Research On Terrain Matching Algorithm Based On KRAWTCHOUK Oments For Streak Tube Imaging Lidar

Precise navigation and positioning have attracted a lot of interests around the world. Terrain-matching aided navigation based on lidar image matching which has a higher positioning accuracy and a smaller length of the matching area can acquire the range information and intensity information of the terrain below the aircraft in real-time compared with the traditional terrain aided navigation. This technology becomes a research hotspot in recent years.As one of the latest flashing non-scanning laser imaging technologies, streak tube laser imaging technology is capable of achieving larger field of view and higher frame rate than scanning laser imaging technologies, and both range information and intensity information of the target aera are obtained easily by this technology. However, how to make full use of the wealth of information provided by lidar images is still a problem. In addition, when the streak tube laser imaging lidar (STIL) is used as an aided navigation system, there may be a scale and rotation distortion because of the different scanning height and inconsistent of the scanning direction between the real-time image and the reference image. What's more,the deviation of the scanning direction named as the angle of rotation is generally within a few degrees in the inertial guidance system. Therefore, studying a set of anti-scale and anti-rotation distortion methods for lidar image matching are the main subject in this paper.First of all, to solve these problems, the Krawtchouk moment was introuduced to the research for lidar image matching for the first time. Through extracting Krawtchouk moment invariants and radial Krawtchouk moments of STIL images, two anti-rotation lidar image matching algorithms was constructed. Secondly, the scale of real-time image was adjust by interpolation to suppress the scale distortion. After eliminating scale distortion rotation invariants, lidar images was extracted such as Krawtchouk moment invariants, radial Krawtchouk moments and circle invariants to form feature vectors. The Camberra distance of feature vectors was used to measure the similarity of images. The lidar images were searched with specific condition and finally were matched successfully. In particularly, a set of matching methods for lidar images were introduced in this paper. In addition, the feature vector of range image and the feature vector of intensity image were used to form a joint-feature vector, by this way the lidar image with fused information of range information and intensity information was matched. The fusion method of the range image×intensity image was also tested with real STIL datas. After programming the matching algorithms, firstly, lidar image matching algorithms was tested respectively based on Krawtchouk moment invariants, radial Krawtchouk moments and circle invariants with the simulated lidar images as experimental datas. Matching algorithms were proved to be feasible. Time using of computation for feature vectors, the matching accuracy, the matching precision and noise immunity were tested in experiments. The basic properties of these algorithms were analyzed and summarized. Finally, validity of algorithms was verified by the forward looking scanning streak tube lidar images instead of the vertical looking scanning streak tube lidar images.