Study On The Pedestrian Detection Technology Based On The Information Fusion Of Infrared CCD And Laser Measurement System

At present, the frequent occurrence of accidents at night has brought the very big threat to pedestrian's personal safety and property. In order to improve this kind of situation, various countries are constantly putting in a lot of manpower, material resources, to research and develop intelligent transport systems. At night, the vehicle with a pedestrian detection system has special and the vital practical significance to reduce accidents and protect the safety of pedestrians and improve the traffic situation.As the pedestrian detection system which is based on vision conditons is restricted in the night, it is essential for intelligent transport systems to detect pedestrian at night with the sensor which is indispensable to visible light. In this paper, we use two kinds of sensors which are the near-infrared CCD camera and the Laser Measurement System to detect pedestrian through the brightness and depth of information they acquired. The following three aspect researches are the main content of the paper:1. The Pedestrian Detection System based on infrared image This paper adoptes the initiative infrared CCD camera of JG-968A.It brings an infrared lamp. In this paper, our algorithms can identify pedestrians who are within 15m from the camera. This paper focuses on the near-infrared image processing algorithm, including image acquisition. Owing to the low pixel contrast and low signal-to-noise ratio of the infrared image, the paper has adopted the unlimited mean filter and gray stretch algorithm to filter out of the background noise of the infrared image and enhance pixel contrast of the infrared images. According to the different distance the pedestrian away from the infrared camera, we use two kinds of image segmentation methods, Otsu andinvariant moment algorithm, to extract the interest region of the infrared image. Then we extract the area, location, aspect ratio, dispersion degree of the connect region of the segmented image, and statistical methods are used for the scope of its threshold. And designs the multi-sorter cascade, progressive filters negative sample according to the characteristic constraints, examines the pedestrian finally.2. The Pedestrian Detection System based on As the Laser Measurement System has the advantages of fast scan speed, low cost, and being able to work under all kind of weather, the paper uses of to detect Pedestrian through acquiring the distance information of the road in front of the vehicle and then analyses and processes it. First data from the laser is processed with median filter to remove noise point, and then the information of obstruction line-segment area in the interested region of front vehicle is detected by edge extraction. In this paper, the pedestrian criterion based on laser image is obtained by analyzing statistically the information of pedestrian samples'width.3. The Pedestrian Detection System based on data fusion Because the Low infrared image quality, pixel threshold segmentation methods the paper used will easily lead inseparable integrity, as well as the phenomenon of divided redundant of the segmented image.And this will result of infrared images pedestrian identification inaccurate. Even with the improved segmentation algorithm consumed a great deal of time, we can not enhance pedestrian greater recognition accuracy. As automotive systems'real-time request, this information will be infrared and laser fusion, thereby improving the pedestrian detection effect based on individual sensor information.The multi-sensor information fusion technology opened a new efficient path for the infrared target recognition; it means using different sensors work on the complementary nature of the complex scenes can improve in the target detection, identification and tracking capabilities. This paper proposed that based on infrared image's aspect ratio, the dispersion degree characteristic as well as the laser obstacle line segment region information carries on the multi-sensor decision-making level information fusion method, used the D-S evidence theory the data fusion method to carry on the pedestrian to examine, enhances the pedestrian to examine system's robustness and the reliability.Although this is mainly based on the static image of the pedestrian detection method did not use tracking algorithm, However because receives the existing condition and the research time limit, the present paper research work has the insufficiency in the aspect, waits for further consummating and improving.