Improvement of Complex Background Crop Image Segmentation using Sparse PSO

Crop production has a direct impact on the nation’s economy. At the early stage, the crop is affected by bacterial and fungal infections. Hence, real-time application needs quick identification of disease to reduce the production loss. In this paper, multilevel thresholding of crop images is achieved by recursive minimum cross entropy (R-MCE). As the level of thresholding increases, the computation time increases. Hence, Sparse PSO has been amalgamated with R-MCE to reduce the computation time. The color crop images are converted into L*a*b* color space to realize it devises color independently. Moreover, the proposed technique uses a Gaussian filter for smoothing. The segmented image is evaluated at different levels of thresholding which helps to analyze the crop images by color variation. It is compared with PSO, and PSO variants such as constriction factor, linear decay inertia, dynamic inertia, nonlinear inertia, and constant weight inertia to confirm the accuracy. The outcome of the proposed method validates the accuracy in terms of fidelity parameters. It takes less time as compared to other reported algorithms which meets the requirement of real-time analysis.