Study on three-dimensional surface reconstruction of black melon seeds using photometric stereo

In this paper, the 3-D surface reconstruction method of black melon seeds based on photometric stereo was studied, which was then used to identify the wrinkled seeds. After the image capturing system being set up, the system accuracy was calibrated by reconstructing a standard hemispherical object. Then, the whole 3-D surface of a seed was built by reconstructing its central and periphery region with PS separately for the two regions are of great colour differences. Finally, experiments were done to show that cross surface area of the reconstructed surface can be used as wrinkle identification parameter with the identification threshold of 9. Under this threshold more than 90% of wrinkled seeds could be successfully identified while only 5% of normal seeds being identified as wrinkled.