Research Of Pedestrian Tection Method Based On Active Contour Models

The motion analysis of moving objects in video sequences is an important part indomain of computer vision technology. The motion analysis consists of initial positionof moving object, extraction of object contour, object detection and tracking, objectrecognition, and understanding. Initial position of moving object and extraction ofobject contour are the precondition of object tracking, object recognition, and objectunderstanding, and are very important to the latter. Pedestrian detection in videosequences is an important direction in the moving objects analysis. It involves wideapplication foreground and great economic value in intelligent monitoring andanalysis of human behavior, etc.Active contour model is paid close attention since it was proposed in 1987. And itis widely used in domains of computer vision. Active contour model has a naturaladvantage to study the deformed objects, because its contour is"active". Pedestrian inthe video sequence is the typical deformed target. And using the"active"uniqueadvantage of active contour models can effectively achieve the pedestrian targetdetection.In this paper, it proposes an improved active contour model to detect pedestrian invideo sequences. First, it obtains the initial contour of the pedestrian by backgroundsubtraction instead of manual calibration, overcoming the existing range of issues inmanual calibration, such as instability, etc. Second, improve the internal energycomputation. It improves the internal energy computation by using the squares of thedistance between average distance of all adjacent control points in the contour curveand the control points, overcoming the disadvantage of the existing methods. Andconstruct local energy window to search for optimal solution. It could increase thecontinuity of the object contour and make control points distribute evenly to avoid aggregation. Third, Greedy method is used. Greedy method is proposed to obtainminimum energy and find image contours differing from the traditional variationalapproach. Greedy algorithm is a kind of method that it does not need to find anoptimal solution, just wants to find more satisfied solution. Greedy algorithm canmore quickly get the general satisfactory solution, which eliminates the need to findthe optimal solution but to exhaust all possible solutions that must be spent a lot oftime. Greedy algorithm bases on the current situation to make the best choice,regardless of the overall possible situations, so the Greedy algorithm does not requirebacktracking. Compared to others, Greedy algorithm is reliable, stable, and allows theaddition of external constraints, etc., and the complexity of the algorithm is greatlyreduced.The experiment results prove that the algorithm has a better performance indetecting pedestrian compared with other object detecting methods and traditionalactive contour models.