Weed identification based on shape features and ant colony optimization algorithm

In order to improve the accuracy and efficiency of weed recognition, an identification method based on ant colony optimization (ACO) algorithm and support vector machine (SVM) is proposed. Firstly, shape feature parameters are extracted from the plant leaves after a series of image processing such as threshold segmentation, smooth processing and edge detection etc., and five geometric parameters and seven Hu-moment invariants which have useful properties is utilized to produce feature vectors. Then ACO algorithm in combination with SVM classifier is used to select the optimal feature set for classification. Finally, proposed approach has been applied on lab plant image database of cotton field and the experimental results have shown that the method can optimize feature subset and achieve an identification rate over 94% which is higher than using the original feature set.