Abstract: In order to improve the recognition rate of plant species, leaf image algorithm was proposed. Firstly, the plant species characteristic model was established, including leaf color, shape and texture characteristics. Secondly, relations of input, output and hidden layer of radial basis function neural network was built. Thirdly, the numbers of radial function, center and width was optimized, weighting parameter was calculated based on gradient descent method, and learning rate was adaptively adjusted. Finally, plant species recognition process was given. Plant leaf color, shape and texture characteristic numbers of simulation were selected as 6, 7 and 7, the three characteristics average recognition rate of improved radial basis function neural network algorithm was 93.5%, higher than single and two characteristics, the shape characteristic had maximum recognition rate.

Key words: leaves, plant species, radial basis functions, image algorithm