Analysis Of Image Segmentation Algorithms For Forest Diseases And Insect Pests

Forest is one of the most important resources on earth, forest diseases and insect pests is almost the biggest threat to forest health. How to monitor and prevent forest plant diseases and insect pests is an important issue to forestry experts for a long time. The traditional methods of monitoring forest pests mainly rely on artificial means like patrol detection and so on, which is subjective and has time lag, and also greatly be influenced by the terrestrial environment.This thesis proposed a new way to monitor forest plant diseases and insect pests aiming at the sea-buckthorn & Chinese pine forest of Jianping district, Liaoning province. At first, we use unmanned aerial vehicle(UAV) to collect images of the stricken forest, and preprocess the images, then segment the image and extract the stricken area, then do the disaster classification work combined with the ground survey results (leaf loss rate). For the issue of UAV image of stricken sea-buckthorn field, we use the marker-based watershed algorithm using fractional calculus (FC) to segment the images, this algorithm first calculate the FC value of each pixel, which can replace the gray value or gradient value in traditional algorithms. Experimental results show that this algorithm can accurately extract the stricken area, and the segmentation accuracy can reach more than 95%; For the issue of UAV images for severely affected pine field, we use the Type-2 Fuzzy Clustering Algorithm (Type-2 FCM) algorithm to do the segmentation work, this algorithm can stretch the membership function linearly, as a result, the algorithm can extract the affected area correctly; For the issue of UAV images for lightly affected pine field, we use the texture characters to judge the disaster classification. Experiments results show that fractal dimension (FD), lacunarity (lacu) and Fractal Dimension Gradient (FDG) three texture characters can represent the disaster classification of pine-tree UAV images. Finally, this thesis do the comprehensive analysis and comparison work of the proposed UAV image segmentation algorithms based on the forest diseases and insect pests issue, and summarizes the application range of the algorithms.