New Image Segmentation Algorithm Of Nonuniform Illumination Image In Video Surveillance Of The Aged

Nowadays,intelligent video surveillance systems have been widely used in transportation,finance,manufacturing and other industries.As the population ages,the video surveillance system for the elderly has also been studied in depth.The commonly used video surveillance system for the elderly mainly includes four modules: video acquisition,algorithm module,network transmission and human-computer interaction.The algorithm module can handle the identification of abnormal behaviors such as falls and sudden illnesses,which makes the video surveillance system for the elderly implements real-time monitoring,emergency alarms and so on.The algorithm module includes image preprocessing,target detection,target tracking and recognition,which makes the video surveillance system maintain high accuracy and high real-time.But most of the existing algorithms do not take into account the uneven illumination caused by multiple light sources,low-light or sudden changes in illumination,resulting in performance degradation of system severely.So,based on the research and summary of existing non-uniform illumination image enhancement algorithms and moving target detection algorithms,the image preprocessing and target detection algorithms modules of the system are improved and a better video surveillance system for the elderly under nonuniform illumination is designed in this paper.The main work of this paper is as follows.1.Aiming at the problem of block artifact and edge information loss in low illumination image enhancement algorithm based on dark channel prior,this paper optimizes the guidance image of the guided filter by means of bilateral filter and median filter,and then smooths the transmission map.This improvement not only improves the enhancement effect,but also reduces the computational complexity of the bilateral filter.2.Aiming at the problems of abrupt edge transition,loss of edge information and color imbalance caused by traditional image enhancement algorithm based on wavelet transform,an improved wavelet gray-scale stretching method is proposed,which integrates the information of neighbor wavelet coefficients and adjusts the low-frequency coefficients of the image based on the idea of Otsu algorithm,and also amplify the high-frequency coefficients except noise coefficients of different scales based on the wavelet energy function to improve the contrast of the image effectively.3.Aiming at the problem that the robustness of ViBe algorithm is not high because of the sudden change of illumination between video frames,this paper uses rgI color model by mixing rgb color model which is from RGB normalization and HSI color model,combines the rule of the change of the average value of the I component of the before and after the current frame,applies the idea of Otsu to find the critical point of sudden illumination change.And this paper uses three-frame difference method in video frames with sudden change of illumination to segment the target.In addition,the background modeling strategy and the background updating strategy in ViBe algorithm are improved to deal with the sudden change of illumination effectively and improve the accuracy and robustness of algorithms of moving target detection and segmentation.4.In this paper,a strategy combining the above non-uniform illumination image enhancement algorithm is proposed to adaptively enhance the non-uniform illumination image in the video sequence,and the improved ViBe algorithm is used to detect the preprocessed video sequence.Finally,this paper designs an intelligent video surveillance system based on Raspberry Pi and cloud server for the fall detection of the elderly,and verifies and evaluates the above algorithm.The system test shows that the algorithm module designed in this paper can effectively improve the accuracy and robustness of the elderly video surveillance system.