Research On Road Scene Object Detection Based On Depth Convolution Neural Network

With the rapid development of economy,the automobile has become the main means of transportation.However,with the rapid growth of car ownership,traffic safety is becoming more and more serious.One of the main factors leading to traffic accidents is the driver's distraction and negligence.The vehicle mounted auxiliary safety system can find the potential risks of the vehicle and the effective system and tools for the early warning to the driver.Automatic driving technology is a complex system with many technologies,such as location,perception,decision-making,control and so on.The perception system provides the basic supporting data for the automatic driving system and the auxiliary driving system.In this paper,we design an object detector model based on depth convolution neural network,which is used to detect the target which is of great value to the driving system.At present,the object detection framework based on depth convolution neural network is divided into two research directions.One is to build the frame of object detection by using the method of region recommendation and the feature extraction method of deep convolution neural network,and the other is to direct the region frame and object recognition from the depth convolution neural network.In this paper,we will do further research on the two methods,and discuss how to train and test the performance of the road scene detector model in different ways.The design of the road scene object detector using depth convolution neural network layer convolution feature extraction of image data on the surface,and then use the specific location frame extraction algorithm of positioning road scene objects in the image,and then use specific categories of feature extraction of object recognition layer convolution.Finally,in the high performance computer trained detector model into the embedded system,and test the road scene object detection model for day and night in different light environment of road object detection and get better results.The accuracy of pedestrian detection in the road scene is 83%,and the average detection accuracy of small cars,buses,motorcycles,bicycles and other vehicles in the road scene is about 81%.