A BINARY COLOR-SIFT DESCRIPTOR FOR PRODUCT RECOGNITION SYSTEMS

Wei-Jong Yang ； Wei-Hau Du ； Pau-Choo Chang ； Jar-Ferr Yang

《TANET2016 臺灣網際網路研討會》 2016 (2016/11) Pp. 328-331-060

https://doi.org/10.6679/TANET.2016.60

並列摘要

The demands of smart recognition technologies realized in smart devices for human assistances are increased daily in recent years. This paper proposed a product recognition system which combines binary scale-invariant feature transform (SIFT) [1], [2], bag of words model (BoW) [3], [4] and support vector machine (SVM) [5]. Since the traditional SIFT features are mostly in the gray scale, the color information is missing. In image processing, color information is still an important information for recognition of products. Thus, a color- SIFT descriptor which is based on the traditional binary SIFT is proposed. The color-SIFT descriptor combines the original 128 dims descriptor of SIFT with different color information bits around the feature point. With the color- SIFT, we can discard some unreliable matching pairs and increase the robustness of matching tasks. The experimental results show that the proposed product recognition system with color-SIFT achieves higher recognition rate than that with the traditional SIFT and other features in various situations.

並列關鍵字

SIFT ； color-SIFT ； color moments product recognition system