Indian Plant Species Identification under Varying Illumination and Viewpoint Conditions

The emergence and development of plant diseases and pest outbreaks have become more common nowadays due to the unsettled climate and environmental conditions. Actions controlling diseases or remedial measures can be undertaken if the symptoms are identified at an early stage. This would help the farmer in detecting and controlling plant diseases, thereby controlling the financial losses. We present a method for automatically recognizing the plant species based on leaf shape for five different species of common garden plants, Anant (Gardeniajasminoides), Aboli (Crossandra), Chandani (Crapejasmine), Jui (Common Jasmine) and Jaswand (hibiscus). The work focuses on identifying garden plants species which will act as an input to a decision support system (DSS) that would be developed for giving advice to farmers as and when required over mobile internet. The proposed system is comprised of four main stages. First, is image acquisition that can be done using mobile camera with minimum resolution of 2 mega pixels. The images are captured with differing scales, different viewpoint and at different time of day. Second stage is pre-processing and segmentation. Third stage detects the most discriminable set of features. Scale invariant feature transform (SIFT) is used for detecting SIFT features and finally using these features SIFT feature matching is done using k-d tree. The experimental results gave an accuracy of about 85% suggesting that SIFT can be used for plant species identification.