Research On High-speed Vehicle Detection Based On Convolution Neural Network

Deep learning algorithm can be effectively applied to the study of large-scale data,and the convolution neural network is an application of deep learning algorithm in the field of image recognition.The convolution neural network has robustness in the image classification and detection of big data sets,and can effectively avoid the pretreatment of complex images.R-convolution based on regional neural network(CNN)--is a kind of combination with regional nomination and convolution neural network methods of target detection,for nomination and size normalization in the region and improve the performance of ima ge classification and detection.The faster-R-CNN directly calculated candidate boxes using regional choice network RPN,which is a picture of a arbitrary size as input,with the size specifications of the rectangle as the output of a batch of selecting network,each field mapping a goal scoring and location information.Based on this theory,this paper raises a convolution of the neural network based freeway scene vehicle detection algorithm.Through artificial tagging highway vehicle images,the Faster-R-CNN training learning algorithm model,a convolutional neural networks for vehicle detection appears.In this paper,the problem of vehicle detection in highway scenes is transformed into two classification problems of vehicle and background,and the classification and detection of vehicle targets are carried out in combination with the image data set of big data size.In this paper,a picture annotation system is used to train the Faster R-CNN.After a large number of experiments,it is proved that the vehicle detection algorithm based on convolutional neural network has good classification and detection accuracy,and verifies the practicability and effectiveness in image classification and Detection of the implementation and effectiveness.