Computer Vision and Machine Learning in Agriculture

Abstract

The productivity and efficiency of a traditional agriculture crop production cycle can be improved by computer vision and machine learning. Innovative Information and Communication Technology (ICT) solutions in sensing, processing, and learning have attracted significant attention in precision agriculture. The power of visualizing the real world allows spatial or spatiotemporal information gathering and its representation. On the other hand, learning allows processing the information for planning, reasoning, and inference. The combination of computer vision and machine learning facilitates the automation of various tasks in the crop cycle. Compared to conventional agricultural practices, tasks are performed automatically and with greater fidelity. Computer vision has improved due to the following: availability of solid-state photo-detectors, increased sensor size, better spatial and spectral resolution, increased frame rate, and shutter speed. The development of remote sensing platforms like Unmanned Aerial Vehicles, terrestrial robots, and satellites allows capturing images at different scales. All the above factors increase the quality of the visual information. At the same time, information processing has become more accessible due to the growth of parallel computing (Graphics Processing Unit), efficient open-source machine learning libraries, large-scale annotated datasets, and reproducible research. Enrichment in information and increased processing power have seen innovations in precision agriculture like crop health and growth monitoring, prevention and control of pest and crop disease, automatic harvesting of crops, automated crop quality testing, and automated farm management. Computer vision and machine learning go hand in hand with precision agriculture, and therefore, they are seamlessly integrated into this chapter. It showcases publicly available large datasets and state-of-the-art algorithms developed while using them. After thorough analysis, it is evident that computer vision combined with intelligent techniques allows processing in more complex environments. It, in turn, will produce generalized and robust automation systems for precision agriculture.