The identification of leaves is of great significance to the identification and classification of plant species and the exploration of the phylogenetic relationship between plants. The accuracy of existing leaf identification technologies needs to be improved. This paper quantifies the image samples, optimizes support vector machine parameters through an improved grid search method, and establishes a GSM-SVM prediction model. The shape features and leaf vein texture are used as discrimination indexes, and the types of leaves are discriminated based on the extracted data information. At the same time, the comparative analysis of traditional BP neural network algorithm is introduced to consider the superiority of the model. Finally, the improved GSM-SVM classification prediction model has a classification accuracy rate of 97.5%, which is obviously better than traditional recognition methods. This method embodies the application of intelligent recognition technology in botany, and it is also conducive to the widespread development of recognition technology.
