Plant-Seedling Classification Using Transfer Learning-Based Deep Convolutional Neural Networks

Background

In recent years, various projects have dealt with automated recognition of weeds using cameras with the aim of developing new farming machinery that can control the weeds more intelligently. It is then of major technical and economic importance to implement computer based methods for reliable and fast identification and classification of weed/crop for empowering machinery that can control weed invasion. The first step in both of these is to identify weeds using their images. This is accomplished using computer vision. Automatic systems can be based on images of weed and crop plants from which classification features associated to size, shape, texture and colour can be readily obtained. For this task numerous image processing algorithms are available which complement with classification methods to make the field of machine vision suitable for weed identification. In this section, the literature related to this area is presented. Deep learning is the current favorite choice of researches working around the field of image processing. It can provide great deal of help in the problem of classifying images of plants as weed or crop. Application of deep learning in agriculture-related literature is presented to observe how deep learning has helped to resolve problems in agriculture.