Research Of Forest Aerial Image Mosaic Algorithm Based On UAV

Affected by factors like flight height and camera pixels,when the drone is performing aerial photography in a low-altitude forest area,the range of images taken is usually small,and it cannot cover all the image information in the forest area to be studied.In order to obtain high-resolution wide-angle aerial images of forest areas,the UAV must complete the image capture along the designated route during aerial photography,the aerial photography range covers the entire forest area to be studied,and then use the panoramic image stitching method to transform a single aerial image into The entire wide-field image contains richer forest information.The main research topics of this article are as follows:1.A CSURF(Circled Speeded Up Robust Features)feature extraction algorithm based on circular neighborhood is proposed.For the traditional SURF(Speeded Up Robust Features)algorithm,the matching process is complex,and the feature points are relatively fuzzy.In the process of determining the main direction of the descriptor,the CSURF operator proposed in this paper selects the circular area instead of the square area in the traditional SURF algorithm.The main direction deviation of the feature descriptor is effectively reduced;at the same time,the secondary gradient of the image is introduced as an extension of the details of the image feature operator,which enhances the robustness of the algorithm.Compared with the traditional SURF algorithm,the algorithm has a significant improvement in image evaluation indicators such as grayscale variance,peak signal-to-noise ratio,and structural similarity.2.Establish a complete image stitching effect evaluation system.In order to make the evaluation results more convincing,this time a complete evaluation system is built around the objective aspects of the stitching results.This paper introduces SMD PSNR SSIM and stitching time-consuming Four indicators including the double image stitching experiment,the multi-picture stitching experiment and the dynamic image stitching experiment complete the comprehensive evaluation of the aerial image stitching results in the forest area.This evaluation system can effectively ensure the fairness and reliability of the evaluation results,and then verify the superiority of the algorithm proposed in this paper.3.Developed a forest aerial image stitching software based on MATLAB GUI.In view of the problem that the aerial image stitching operation in forest area is not intuitive,this paper developed a host computer based on CSURF feature point UAV forest aerial image stitching algorithm software.The software integrates three image stitching algorithms,supports image display,parameter adjustment,system control and other functions,which can greatly enhance the image stitching interactivity and operability.The specific modules of the software include: one,display part;two,parameter setting part;three,control part;four,splicing algorithm selection part;through experimental analysis,it is known that the software has good practicability and shows a large-scale promotion and application value.This article deeply studies the image stitching algorithm,combines drone aerial photography and image stitching technology,and obtains high-resolution,wide-field forest aerial images,which creates favorable conditions for the next image research.At the same time,this research has developed an aerial video stitching software in the forest area of the host computer,which has realized the practical application of the CSURF algorithm.