Smoking behavior detection method in monitoring scene based on computer vision

Abstract

The invention discloses a smoking behavior detection method in a monitoring scene based on computer vision. The method comprises the steps: collecting a video image comprising a smoking behavior and avideo image of a similar behavior, such as a call, and employing a pre-designed target detection network model to detect the position of a human body of the image, using a preset cutting algorithm tocut the human body detection frame, using a preset classification network model to identify the image information of the cutting frame area, and then obtaining a smoking identification result; the method has the advantages of low cost, real-time detection, high accuracy and the like in implementation.

Description

technical field [0001] The invention relates to the technical field of target detection in deep learning, the technical field of image classification and recognition, and a set of action detection rule methods and systems. Background technique [0002] At present, under the technical conditions of video behavior understanding, human behavior detection technology can be used for smoking detection under surveillance cameras. The main technical solutions are as follows: [0003] (1) Based on the C3D (three-dimensional convolution) method, a 3D convolutional deep network is used to perform spatiotemporal modeling of the entire video and directly identify smoking behavior. [0004] (2) Based on the method of CNN-LSTM (convolutional neural network-long short-term memory neural network), CNN is used to extract the feature information of each frame in the video, and the feature information of each frame is processed by LSTM to finally identify the behavior. [0005] (3) Based on th...

AI Technical Summary

Problems solved by technology

[0009] (1) Although the methods (1), (2), and (4) mentioned above have improved the accuracy of smoking behavior detection / recognition to varying degrees, it is difficult to achieve real-time detection due to the high time and space complexity of the algorithm. , smoking detection requires a certain real-time

[0010] (2) The method of method (3) described above, due to the need to capture gesture information at close range, it is difficult to detect smoking behavior under the condition of a surveillance camera

[0011] (3) The method (5) described above has a low accuracy rate and cannot perform accurate behavior detection on the human body

Images