A Effect Method For Pedestrians Detection Based On HOG And Haar

Vision algorithms are used widely now.In many areas it has achieved very good results.But it faces many challenging issues.The problem is particularly serious when it comes to analyze human activities in video surveillance applications.In the field of pedestrian detection,the most widely used detection methods are algorithms based on Histograms of Oriented Gradients(HOG)and algorithms based on Haar-like features.However,these two methods have their own shortcomings.The algorithms based on HOG get more accurate results,but it calculate too much.Haar-like features is simple,better for face recognition,but not applicable for the detection of the back.The result is not ideal to use these two features separately.The main contribution of this paper is an effect method for pedestrian detection based on histogram of oriented gradient(HOG)and haar-like.In order to combine the two methods to achieve the purpose of taking their advantage,this paper uses the Adaboost algorithm.It is composed of weak classifiers.We use HOG and Haar-like features make weak classifiers to train samples.The Adaboost algorithm is used to weight the weak classifier to achieve the best detection effect.The experimental results show that this new method has a great improvement in accuracy and detection speed.In the pedestrian tracking part,this paper performs the target scale adaptive processing on the target tracking algorithm based on Mean shift.Mean shift algorithm belongs to kernel density estimation methods and it doesn't need any prior knowledge and relies entirely on the calculation of the sample points in the feature space.The principle of the Mean Shift algorithm is a statistical iterative process.Firstly calculate the offset of the current pointl.Secondly move the current point to the position of the calculated offset mean.Then continue to move as this process until the stopping conditions are met.At last,it can get the final position of the target.