| At present,the car has become an indispensable vehicle,and the intelligent assistant driving system is an important part of intelligent transportation.Car and pedestrian detection is a basic part of vehicle perception of the external environment and an important branch of computer vision and image processing.Along with the rapid development of detecting technology,people also put forward higher request to the accuracy of detection,in order to be able to better the implementation of the road scenes of the vehicle and pedestrian detection,people began to consider add depth information in ordinary images,and the way to carry out the road scene depth information estimation also became an important research topic in the field of computer vision.The main work of the paper is summarized as follows:First of all,aiming at the problem that the traditional car and pedestrian detection algorithms have too many parameters when extracting features,this paper proposes a car and pedestrian detection algorithm based on deep learning,especially the method based on the open source Faster RCNN network framework.Then,considering the importance of the distance to car and pedestrian detection in road scenes,this paper mainly considers the depth estimation of monocular road images.Aiming at the problem that the prediction accuracy is not accurate enough when the depth information is recovered from the monocular vision image,a depth estimation method based on multi-scale convolutional neural network is proposed in this paper.Finally,considering the significance of local information and global information,and considering the insufficiency of current model,anetwork model of monocular road image depth estimation based on pyramid pooling network is proposed.In conclusion,this paper mainly implements car and pedestrian detection based on deep learning and depth estimation of road scenes,and it is proved that good results can be obtained through multiple tests. |
PDF Full Text Request