Study On Detection And Identifier Of Vehicles At Night Based On Laser And Vision

In the 21st century, high-tech has advanced very rapidly , it has triggered a series of great transformations in the military field. To reduce people's casualties arising, Unmanned ground combat platform as the technology advances and the growing demand. Military unmanned Ground Vehicle was a kind of wheel, crawler-type vehicle that could make decisions and adjust itself moving states automatically by environment perception systems of the vehicle on outdoor environment. Unmanned Ground Vehicle's development could be divided into two stages: remote sensing vehicles, intelligent vehicles.At present, we can see the military robots' development instance has the tendency to subvert traditional warfare tactics at one stroke. However, the discipline of robot technology is not very mature, many theories and important technologies have been not yet broken through. Therefore, the robot which is used in the military field also has a quite long way to take service on a large scale.How to detect moving or quiescent targets to complete the tasks at night has important significance. Environment perception is the key technology of Unmanned Ground Attack Systems. In daylight, the ordinary sensors can resolved all the tasks, However, at night the ordinary sensors can not resolve the problem of the environment perception.In this paper, because of the difficult of the common sensor detecting vehicles at night, used the infrared CCD and the line scanning laser to detect the front vehicles, and then make the data fuse.The main contents of this paper are introduced below in detail:1,Because the infrared images have many shortcomings, such as: the low contrast , the big noises , the fuzzy boundary and so on, the paper used gray extend and the histogram correction method to intensify the image , and then according to the character of the infrared image, used the two-dimensional variance threshold value division method and area threshold arithmetic to eliminate the noises.2,Based on the grey information of the images and the shape feature of vehicles to recognize the frontage and the side of vehicles. First, based on the character of the circular wheel, symmetrical, the rate of the beam-to-depth and the rate of the area of the body and the wheel, we can recognize the side. Second, at night infrared image was not clear, the traditional segment method cannot meet the requirement, in this paper, used the idea of isolated point elimination to segment vehicles' headlight and license. In the license, the rectangle is used for recognition; in of the headlight, the circular and the symmetrical is used for recognition, the comprehensive rectangle, the circular is used to recognize the vehicles'frontage.3,Used the LMS line scanning laser to detect the front of obstacles, obtained the distance image under the polar coordinates, the distance images are transformed to the common distance images by Descartes coordinates, the D-S theory is used into the data fusion. First, the three methods of the line scanning laser images, the vehicles'full-face images and the side images are defined to the three feature of fusion. The D-S theory was the method that made random two characteristic of three characteristic to fuse, then fused the result with the third characteristic.4,Experiments are implemented on the platform based on VC++6.0 and the results have proved excellent performance of the algorithms.