Research Of Ilumination Invariant For Color Image Of Mirosot

The MiroSot robot system targeting identification is mainly based on target color feature, and as the competition venues are large range of changes and there is uneven illumination such as the color of an object under the illumination changes.It is easy to form shade, shadow and highlight areas, which seriously affect the color image segmentation algorithm. So the color image processing in robot soccer become one of the bottlenecks.This paper analyzes the structure of the MiroSot robot system, highlights the importance of the visual system, and studies the light used for color image processing method. On the theory of color vision and its development has been studied, analyzed the models based on dichromatic reflection model, and studied several kinds of color space. By using the color histogram based on similarity calculation method in all components, they have been compared and researched. Experiments show that In the light of the considerable change in energy, the color space of rgb,h,s,yuv,clc2c3 ability to identify lower. The color space of 111213,m1m2m3,P1P2P3 are Not sensitive to light but They are nonlinear combination of RGB, is mainly used in image retrieval and other fields, not suitable for robot vision system.According to the color characteristics in the MiroSot robot system, the distinguishing method of the least relevant to 8 color components are extracted and used within-class scatter matrix and between-class scatter matrix and the trace criteria for the final proposed HSv model based on the dichromatic reflection model HSv color space.Finally, the paper introduces the vision system software component module functions. In the actual game platform validated color-based model of reflection HSv color model has a good color invariance.