Trees Image Stitching Technology Based On The Harris Feature

At present, China is in a critical period from traditional agriculture to modern agriculture, and a variety of broad-based research and promote intelligent machine vision agricultural machinery and equipment, such as quality inspection of agricultural products, pesticides precisely on the target application techniques, pest identification control, fruit picking, the core technology of these systems is digital image processing. Through access to agricultural image analysis, processing, agricultural production, mechanization and intelligent management phase. Important prerequisite for these technologies is the ability to get a clear and complete picture, but by the mechanical limitations of camera equipment, a single camera often can not meet the demand to get the full picture. To solve this problem, the text proposed by image stitching technology to solve the actual production process problems in image acquisition stage.This paper analyzes the use ofprospect and research value image stitching technology in agricultural production, management, and by way of example analysis describes the technology for pesticide trees precisely the target application system specific application method, effective solution to a single camera can not Get the complete tree image problem. Help to improve the usefulness of the system, to promote the mechanization of pesticide use, intelligent development and improving the process of using the issue of the presence of pesticide waste and environmental pollution. The contents and results of the text is as follows:First, this paper describes in detail the theory and methods of image stitching technology, describes the basic flow of image mosaic includes image preprocessing, image registration, image fusion. Focuses on the image stitching technology based on feature points by analyzing common feature point detection operator (eg Moravec operator, SUSANC operator, Harris operator and SIFT operator) the advantages and disadvantages of the implementation process, the proposed Harris operator trees suitable for stitching images.Secondly, focusing on the use of Harris feature point detection technology tree image stitching. In order to improve the real-time algorithm is proposed splicing method for binocular vision ranging combining with Harris operator:First use of DME or pre way to measure the distance of the camera and the target trees, and then use the principle of derivation binocular vision the overlapping area between the image of the last picture in the overlap Harris extracted feature point, not only effectively reduces the amount of computation Harris operator, and also in addition to the feature points of the overlapping area does not reduce the matching process.Finally, NCC match search process is complex, large amount of computation, the problem does not apply to real-time processing, an improved NCC matching process. Firstly, the gradient direction histogram feature point direction parameter extraction, filtration is not the same gradient direction feature points, and then create matching matrix to find the right match right, to avoid duplication of the search process. Experimental results show that the algorithm not only has a good match results, and effectively improve the computing speed, real-time processing is conducive for small amount of rotation image registration.The proposed method has a certain robustness to noise, suitable for real-time processing of large image does not exist between the rotation angle of the stitching better, does not apply to changes in scale and angle larger image. Text study on agricultural production management process to solve the problem of the narrow perspective of the machine provides a theoretical basis and reference value.