Kesearch On Image Mosaic Algorithm Based On UAV Visible And Infrared Images

Invasive species makes China become one of the most serious countries affected by plant diseases and insect pests in the world, especially pine wood nematodes, which threaten the development of the forestry ecological construction severely. UAV, as a new low-altitude remote sensing technology, has been widely applied to the fields of forest fires monitoring, natural disasters area detection, with the advantages of high flexibility, low-cost and simple structure. It plays an irreplaceable role in remote sensing monitoring system. Combined with the visible and infrared images information taken by UAV, this study primarily researched the image stitching problem in the research on the application of UAV in pine wood nematode disease investigation and monitoring. With regard to a large number of high-rate remote sensing images stitching technology, the main research contents as follows:First, this study introduced the general process of image stitching technology. Then image matching problem, the key technology of visible and infrared images stitching, was fully discussed. And the function of several common SVD matching algorithms was also analyzed.Second, a new SVD image matching algorithm based on CS-LBP was proposed. Firstly, the Hessian-Affine detector was applied to obtain the feature region from the image and the CS-LBP descriptor was used to describe the local feature. Then adjacent matrix was constructed by using the similarity of the local feature description as a measurement. The matching relationship between the images was acquired according the SVD matching algorithm. Substantial experiments indicated that the proposed algorithm outperformed the previous SVD matching methods, when the images have large transformation of illumination, scale, rotation.Third, area for the unmanned airplane flight field test and the acquisition method of visible and infrared images based on UAV platform were introduced. In experiment, the UAV flight field test implemented in Feidong forest farm was introduced. Then the methods of how to acquire the visible and infrared images in the same scene by airborne visible and infrared camera respectively were stated. Finally, the imaging characteristics of visible and infrared image were analyzed.Fourth, an accurate matching algorithm combined visible with infrared images information was proposed. Firstly, the initial image matching was made by using the SVD matching algorithm based on CS-LBP. Secondly, mismatching was eliminated by combined visible with infrared images information, and an accurate matching result was acquired. Then the transform matrix between images was estimated by RANSAC. Lastly, the panoramic images of the forest were obtained by image fusion technology.