Swimming pool drowning behavior identification method based on video time sequence feature analysis

Abstract

The invention belongs to the technical field of video monitoring computer image processing, and relates to a swimming pool drowning behavior identification method based on video time sequence feature analysis, which comprises the following steps: S1, constructing a data set; s2, preprocessing the image; s3, constructing a neural network model, including constructing a feature extraction part, constructing a space semantic modulation module, constructing a time semantic modulation module, constructing a feature fusion module, and constructing a behavior identification module; s4, training the constructed neural network model; s5, preprocessing the collected swimming behavior video of the target person, inputting the trained neural network model, and judging whether the behavior belongs to a drowning behavior; according to the method, spatial information features and time sequence information features of human behaviors are fused, image information can be processed in real time, and the constructed neural network model is high in calculation convergence speed, high in generalization ability and high in robustness; the method is ingenious in conception, and the accuracy of drowning behavior recognition reaches 90% or above.

Description

Technical field: [0001] The invention belongs to the technical field of video monitoring and computer image processing, and relates to computer vision algorithms, in particular to a swimming pool drowning behavior identification method based on video timing feature analysis. Make effective early warning to reduce the occurrence of drowning. Background technique: [0002] Along with the development of society and the raising of people's living standard, the sports of this whole body exercise of swimming are more and more subject to people's welcome. However, because it is carried out in water, beginners often cause choking and even drowning accidents because they cannot breathe freely and move freely. In addition, swimmers may experience muscle spasms, collisions, and physical exhaustion in the water, all of which may lead to drowning events. Once a drowning incident occurs, it will directly affect the life safety of swimmers. At present, research on identifying drowning b...

AI Technical Summary

Problems solved by technology

[0004] (2) At present, most of the human behavior and action recognition based on monitoring adopts the technology of static image analysis, manually marking a large number of foreground objects, and then using the 2D CNN network framework to extract the foreground object frame of the image. Although the calculation cost is low, it cannot capture the behavior. The time relationship of the target, and the lack of the behavior characteristics of the target, makes the important behavior characteristics missing seriously, resulting in low accuracy of behavior recognition; especially for complex environments such as indoor swimming pools, the installation angle of the monitored equipment, lighting, water surface fluctuations, pool Due to the influence of various factors such as the shaking of the bottom pattern, the position of the moving target is sometimes above the water and sometimes under the water during the swimming process.

[0005] (3) The 3D CNN network framework is proven to be effective in spatiotemporal modeling, but it cannot capture enough information contained in the video. By increasing the optical flow information, the performance can be significantly improved compared with the single-flow network framework, but the optical flow calculation introduced are too expensive to deploy on real-world applications;

[0006] (4) Part of the designed 3D network is affected by the size of the data set, resulting in the problems of overfitting and slow convergence of the trained model

Images