Convolution Neural Network based Weed Detection in Horticulture Plantation.

Abstract

Weed classification plays a vital role in horticulture plantation to optimally use herbicides on the weeds and maintain the quality of crops. This paper discuses weed classification with the combination of Image Processing and Convolution Neural Network (CNN). The image with both crop and weed is segmented and CNN is used for classification of crop and weed. The experimental database comprises rgbweeddetection dataset from the github portal which consists of images of carrot plant along with weed. The complete approach is implemented using Python Programming along with keras and tensorflow Libraries of Python. The experimental results show 95% accuracy in classification using CNN with lower order of convolution and maxpooling layers supported by reduced rate of misclassification of weed and crop.