Study On The Vehicle Front Pedestrian Detection Based On 3D Laser Scanner

The China's economy continues to develop, and people's living standard continues to improve over the next decade, the whole number of car in China will continue to maintain high growth. With the number increasing, road traffic problems caused a more prominent and increasingly serious security situation. The number of road traffic accidents, deaths and mortality in China are all highest in the world. This is because of road traffic caused by inadequate infrastructure, and pedestrian and different types of vehicles hybrid in same road, which can easily lead to the collision of people and vehicles; and partly because laws and regulations on pedestrian protection and the related lack of road safety technology. In order to effectively protect the safety of pedestrians and carry out the research of pedestrian detection in front of the vehicle, it is important significance to develop a kind of collision warning system between vehicle and pedestrians reducing the number of pedestrian casualties in road traffic.To realize early warning of the impact that may occur between pedestrians and vehicles in time, pedestrian detection system not only requires the timely detection pedestrians in a complex environment, but also can identify the precise positioning of the pedestrian. So timely warning can be sent to drivers of vehicles and the automatic steering and brake system, which can assist drivers deal with emergency. Machine vision, infrared imaging technology and millimeter-wave radar are mostly adopted as the detection devices in traditional pedestrian detection system. However, these devices exit some obvious drawbacks, such as machine vision is vulnerable to light, weather, lack the location and orientation information infrared imaging system is susceptible to heat effects, positioning needs assistive devices, and millimeter-wave radar is not sensitive to small obstacles and not conducive to the goal classification. With the development of three-dimensional laser scanning technology rapidly, scanning speed is increasing greatly. Three-dimensional laser scanner can sense the space information, and obtain the distance, angle, and reflectivity of the points of space. This paper research the vehicle front pedestrian detection system based on 3D laser scanner. The study involve the three-dimensional laser scanning, communication control, data acquisition and storage, data processing, feature extraction, classification and other aspects of technology. In this paper, the study included the following aspects:1) Three-dimensional laser scanner serial and parallel communications programming:This section includes the main principle of the serial and parallel communication, features and corresponding ports open, setting and other functions, in addition, the control and acquiring data of the system is achieved by VC++ programming.2) Data acquisition, storage and scanning accuracy analysis:Firstly, the data acquisition process of the three-dimensional laser scanner and the principle of measuring distance are presented. Secondly, in order to facilitate subsequent processing and use of data group, the different types of collecting data are real-time stored in the system. Finally, in order to guide and control the data acquisition accuracy, the reasons leading to the data error are analyzed from the perspective of the system design and manufacture factors and the use factors of the system.3) Imaging laser range data:Derived from the data to the gray image of laser space mapping function, and according to the mapping function, designed the algorithm process, and ultimately by the VC++ programming, the image generation is realized. According to the characteristics of laser range image, deal it with noise filtering, histogram equalization, edge sharpening, threshold segmentation, etc.4) Feature extraction and classification:Including laser range image Characteristics of the mathematical description according to the characteristics of the target of treated laser range image, extract the target in the image height, width, perimeter, and area, axis of symmetry and other geometric features and shape feature. Theses features were used to designed the linear classifier, to achieve the target recognition of range image.