Research And Application Of Forest Aerial Image Mosaic Algorithm Based On UAV

As an emerging low-altitude remote sensing technique in recent years,UAV remote sensing is capable of offering various advantages including low cost,high timeliness and high security level.It has been widely applied in forestry-related fields such as forest information monitoring,pest and insect preventing and wildfire monitoring,etc.However,due to the limitations of flying height,digital camera focal length and image resolution,the valid coverage of acquired aerial images is limited.Consequently it is difficult to contain whole image information and acquire an overall recognition of the target research areas.Therefore,the UAV aerial images need to be stitched to obtain forest aerial images with high resolution and broad vision.This paper aims to study the image mosaic algorithm of forest aerial images to meet the image mosaic requirements.Key research work is shown as follows:1.An improved SIFT image mosaic algorithm was proposed in this paper.Traditional SIFT algorithm in UAV stitching forest images practices has disadvantages such as low contrast,complex matching process and low accuracy.This paper presented an improved SIFT image mosaic algorithm to address these drawbacks.By introducing adaptive histogram equalization pre-processing,feature matching by vector dot product inverse cosine function and transformation matrix refining by L-M algorithm,the number of feature points was increased.Therefore the efficiency and precision rate of image mosaic were improved.2.A system for aerial image mosaic results evaluation was built in this paper.Based on the mosaic experiment of UAV forest aerial images,the results evaluation system was proposed by combining subjective and objective evaluation.Among them,the subjective evaluation was achieved by rating the stitched images with human visual experience.The objective evaluation was to evaluate the stitching effect through four evaluation indicators,namely SMD,PSNR,SSIM and stitching time consumption.The combination of both subjective and objective evaluation methods can effectively improve the fairness and reliability of the result evaluation.3.A mosaic software was designed for UAV forest aerial images in this paper.The proposed software was based on the MATLAB GUI tool.It integrated a variety of image mosaic algorithm as a whole body by integrating image display,parameter setting,debugging and software control functions.It can effectively improve the interactivity and operational ability of aerial image mosaic task for forestry purpose.In this paper,the improved SIFT algorithm was verified by using forest images acquired in wild experiments.The experimental results showed that the proposed algorithm can complete the forest aerial image mosaic work fast and effectively.The subjective evaluation is promising.The objective evaluation indicators,such as SMD,PSNR,SSIM and stitching time-consuming are also superior to both traditional SIFT algorithm and block matching algorithms.At the same time,the image mosaic software designed in this paper greatly simplified the operation process of image mosaic and provided ideal convenience for the application of UAV forest aerial image mosaic algorithm.