Design of Automatic Recognition of Cucumber Disease Image

An automatic recognition method for cucumber disease images is presented. Threshold for image segmentation was generated with 2 dimensional maximum entropy principle and optimized with differential evolution algorithm. With threshold values generated, the authors have segmented cucumber disease images and picked up the lesion with maximum area from segmentation results as representative lesion. Then they analyzed representative lesions of disease images and extracted theirs color features and texture features. Bayes classifier model for cucumber disease recognition was built with information of color features and texture features of representative lesions. They have applied their method to recognition of cucumber powdery mildew leaf image, anthracnose leaf image, botrytis leaf image and downy mildew of leaf image which were collected under natural conditions. Experimental results show that, automatic recognition approach for cucumber disease image can recognize cucumber disease images without human interaction and high performance of their method.