IDENTIFICATION OF CASSAVA LEAF DISEASES BASED ON OPTIMIZED SUPPORT VECTOR MACHINE

ABSTRACT

Cassava is one of the principal crops farmed and extensively consumed in Nigeria, especially in the Southwest. While cassava is widely used and gives many households a source of income, it is prone to several diseases, including cassava brown streak disease (CBSD) and cassava bacterial blight disease (CBBD). Farmers find it very difficult to diagnose these illnesses with their unaided eyes because it takes a lot of time, and effort, and is inaccurate. As a result, this study developed Binary Particle Swarm Optimization-Support Vector Machine (BPSO-SVM) and Reptile Search Algorithm-Support Vector Machine (RSA-SVM) models to categorise cassava leaves into brown streak disease and bacterial blight disease. 1,248 datasets were obtained from the Kaggle village datasets which were later resized and converted to grayscale images. The images were further sharpened using morphological filtering, denoised using adaptive median filtering, and contrast-enhanced using bi-histogram equalization. Segmentation was done using Sobel edge detection, and texture and colour features was extracted using the gray-level co-concurrence matrix with the haralic statistical feature and the four-colour moment method, respectively. It is found that the accuracy of BPSO-SVM is 95.27% and 95.78% for the CBSD and CBBD cassava datasets, respectively, while the accuracy of RSA-SVM is 93.73% and 93.99%, respectively. Nevertheless, only two of the many diseases affecting cassava were identified in this study, and no strategies for preventing these diseases were provided.

Keywords: Binary Particle Swarm Optimization (BPSO), cassava bacterial blight disease (CBBD), cassava brown streak disease (CBSD), Reptile Search Algorithm (RSA), Support Vector Machines (SVM)