Disparity estimation on stereo vision based fruit inspection

Abstract

Star fruit is one of the most popular fruits exported by Malaysia. From 1965, Federal Agricultural Marketing Authority is given the authority to regulate the quality of the star fruit exported by Malaysia. The great interest among the importers of Malaysia’s star fruit becomes a great motivation to increase the yield and quality of the star fruit for export. One of the areas where the process can be improved is the fruit inspection. Universiti Teknologi Malaysia Computer Vision, Video and Image Processing Research Group (CvviP) had successfully invented an automatic star fruit grading system. Now, CvviP try to extend the system by experimenting the application of stereo vision in starfruit inspection. Disparity map obtained from the stereo vision can be used to find the size of the star fruit. Thus, this thesis objective is to experiment with stereo vision in order to obtain the disparity map of the star fruit. The experiment result should dictate the feasibility of the stereo matching to be integrated into the current automatic star fruit grading system. The proposed approach consists of hardware and software implementations. The hardware implementations focus on the hardware modeling and setup to acquires the starfruit images. While the software implementation focus on obtaining the disparity map from the images acquired. Based on the result obtained, the proposed approach has potential for practical implementation with some improvements needs to be done.