The Research On Intelligent Assistant Lifesaving System Of Swimming Pool Based On Computer Vision

Alone with the continuous applications of video surveillance system in underwater life-saving areas.Underwater human target tracking algorithm based on computer vision has been widely used in real-time tracking of drowning target.The diversity of human target characteristics and the complexity of the environment in which they are located become the inevitable constraints in the tracking of human targets in the actual tracking application.Therefore,the research on moving object detection algorithm has great significance.Aiming at the complexity of the underwater environment in the pool.This paper makes a deep research on moving object detection algorithm,the specific work is as follows:(1)Theoretical analysis and experimental simulation are carried out for three common background modeling algorithms such as: CODEBOOK,VIBE and GMM.The accuracy and time efficiency of the three algorithms are analyzed.By setting up a sliding window,by calculating a mixed Gaussian model of a historical image with a certain step,the accuracy of the traditional mixed Gaussian model is enhanced.The parameter updating part of the mixed Gaussian model is improved.(2)The characteristics of reflection,and ripple in the swimming pool are analyzed.The adaptive threshold method such as: OTSU is used to solve the influence of reflection.The effect of ripple is solved by subdividing the grayscale at the bottom of the pool.And the value of ripple pixels are replaced by the median value in the gray level.(3)The influence of drowning target size and the distance between drowning target and underwater camera on the efficiency of drowning detection is analyzed.The efficiency of drowning detection is improved by setting the number of search points in the contour detection.The improved algorithm is used for underwater moving target detection.And a systematic experimental analysis is carried out.(4)The background modeling algorithm and target detection algorithm based on sliding window of mixture Gaussian model are encapsulated into DLL dynamic libraries for system calls.The design of underwater detection system is explained.The overall scheme of the system design,the hardware devices used in the system and the human-machine interface are introduced in detail.