Applying computer vision for detection of diseases in plants

Early detection and quantification of diseases in food plants are critical to agriculture industry and national food security. However, limitation in technology and cost has limited the success of applying Computer Vision in Plant Science. This research builds on the recent advance of Machine Learning, GPU and smartphones to tackle the problem of fast and low cost diagnosis of plant disease. In particular, we choose soybean as the subject for applying automatic disease detection. The reason is because soybean is an important crop for the state of Iowa and an important source of food for America. The plant is however, highly vulnerable to several type of diseases. This thesis consists of two sub-analyses of soybean diseases, which are: First, detection of a single disease in soybean, particularly Sudden Death Syndrome (SDS) with high detail (including location and severity). Second, detection of multiple diseases in soybean, using mobile phones which are resource- constraine