CNN Based Technique for Systematic Classification of Field Photographs

Computer vision is the science that aims to contribute a similar, if not better, capability to a machine or computer. It is also interested with the idea and technology for building artificial systems that secure information from images. The digital image processing, image analysis and computer vision have become a principle part of artificial intelligence and the interface between the human and the machine grounded theory and enforced technologies. Implementing machine learning such as deep neural networks on agricultural data as immense attention in recent times. Convolution Neural Networks (CNN) are most commonly used in pattern and image detection problems as they have a number of advantages compared to other techniques. This paper proposes CNN architecture to segregate different plant images from the sequences collected. Following the preprocessing steps that will be including the elimination of blurriness or introducing the illumination change; CNN architecture is employed to extract the traits of the images that are present in the dataset that is created. Image classification using CNN involves dataset creation, training CNN, validating, testing CNN, prediction and at last classification. Here we use Keras software with backend as Theano and Tensor flow and we are able to predict the overall classification report with the accuracy of 43.98% for one cast of field photographs.