Research On Vehicle Target Detection Based On Embedded System

Vehicle detection is an important part of environmental awareness in Advanced Driving Assistance(ADAS)system.Traditional vehicle detection is usually implemented on PC.With the rapid improvement of computer software and hardware,deep learning and target detection algorithms continue to develop,and vehicle detection is gradually applied to embedded devices.Embedded devices have the advantages of small size,low price and good stability.If the embedded devices can achieve accurate and real-time vehicle detection,the development cost of ADAS system will be significantly reduced.The hardware performance of embedded devices is limited,and the algorithms used have considerable limitations.In recent years,target detection algorithms based on in-depth learning have developed rapidly and achieved high-speed detection with high accuracy.However,these algorithms can not be directly applied to embedded devices.Aiming at the characteristics of vehicle detection and the special place of embedded equipment,this paper improves and creates SSD network based on SSD target detection network and other technologies.Secondly,it builds the software and hardware development environment of embedded system and realizes vehicle detection on embedded equipment.Firstly,according to the distribution of vehicle data and the characteristics of embedded devices,the original SSD network is improved as follows: the K-Means clustering algorithm is used to set up a new regional candidate box of SSD to make it more in line with the scale distribution of vehicle data and improve the detection accuracy of the model;On the basis of the original SSD loss function,exclusion loss is added to improve the detection performance of overlapping vehicles.MobileNetV1 depth neural network is used as the feature extraction network of SSD,which can greatly reduce the amount of calculation to meet the real-time requirements of vehicle detection algorithm running on embedded devices without reducing the detection accuracy.Embedded system takes Samsung Exynos4412 microprocessor based on ARM as hardware platform,transplants Linux operating system and all drivers involved on this platform,and completes the transplantation of QT graphical interface library and Caffe deep learning framework library in application development.