Application Research Of UAV Aerial Image Recognition Technology In Agriculture

China is a big agricultural country.Accelerating the development of agricultural modernization and improving the agricultural competitiveness of big countries are the foundation of China’s economic development and comprehensive national strength.With the rapid development of technology,the application of drones has become more and more widespread,and high-resolution,large-scale images can be obtained by shooting farmland using aerial drones.In the field of machine vision,technologies such as image processing and pattern recognition have also been widely used.The combination of drone technology and image recognition technology is the starting point.Committed to solving practical agricultural problems is the focus of this article.This paper studies the aerial image recognition algorithm of drone.The main research contents are aerial image preprocessing and segmentation,aerial image mosaic,image feature extraction and matching.The factors affecting aerial image quality were analyzed and preprocessed by image distortion correction and smooth noise reduction.The morphological processing is introduced.Two methods of segmentation,image threshold segmentation and region segmentation are described.Combined with the characteristics of the UAV image,a segmentation method is proposed to divide the image into small pieces according to the checkerboard pattern.The main techniques of image stitching are summarized.Harris corner detection is used to extract features.MAD feature point matching is used to implement feature coarse matching.The RANSAC algorithm is used to accurately match features.Finally,the image is fused by a weighted average fusion method to achieve an aerial sequence.Quickly and accurately stitch images together.The visual features of the colors and textures are used as observation points based on the characteristics of the aerial image.The color feature and texture feature extraction methods are introduced,and feature matching is i mplemented by nearest neighbor classification.Aiming at the problem that the single feature recognition rate is not high,a HSI-GLCM weighted feature model is proposed,and the recognition accuracy under different weights is analyzed through experiments.On this basis,the UAV aerial crop image recognition experiment was completed,and the crop classification and wheat ear identification were taken as examples to carry out experimental verification,and the recognition accuracy experiment was analyzed.In the crop classification,according to the image acquisition-processing-interpretation complete technical process,the regional division of different crops is realized;in the wheat ear identification,there are different shapes or adhesions of wheat ears,and the cornicles are refined and the corners are detected.Methods,the relationship between the corner point and the number of wheat ears was established to realize the automatic counting of wheat ears,which provided a useful reference for wheat yield estimation.The recognition experiment was carried out in the MATLAB environment,and good results were obtained.