Research Onmulti-scale Pixel-level Information Fusion For Lidar Four-dimensional Image

Streak tube laser lidar is a new laser imaging technology that has a flash type focalpiane. It can obtain a four-dimensional image which contains the information about thetarget material information and the spacial information, and has a good applicationprospects in the field of target detection and recognition. But in the process of imaging,the quality of four-dimensional image decreased because of the atmosphere, the targetcharacteristics and other factors. Therefore, this paper makes use of the relativeconsistency of strenth image and distance image,adapts the core ideology of multi-scaledecomposition algorithm, by studying the pyramid decomposition algorithm andwavelet decomposition algorithm, and fuse images of visible light, strength anddistance on pixel level by streak tube laser lidar, and then obtains an image with thefusion of intensity and distance, and an image with the fusion of strength, distanceand visible light image. This image process improves the image contrast and retain theimage details, and provides a more reliable image information for the subsequent targetdetection and recognition.This paper includes the following four aspects:(1) to study the pyramid andwavelet multi-scale decomposition algorithm, improve the fusion rules ofmulti-resolution image restoration process, including.①in the recovery process, the toplayer focuses on strength image.②the laser lidar equation as the core of modulationfusion rules to improve the image contrast.(2) to propose a method of three pixel-levelinformation fusion on visible image, the strenth image and distance image by streaktube liars.(3) to present a contrast modulation distance image denoising algorithm.(4)to study the objective evaluation index. After the comparisons on subjective andobjective image fusions, the edge of the trust is proved to be a more suitable asfour-dimensional laser lidar information fusion evaluation.Using the proposed algorithm, a series experiments of different types of streaktube laser image pixel-level information fusion on land, underwater and land buildingsand other targets. The results show that the proposed algorithm is suitable for differentscenes with a better universal applicability. It also improves image quality, increasescontrast, enhances profile targets, and can provide more reliable image data for thefollow-up target detection and recognition.