Plant identification method

Abstract

The invention discloses a plant identification method, and the method comprises the steps: carrying out the preprocessing: enabling a plant image to be converted into a binary image; feature extraction, which is used for extracting relative features in the image to form a feature data matrix; the classification and identification module is used for constructing a neural network model based on deeplearning, the model adopts a five-layer BP neural network and four full connection layers, and the last layer is a normalized layer; and the model is trained for subsequent plant identification. According to the method, the representative eight relative characteristics are adopted, so that the method can be suitable for identification of most plants, and compared with the prior art, more plant types can be identified to a greater extent, and the identification accuracy is improved.

Description

technical field [0001] The invention relates to the field of machine vision research, in particular to a plant recognition method. Background technique [0002] Plant cluster recognition technology has practical significance for plant classification and identification, protection and utilization of plant resources, exploration of plant relationship, clarification of plant evolution law, and agricultural application. [0003] Most of the current computer-aided plant classification methods are based on the characteristics of leaf shape. This type of method mainly includes two aspects, feature extraction and classification algorithm. [0004] Feature extraction methods include methods of manually extracting features, methods of extracting features using the curvature of plant leaf edges, and methods of extracting features such as color, shape, and texture. These methods are greatly affected by the shooting angle and distance of leaves, which are not convenient for subsequent ...

AI Technical Summary

Problems solved by technology

However, these methods are not suitable for multi-plant classification, and can only solve the classification of specific plant leaves. When training on large data of plants with various leaves, the recognition rate will drop.

Images