Optical Flow Technique Research And Application For Complex Environment

Optical flow is an important technology of motions estimation, it represents the relative motion information for each pixel, it can not only used in object recognition and object tracking in static scenery,but also in dramatic scenes too. This paper is aimed at resolving the object recognition and object tracking in complex environment using optical flow technologies.In the first, we introduces some common method on object recognition and object tracking, then we made some researching on optical recognition method existed, and we compared all the results for each optical method.The existing optical method has a great detection results in the simple backgrounds, but the traditional methods has some limits under a complex environment. An object detection algorithm based on accumulated discrepancy of optical flow and video windows segmentation is proposing in the paper. According to segmenting the video windows and calculating the optical flow discrepancy matrix, the original video window is divided into moving object view, static view and fluctuation interference view, and merging the moving object view can get the moving object. The experiment verification shows that this method has a effective testing result in object detection under complex environment.According to the inaccurate of optical estimation in feature points under illumination variation or shadow, we proposed a combination method that joint the conditional number of Hessian matrix and pyramid optical flow method, this can eject inaccurate feature points, resolving the problems of inaccurate of optical estimation, enhancing the accuracy of optical tracking.In order to solving the object tracking in short times caused of of target occlusion, shadow, camera shivering, object disappearing in horizon as well as image scaling variations, we proposed making the optical tracking and machine study joint together, the proposed method can track the interested object in a long time.