Research On Ada Boost Algorithm Based On Particle Swarm Optimization

The Adaboost algorithm obtains a strong classifier by combining multiple weak classifiers,which can effectively improve the classification accuracy and a certain ability to resist overfitting.Adaboost algorithm is simple,fast detection,and has a wide range of applications in face detection,signal classification and detection.However,the Adaboost algorithm still has deficiencies in many aspects,such as long training time in large-scale samples,poor resistance to noise,especially the error of the classification label will seriously interfere with the effect of the classifier.There are many ideas and types of improved Adaboost algorithms,such as gentle-Adaboost,xgboost,etc.,which have achieved good improvements in classification performance,training speed,and overfitting resistance.After analyzing the Ada Boost algorithm,this paper puts forward the idea of optimization and improvement.Combined with the particle swarm algorithm to optimize and improve the classifier training and detection method design of the Ada Boost algorithm,the Ada Boost algorithm uses Haar-like features to collect relevant data when performing classification The research improvement is mainly reflected in the following aspects:(1)Aiming at the problem that the Ada Boost algorithm takes a long time in the training process of the weak classifier,the idea of ??combining the particle swarm algorithm to train the weak classifier is proposed.The search method greatly improves the training speed of the weak classifier.(2)Aiming at the problems of low quality and high false detection rate of weak classifiers trained by traditional Ada Boost algorithm,the idea based on dual feature weak classifiers is proposed to improve the quality of weak classifications and reduce the number of weak classifiers in strong classifiers To improve the performance of the overall algorithm.The improvement of the algorithm in this paper is based on the particle swarm algorithm.The improved dual-feature weak classifier consists of two optimal features.Through the unique optimization ability of the particle swarm algorithm,the two optimal feature combinations are found.Form a weak classifier with the best performance.(3)In view of the problems that the detection method of the Ada Boost algorithm gradually enlarges the detection window and traverses the entire picture during the detection process,which takes time and low detection efficiency,the best detection window selection method based on particle swarm optimization is proposed,and the targeted selection detection The window size increases the detection speed.(4)The detection window of the traditional Ada Boost algorithm often traverses the entire picture in a fixed step size exhaustively during traversal.This stepping method is also a key issue that affects the detection speed.Therefore,this paper proposes the idea of optimal step size selection based on particle swarm optimization,and searches for the optimal stepping strategy through particle swarm optimization to improve the detection efficiency.The algorithm proposed in this paper is simulated on the MATLAB platform.The results show that training the weak classifier through this algorithm takes less time,the quality of the single weak classifier is improved,the performance of the strong classifier is improved,and the overall algorithm detection speed and detection The rate has increased,and the false detection rate of the algorithm has decreased.