The Research Of Video Dynamic Targets Detection Based On The Moving Vehicle

This paper mainly describes the visual inspection on the vehicles of movement.Firstly, it introduces the hardware platform to implement detection; Secondly, afterobtaining an image, the hardware platform preprocesses the image by affinetransformation and image filtering; Then it does the regional or edge segmentation forthe target, and two techniques has its advantage and disadvantage in differentinstitutions; After obtaining the target region, it needs to be described and identified bythe means of the shape and the description of the skeleton; That extracts importantfeature of the target image to provide information for subsequent identification. Finally,the mutative target and background image are classified and the improved integrationassay is used in the classified image pixels.This detection method is mainly proposed against some interference problems thattranslational motion of the camera and moving vehicles jitter cause. Specifically, opticalflow algorithm based on the SAD is used to generate a stabilized video by removingunwanted the interference of translational camera motions as well as vehicles jitter.Ultimately, we get the correct detection of video dynamic target on the moving vehicle.Further, SAD algorithm is mainly against the detection of moving region. It uses thisfeature to extract the desired moving target area. The optical flow only calculates themoving target region. So, it could overcome the problem of large computation andreal-time poor and greatly improve the real-time by saving a lot of the space of datastorage and time. The experimental results show that video dynamic target detection inmoving vehicles is very successful.