Research On Multi-camera Array Image Processing System And Technology Of Object Detection

Nowadays,video surveillance technology is developing rapidly and be applied widely in the fields of road traffic,safety monitoring,criminal investigation and others.With the continuous development of image sensors,the quality of camera image on video surveillance keeps increasing,while the cost greatly reduces,which lay the foundation for the development of multi-camera array system.Compared to single-camera system,multi-camera system has the characteristics of wider field of view,more access to information and better reliability.It is the direction of video surveillance development.Multi-camera array image processing system consists of several sub cameras and it is based on the technology of single-camera image processing.In the field of object detection,the multi-camera system needs to synthesize the object detection results of a single camera to complete the handover among multiple cameras.It also achieves the control and coordination of multiple cameras.This thesis is based on the practical application to research and realize the multicamera array object detection system.According to the requirements of application scene,this thesis analyses the system function and completes the system design.The DSP camera based on DM8127 is used to form the camera array.Starting from the field angle indicators,this thesis determines the camera arrangement according to the experiment and uses a combination of cameras to expand the field of view.The system uses a distributed design.Single camera is used to detect object after video images capture,and then transports the video images and object information to the central processing unit-PC machine.PC completes the handover of object among the multiple cameras.It also controls the work status and configuration information of cameras and video image display.This thesis designs and produces the camera based on DM8127,then completes debugging and system shift.The system uses multiple cameras to compose multi-camera array,and PC client to complete the entire construction.Aiming at the false detection problem of the Gaussian modeling background difference method in light mutation environment,this thesis uses a fusion method of Gaussian modeling background difference method and an improved five-frame method,which can be adapt to the light mutation environment rapidly.To improve the traditional three-frame difference method,a five-frame difference method is proposed.In the fusion algorithm,a certain strategy is used to determine whether the current frame has a light mutation,to adaptively select the current target detection method and initialize the Gaussian background model when necessary.The algorithm is used to make up the deficiency of Gaussian modeling background difference method in light mutation and superior to single algorithm and able to meet the application requirement in this thesis.As a control and processing center,PC receives the object information from camera detection,using a method based on the field of vision to complete the multi-camera target handover.MFC programming is used to complete the PC client to achieve the interaction between human and machine,thus controlling the camera working status,parameter configuration,and the video display.