Abstract: Coverage is an important quantitative index for plant community structure. Traditional visual estimation of the coverage of plant communities not only had low precision but was also influenced by subjectivity. After analyzing the RGB (red-green-blue) mode of field photographs from desert grassland, the algorithms of an RGB color decision tree were constructed to calculate the vegetation coverage through distinguishing plant and non-plant digital pixels. Software was constructed with a friendly interface after: integrated human-computer interaction convenience, integrated function blocks of digital photograph import, object identification, coverage calculation, results display, and comparisons in the Matlab 7.0 platform, had achieved accurate and efficient calculation of grass coverage and facilitated grassland field investigations. Compared with acupuncture, the maximum absolute deviation of this algorithms was less than 5%, and the resulting vegetation coverage accuracy was more than 95%. This method could solve the problems of the heavy workload and inaccurate quantitative determination in normal manual estimation of grassland coverage.