A New Features and PSO-SVM Classifier for Object Detection

This paper proposes an object detection method that can be real-time, efficient and accurate. At the first, introducing the new characteristic of locally assembled binary Haar-like, and combined with the idea of local binary patter. Not only can maintain the good real-time characteristics of Haar features, but also overcome the defects that are susceptible to illumination. At the same time, the support vector machine model based on particle swarm optimization algorithm is established. The PSO is used to search the optimal parameters of the SVM model to reduce the generalization impact on the classification and improve the function of the system. Finally, the video of the three data sets is used for experiments. The results illustrate that the detection system proposed in this paper successfully completes the detection task and shows better feasibility and real-time performance than other methods.