Recognition and detection of tea leaf's diseases using support vector machine

• Published in: 2018 IEEE 14th International Colloquium on Signal Processing & Its Applications (CSPA) pp. 150 - 154
• Main Authors: Hossain, Selim, Mou, Rokeya Mumtahana, Hasan, Mohammed Mahedi, Chakraborty, Sajib, Razzak, M. Abdur
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.03.2018
• Subjects: Agriculture; Disease detection; Disease recognition; Diseases; Feature extraction; Image color analysis; Image processing; Image segmentation; Support Vector Machine; Support vector machines
• DOI: 10.1109/CSPA.2018.8368703

Tea is a popular beverage all around the world, and in Bangladesh the cultivation of tea plays a vital role. Many diseases affect the proper growth of tea leaves leading to its reduction, thus hindering of the production of tea. However, if the disease is identified at an early age it would solve all the above mentioned problems through the application of appropriate treatment, or through the pruning of the diseased leaves to prevent further spread of the disease. To solve this problem image processing is the best option to detect and diagnose the disease. The main goal of this research is to develop an image processing system that can identify and classify the two most widespread tea leaf diseases in Bangladesh, namely brown blight disease and the algal leaf disease, from a healthy leaf. Disease identification is the first step; there are many methods that have been used for identifying the leaf disease. In this paper, Support Vector Machine classifier (SVM) is used to recognize the diseases. Eleven features are analyzed during the classification. These features are then used to find the most suitable match for the disease (or normality) every time an image is uploaded into the SVM database. When a new picture is uploaded into the system the most suitable match is found and the disease is recognized. The approach is novel since the number of features compared by the SVM classifier is reduced by three features compared to previous researches, without adversely sacrificing the success rate of the classifier, which retains an accuracy of more than 90%. This also speeds up the identification process, with each leaf image taking 300ms less processing time compared to previous research using SVM, thus ensuring a greater number of leaves can be processed in a given time frame. The proposed solution increases in efficiency of the detection, identification, and classification process will enable the tea industry in Bangladesh to become more competitive globally, by reducing the losses suffered due to diseases of the leaf, and thus increasing the overall tea production rate.