Publication

Abstract : Weed management is one of the prime obstacles faced by most of farmers nowadays. Efficient weed detection methods will cut back the price of weed management. Feature extractors have an important role in the domain of computer vision. The feature extracting algorithm takes the image as its input, and then it gives back the feature descriptors of the image that can be used to discriminate one feature from another. In software systems, there are various binary descriptors that are widely used for face recognition, plant discrimination, fingerprint detection, etc. This paper shows the performance comparison of different binary descriptors like local directional relation pattern (LDRP), local directional order pattern (LDOP), and local binary pattern (LBP) with support vector machine (SVM) for the image set classification. The results indicate that the sequence of LBP and SVM together produce a better accuracy of 84.51% in “bccr-segset” plant leaf database when compared to LDOP which produced an accuracy of 75% and LDRP with an accuracy of 75.56%.