Research On Underwater Human Detection Based On Improved Background Difference Method

At present,the number of swimming enthusiasts is increasing year by year.Meanwhile,the drowning accident caused by the swimming is also rising.Although each regular swimming places will be equipped with the corresponding rescue personnel,sometimes rescue personnel will cause mental fatigue or due to environmental factors such as timely detection of drowning,which leads to the occurrence of drowning accidents.Therefore,the intelligent monitoring system of swimming pool emerges as the times require,in which the underwater human detection technology is the core of the intelligent monitoring system.Affected by water under complicated environment where water as medium will absorb light energy and water particles will form a scattering of light.These reasons make the image interference with high SNR and image color distortion.There is also a dynamic interference in the water such as the halo of the dynamic interference.These disturbances make underwater images in the human body detection,resulting in inaccurate underwater detection.The main contents are as follows:(1)Image denoising.Underwater images contain a large amount of noise mainly from the underwater environment(water particles,water ripples)generated by the random noise which is usually seen as Gauss white noise.According to the characteristics of the underwater image noise and high SNR,the wavelet threshold denoising algorithm is used to denoise.The general function of wavelet and the compromise function of soft and hard threshold are improved.The simulation results show that the improved method can get better denoising results.(2)Image enhanced.Underwater optical characteristics are analyzed,the underwater image becomes blurred due to the scattering of light from water.So the image is moderately enhanced in the background modeling and foreground difference.In order to improve the contrast and clarity of underwater images,an adaptive multi-scale Retinex algorithm(AMSR)is proposed.The simulation results show that the algorithm can improve the contrast and resolution of underwater images,and thus provide a basis for improving the accuracy of human body detection.(3)Underwater human detection.Underwater human detection is to detect the location of the human body in the water.Therefore,it is necessary to improve the background subtraction method to adapt to the need of underwater human body detection.Underwater body detection needs to have good real-time,so the establishment of the background by the mean background modeling method;at the same time compared to the gray image,the color image contains more color information,using the advantage of background subtraction threshold was improved to eliminate static interference in underwater image.And the RGB color difference method is used to eliminate the dynamic interference of underwater images with different color information.Through the simulation experiment,the underwater target is proved to be able to get a good effect on underwater image and even the color image is more serious.In this paper,human detection under water through the acquired image sequence of the auxiliary lifesaving system actual pool intelligence,the simulation results show that the improved background subtraction algorithm can accurately detect the underwater images of the human body.