Abstract: This paper presented a new copper compositions soft sensor using color images of copper and LSSVR（least square support vector regression） modeling, as opposed to its delayed measurement by hardware measuring device. The soft sensor is based on a color vision system which includes computer and a color camera. First, a highperformance color chargecoupled device camera captures color from copper samples and transmits images to the computer. Then RGB, hue and color vector angle are used to characterize copper samples’ color for the comparative purpose. Three different LSSVR based soft sensor were developed using RGB, hue and color vector angle respectively. Finally, the real time estimation obtained by the soft sensor achieves superior performance. The experiments on test copper samples show that the proposed method is feasible and efficient.

Key words: copper composition, color vision, least square support vector regression（LSSVR）, soft sensor