Research And Realization Of Real-time Moving Object Detection Algorithm Based On DM642

Real time moving object detection based on image sequences is one of the hot topics in the field of digital image processing.With the popularization of information and intelligence,it is widely used in the field of video surveillance related industries.Due to the problems of large amount of data processing and complex algorithm for moving objects detection,there are many difficulties in the aspect of integrity and real time at present.This thesis chooses TMS320DM642 processor which made by TI company as the core of the hardware platform,focusing on the hardware function module and the DSP/BIOS system that embedded in the integrated development environment of CCS.Designed an application of video image processing software which is based on R5 framework.Realized the collection,processing,display and process video through the task thread.In the specific implementation of target detection algorithm,the author of the paper make background modeling algorithm become C language and transplanted to the hardware platform based on DM642 successfully,and achieve real-time target detection requirements by optimizing the overall software and hardware on the system.In the algorithm,this thesis studies the background modeling methods which is commonly used in background subtraction,such as statistical method,average method.For the interference problem that caused by the environmental factors,such as illumination changes,background disturbance,proposed the dynamic threshold method based on statistical average,which do the subsequent processing mainly by increasing obtain the dynamic threshold in the image information and use the morphological filtering and logic operation,and to effectively eliminate environmental factors led to the use of the optimized background updating algorithm.The calculation is simple,and it can meet the real-time requirement of DM642 hardware platform.In addition,based on the study of the characteristics of the texture to the shadow,combined with the gray information of the video image,this paper proposes a texture based moving object detection algorithm.After processing the video image processing,calculate the LBP encoding texture histogram and gray information vector of each pixel in each part,and using the principle of mixed Gauss model,as the input parameters to establish the background model and detect the target area,in order to eliminate the influence of noise and shadow detection results in.The experimental results of hardware platform of DM642 based shows that the system detects out of the target area completely,and overcome the complex effects of environmental factors such as light,shadows and other causes,but also can meet the real-time requirements,the system has good robustness.