Research On Vehicle Detection And Vehicle Type Recognition In Image

Vehicle detection and vehicle type recognition are important research topics in the field of intelligent traffic all the time.With the development of image processing and pattern recognition technology,vehicle detection and recognition method based on images have become research hot spots.Vehicle detection in image is a process that detecting vehicles from the road image,vehicle type recognition in image is a process that recognizing vehicle type from vehicle images.The vehicle detection and vehicle type recognition algorithms in image are studied respectively in this paper,which improve two aspects of speed and accuracy.The specific research work is listed as follows:(1)Aiming at the problem that running slowly when using deformable part model to detect vehicles due to its high complexity,an improved deformable part model is proposed and used to detect vehicles.On the one hand,weighted PCA is used to reduce dimension of HOG feature,the basis of deformable part model,so that the model's parameters can be decreased;On the other hand,after the combination of HOG feature-levels,FFT is applied to convert convolution between filters and HOG feature-levels into multiplication in frequency domain,which reduces the computation complexity.The experimental results show that the improved deformable part model gains similar detection rate and false detection rate compared with original model,while its speed increases significantly,whose average consuming time accounts for 29.6% and 25.9% of original model respectively on UIUC dataset and BIT dataset.(2)To improve the speed and accuracy of vehicle detection,a vehicle detection algorithm based on two types of feature,which are multi-scale HOG feature and multi-scale MB-LBP feature,and nested cascade Gentle Adaboost classifier is proposed.The algorithm uses integral histogram and integral image to speed up the feature extraction process of multi-scale HOG and multi-scale MB-LBP feature,respectively.Then,it builds two types of weak classifiers based on two kinds of feature for Gentle Adaboost classifiers,and nested cascade Gentle Adaboost classifier is used to further increase the detection rate and the detection speed.The experimental results show that the vehicle detection algorithm proposed in this paper has faster detection speed and higher detection rate,compared with several existing vehicle detection algorithms.(3)In order to solve the problem that extracting features slowly and lacking of spatial information in original bag-of-words model,a vehicle type recognition algorithm based on improved Dense-SURF feature and FC-VQ coding is proposed.Firstly,Dense-SURF algorithm is used to extract features when strategy of dense sample has been optimized to speed up the process;Secondly,a new coding algorithm named feature context-vector quantization(FC-VQ)is proposed to encode features,from which the spatial location information of feature can be expressed clearly,and recognition rate increases synchronously;Finally,fast histogram intersection kernel is used as kernel function,and the encoded features are trained and recognized by SVM classifier.The experimental results show that the algorithm proposed in this paper has higher recognition rate and faster recognition speedcompared with other vehicle type recognition algorithms based on bag-of-words model.