The Design And Implementation Of Moving Object Detection And Tracking

As a branch of computer vision,moving object detection and tracking have attracted wide attention in recent years.With the increasing complexity of the external environment,it is more difficult to detect and track moving object.In this paper,traditional detection and tracking methods are optimized from both static and dynamic scenes.In static scenes,the traditional Camshift detection algorithm has strong dependence on the color of object.In order to make the traditional Camshfit detection algorithm has a better adaption to environmental changes,this paper proposes an adaptive multi-feature fusion object detection method to make up for the shortcomings of the traditional Camshfit detection algorithm.Firstly,the object foreground is extracted by Gaussian mixture background modeling and frame difference methods,at the same time,the texture feature information is extracted by LBP operator,the edge feature information is extracted by Canny operator,and the color feature information is extracted by HSV histogram.Then,the probability distribution of the feature information is described by the reflection projection of the histogram,and the weight proportion is allocated by the adaptive fusion algorithm to calculate the probability distribution after fusion.Finally,we have a logically calculation between the results of the final probability distribution and the object foreground to detect the object.Aiming at the problem that the tracking window of traditional Camshift algorithm is small(the object exceeds the tracking window),this paper combines the linear prediction algorithm with the traditional Camshift tracking algorithm to adjust the tracking window in real time by linear prediction of the tracking window,and realizes update of tracking window at the same time.Through verification,the above method can solve the problem that the traditional CamShift detection algorithm is strongly depend on the color of the object,and also can solve the phenomenon that the object exceeds the tracking window to realize the detect and track of moving object.In dynamic scenes,the moving object detection is interfered by dynamic background.In order to reduce the interference caused by moving background to moving object,this paper proposes a SURF feature extraction algorithm and FLANN optimal matching algorithm based on adaptive perceptual area for object detection.Firstly,through the analysis of the moving object to determine the moving direction and range,and set the perceptual area of the object.Then,the object feature is extracted and matched by using the SURF feature extraction and FLANN optimal matching algorithm to confirm theobject.Finally,the object is extracted from the video frame image by the known object.Aiming at the problem of tracking window migration when moving object is occluded,this paper combines Kalman algorithm with traditional Camshift algorithm to track the object.Through verification,the above method can reduce the interference caused by background to object detection,and resolve the phenomenon of tracking window migration when the object is occluded.Based on VS2012 software platform,this paper uses OpenCV2.4.13 library to complete the design and implementation of moving object detection and tracking.At the same time,in order to facilitate the use of code,we encapsulate some functional methods and demonstrated in a visual form(MFC).