reduced.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to an unmanned escalator identification system based on a portable cloth control ball. Background technique [0002] Guangdong Power Grid Corporation proposes to create an on-site smart security business card with visualization of the whole process. At present, grid companies have many daily operations, and on-site safety management and control are difficult. On-site safety management and control is not only inefficient but also unable to provide real-time reminders and alarms for on-site operation violations or dangerous behaviors. At present, Guangdong Power Grid alone has as many as 4,000-6,000 power operating points per day, and as many as 40,000 to 60,000 people working on the site every day. The operating points are many and wide, and on-site safety control is difficult. Faced with this severe situation, relying solely on human safety inspections to conduct on-site sa...

AI Technical Summary

Problems solved by technology

[0010] 1. When the existing target detection algorithm is far away from the camera, the size of the object in the picture will be small, which will make the target detection algorithm invalid

[0011] 2. Existing action recognition algorithms rely on 3D convolutional networks, cyclic neural networks (LSTM), graph convolutional networks (GCN), etc., which require computing equipment with high computing power

Images