Application of Improved 2-D Entropy Algorithm in Rubber Tree Image Segmentation

Abstract

The tapping panel dryness (TPD) in Hevea brasiliensis is a complex physiological syndrome and seriously affects the yield of the natural latex from rubber tree. Using image segmentation technology, the proportion of cut and latex area in rubber tree image are calculated to determine grade of TPD. According to the principle of two-dimensional maximum entropy, the global search of genetic algorithm and simulated annealing local hill climbing performance are synthetically selected to determine the optimal segmentation threshold. 276 images selected from the experimental base were tested. The experimental results showed that the method quickly and efficiently realized the extraction of the rubber tree cutting marks and the separation of rubber latex. The accuracy rate of TPD grade was as high as 92.4%, which could effectively prevent the further spread of TPD, and had certain experimental value.