Full sphere recognition enhancement with cast shadow images

Abstract

Image classification and recognition algorithms are useful for industrial manufacturing processes; however, the algorithms have limitations. If the products have no special features, images may be unclear. Then, it is difficult to correctly classify or recognize a product using traditional algorithms. Therefore, this paper proposes a technique to enhance a recognition algorithm by using additional information from cast shadow images of the objects. The technique is applied for classifying full spheres, including hemispheres, cylinders, boxes, octahedron, and nonstandard shape. The study's result shows that the algorithm can successively classify full spheres. By comparing the shape of a full sphere cast shadow, the full sphere can be recognized. The success of this technique can also be applied to other classification techniques and recognition problemss such as fruit and rice grain classification.