Target recognition in different color spaces

Color-based recognition method is very simple for programming and has been deeply studied, but most of the previous works are based on computers. This paper studies the issue of target recognition based on monocular vision using the TI´s (Texas Instruments) ICETEK-DM642-BR (an evaluation board). The algorithms for target recognition are based on three different color spaces. Images and videos to be processed have a large amount of data. Therefore, the complicated algorithms are required to ensure the accuracy of recognition. DSPs have limited computation power and can´t satisfy the real-time requirements. Thus, it is critical to find an optimal algorithm to keep balance between accuracy and real-time requirements. Experimental studies in RGB, HSV and CIE-XYZ color spaces have been carried out. Comparison study of the characteristics and effects of the algorithms in these three different color spaces has been included.