ABSTRACT
With the development of agricultural engineering in the 21st century, modern technologies of electronics, information and automation are introduced into traditional agricultural vehicles to realize automation and intelligence, so that the labor can be reduced, pollution caused by agricultural chemical can be avoided, and agricultural cost can be decreased. Based on these ideas, we have studied the method of cotton recognition for cotton harvesting robot and developed an image processing algorithm of cotton recognition. The main and creative achievements include the following aspects:
1. The color spaces widely used in the computer vision are analyzed. The existing color spaces have RGB, normalized rgb, HIS, YCrCb, L*a*b*>and I1I2I3 color space. Color data obtained from three parts of cotton, namely cotton fruits, leaves and cotton stems, is analyzed using SPSS for Windows in different color spaces. And we have drawn several conclusions:
1) In RGB color space, the mean values of R, G and B of cotton fruit are almost equal. However, the mean value of Green of cotton leaves and that of Red of cotton stems has a bigger weight. In general, there exists superposition between R, G and B. It would lead to high error rate if only one of variable of R, G and B was as a threshold to segment the images.
