Locust images detection based on fuzzy pattern recognition

Locust control is the focus of agricultural pest management. As a complement of manually monitoring, low-attitude airborne early warning system can be used to monitor locusts by identifying and counting them from the captured locust image. The experimental field was at Qingyuan, Guangdong. Locust images were captured by digital camera. By contrastive analysis of the average of R, G and B value of locust area and background, a method of extra-green absolute value was adopted to segment locusts from the background. The area and perimeter of each locust were obtained by comparing area statistical value. Fuzzy sets of individual locust object and connected locust regions were established respectively. Individual locust or connected locust regions were determined by the maximum membership degree principle. The accuracy of the fuzzy recognition of individual and connected locust region was 89%, which could satisfy the requirement of locust pre-warning.