Abstract
The intelligence for harvesting machine of tropical fruits requires the ability of image analysis on natural background, which more complicate than image analysis on unique color background that is used for grading fruits and vegetables. In some case, we can easily distinguish fruit areas in natural background image by color. However, finally, we have to use shape analysis for identification and getting exactly boundary positions. This research intends to detect position of fruits by its shape and used Elliptic Fourier Descriptors to describe the typical shape. Then, we deform the typical shape in the spatial frequency domain by scaling, rotating, and phase shifting and match with the interesting image to get the maximum likelihood. However, the matching process to determine all likelihood takes a long time because we have to inverse Fourier series and deform the parameters as many as possible to get enough accurate result. To optimize these problems, we use FFT for inverting the typical shape and the Genetic algorithm for searching the most likelihood because this problem is a problem depending on natural environment. Using Genetic algorithm, which is a random search algorithm, is suitable and can decrease position-searching time.
