Research On Objects Detection Algorithm In Moving Background

In order to detect the moving objects in moving background, this thesis firstly uses the optical flow method to finish the initial detection on objects. This method can only get the approximate area of objects, and also can not detect the multiple objects correctly which move by the same directions and velocities. So, geodesic active contour model is next to be used, which can achieve the curve evolution. The model is classified as based on level set method and based on variational level set method. It makes the objects detection results of the optical flow method as the initial curve, and realizes the accurate detection on objects when the initial curve surrounds the objects nearly. However, the error of the optical flow method exists. It is likely to appear the phenomenon that part of the curve is across the objects and away from the edge, and lead to the detection failed. In view of this situation, an improved geodesic active contour model is proposed in this thesis.Improved model mainly deals with the videos that most background is different from objects on gray scale. It can adaptively control the various points on the curve to shrink or expand independently. Therefore, the part of curve in the background shrinks inwardly and the other curve in the objects expands outwardly until they all reach the edge of the objects. The experimental results show that the improved model is indeed able to achieve the objects detection correctly when the initial curve is across the objects and away from the edge. Moreover, the stabilization of the improved model is more than that of geodesic active contour model. It means that if the curve evolution speed increases gradually, geodesic active contour model may fail, but the improved model succeeds under the same conditions.