Abstract
The discoloration of rice kernels by the salivary secretions of stinkbugs causes "pecky rice", which is rated low on taste and quality. Despite the urgent need to develop measures for preventing such damage, many stinkbug species are difficult for non-experts to identify in the field. So when accurate identification is necessary, it is common practice for farmers to submit samples or photos to entomologists. However, this process is time-consuming and places demands on the entomologists, particularly when there are many specimens relative to the number of experts available. We evaluated the potential of automated digital image analysis to identify stinkbugs. We captured digital images of samples and assessed how to discriminate among the species on the basis of five fundamental morphological characteristics. We then used genetic programming methods to generate a novel algorithm to identify species by these five characteristics. Our results demonstrate the potential for using digital image analysis for automated identification of the stinkbugs responsible for pecky rice.
