Smoking behavior identification method for an open space

Abstract

The invention discloses a smoking behavior identification method for an open space. The method comprises the steps of performing training by adopting a sample picture comprising a face, hands and a cigarette to obtain a first model; performing training by adopting a sample picture comprising an upper arm and a lower arm of an arm to obtain a second model; performing training by adopting a smokingbehavior time interval data set to obtain a third model; obtaining an infrared image of a to-be-identified region; determining the position of a heat source firstly; further obtaining a video sequence; inputting the video sequence into the model to obtain static similarity and dynamic similarity; and finally, comprehensively judging whether a smoking behavior exists or not. According to the method, mainly based on static picture analysis and identification when a key behavior occurs, and supplemented by key behavior occurrence time analysis, the problem that a smoking behavior in a place is not easy to judge by using smoke is solved; and dynamic similarity is combined on the basis of static similarity judgment, so that the judgment precision of the smoking behavior can be greatly improved.

Description

technical field [0001] The invention belongs to the technical field of intelligent identification, and in particular relates to a smoking behavior identification method in an open space. Background technique [0002] Smoking behavior will not only cause great harm to the health of oneself and others, but also induce many dangers and cause great risks to social development and safety. For this reason, gas stations, oil pipelines, special construction sites, public Smoking is explicitly prohibited in traffic places, forests, etc. With the progress of the times, artificial intelligence technology has been greatly developed, and the development of science and technology has also brought new methods and new ideas for better regional control of smoking behavior. [0003] For the detection and warning of smoking behavior in a large range, there are three main existing methods: [0004] Option 1. Use infrared detection thermal imaging equipment to detect high-temperature cigarette...

AI Technical Summary

Problems solved by technology

[0007] The disadvantage of Option 1 is that the detection accuracy is limited. Since this solution must use special equipment, the cost is high and the cost performance is poor. It is not suitable for large-scale detection, and it cannot effectively predict smoking behavior;

[0008] The disadvantage of Option 2 is that it is not suitable for smoking behavior detection in open areas. In the case of rapid outdoor air circulation, it is even more difficult to use smoke recognition methods to judge smoking behavior, and the missed detection rate is high;

[0009] Although Option 3 is very suitable for detecting individual smoking behavior, it is extremely difficult to accurately determine the angle of the upper arm and lower arm for different individuals and different perspectives. Therefore, Option 3 is suitable for making personal smoking cessation devices, but it is not suitable for large-scale smoking behavior detection

Images