School of Automation, Guangdong University of Technology, Guangzhou, Guangdong, China

---

Abstract—In the process of color classification of vehicles, the accurate segmentation of color regions and the elimination of non-color interference regions remain to be dealt with. Therefore, a vehicle color algorithm based on two-layer saliency map, illumination invariant transformation, and adaptive KNN is proposed in this paper. A two-layer saliency map is used to remove interference regions independent of the color of vehicles. The graph is transformed and finally classified based on the adaptive k nearest neighbor algorithm. The ex-perimental results demonstrate that the method can accurately extract the body of the vehicles to a certain extent, and preprocessed with illumination invariance transformation, colors of vehicles can be accurately classified even in dark and reflective environments. The further work of this study is to extract slightly deeper features and directly obtain the preliminary saliency graph based on the decoder processing

Index Terms—Color classification of vehicles, two-layer saliency, illumination-invariant transformation, adap-tive KNN.

[PDF]

Cite: Qihua Huang, Qilv Li, Guoheng Huang, "Color Classification of Vehicles Based on Two-Layer Salincy, Illumination-Invariant Transformation, and Adaptive KNN," Journal of Software vol. 14, no. 10, pp. 479-487, 2019.