A Research Of Real-time Object Detection And Tracking Based On Shape-based Template Matching

With the rapid development of computer vision and image processing technology,template matching has become an irreplaceable technology in image information processing.Template matching is the most commonly used method for target detection and moving object tracking because the method is simple,adaptable and can handle complex scenes.At present,template matching has penetrated into all aspects of life,such as industry,medicine,transportation and so on.At the same time,the rapid development of the national economy has accelerated the growth of the number of vehicles,followed by urban traffic congestion and safety issues.Traffic sign recognition and vehicle tracking are two important parts of traffic safety.In this thesis,a template matching method based on shape is studied,which is applied to traffic sign recognition and vehicle tracking.The main contents and conclusions are as follows:Firstly,the basic theory of shape based template matching is analyzed,such as edge detection algorithm,template creating and similarity measurement.Secondly,the factors that affect the matching efficiency and accuracy are analyzed,including the rotation,scaling,polarity and search strategy.Secondly,a traffic sign recognition algorithm based on color and shape based matching is proposed.Firstly,the traffic signs are detected and located by the color and geometry information.Then the traffic sign is matched by extracting the edge gradient feature.The method creates templates in different scales and degrees from a small number of samples,which solves the problem that the machine learning methods need a large number of samples and the training stage is too time-consuming.At the same time,this method can solve the problem that the template matching based on gray level is too sensitive to illumination changes.The experimental results show that compared with the traditional SIFT and SURF features matching methods,the proposed algorithm has high recognition accuracy and matching efficiency.The Recognition accuracy rate is up to 86.65%,and it can also meet the real-time requirements.Finally,a method of moving vehicle tracking based on shape based template matching and particle filter is proposed.The algorithm combines the shape-based template matching and particle filter,which can solve the problem that the color based particle filter can't track the object with change of illumination.In addition,this paper proposes a dynamic model updating mechanism based on scale and rotation,which can get a good performance in occlusion,rotation and scaling.Experiments show that particle filter based on the shape based template matching can get a better tracking effect than color based particle filter method under illumination change and occlusions.