|
This paper describes a method to discriminate objects based on scratches and patterns of objects photographed by smartphones. Researches on object fingerprints have been progressed as a technique for enabling identification of objects based on scratches and patterns, but there are two problems to determine object identity. First, if images are checked with strong feature points such as labels, fine feature points on the surface are ignored, causing misrecognition. The second is that it is impossible to extract sufficient feature points for discrimination when the angle of inclination is different between the database image and the captured image. In the experiment using the proposed method, performance evaluation was conducted by using 25 AC adapters and performing 625 collation. As a result of the experiments, we succeeded in classifying 25 AC adapters 100%.
|
