Applied Research On Moment Invariant In Lader Image Recognition

Currently,research on laser radar image target recognition has came in for more attention. The feature of laser radar's initiative imaging is spanning range and intensity at the same time. Initiative image possess the feature of 4 dimension information(corner-corner-range-intensity), exuberance in image information quantity and in favor of automatism target recognition,along with it the difficulty and complexity of laser radar target recognition. Invariant's characteristic diagnostic abstraction and construction is one of the key technology to image target recognition, which will make later recognition more easier. To research all sorts of laser radar image feature extraction means is to be of far reaching importance.This article has introduced moment invariant to the research for laser radar target recognition, therefore has carried out the algorithm routine compilation for Hu moment and Zernike moment's feature extraction. It has carried the emulation experiment respectively for the stability of amphi-moment's each category at noise-free and 10dB, 20dB noise. Experimental results have shown that under noise-free circumstance Zernike moment and Hu moment behave ideal, but are both short of the ability to noise suppression. Zernike moment behave better just in low-order situation. Research and stability experiment for this part has proved that traditional immovability moment behave poor at laser radar noise suppression and has made preparations for further research.According to the stability insufficiency of both moments and the relativity of laser radar intensity image and range image, this article has refined the algorithm of Zernike moment and also framed a new algorithm for moment invariant abstraction utilizing range image and intensity image concurrently. It has carried stability emulation experiment for new algorithm at 10dB,20dB noise. Results have shown that the extractive moment feature under new moment invariant algorithm at noise behaves ideal, which is a great advance compared with traditional moment algorithm.Subsequently, this article has established a type of experimental program to estimate laser radar image accuracy requirements of moment feature extraction scheme used the relation between diagnostic changes on precision parametric and extractive moment of emulation video. By using special object scene emulation experiment, distance accuracy and requirement for resolution ratio of six exponent Zernike loop post invariant feature extraction's scheme has been figured out.Finally , it has preliminary studied the establishment means of tri-dimensional clarification of objective space. Henceforth makes preparations for further research.