Plant Species Classification through New Feature Extraction Model-Velocity Clamping Based Intersecting Cortical Model

Abstract

Plant classification is an active research area. The purpose of our current work is to develop a suitable feature extraction model. This paper suggests a technique to extract the geometric invariants of leaf images using a new velocity clamping based particle swarm optimized intersecting Cortical Model (VCPSO-ICM). Earlier geometric moments were assessed by transforms, separate normalization was used and they were costly. Intersecting cortical model (ICM) is used to avoid the usage of separate normalization for moment invariants of leaf images. In this model, the image is directly processed, as there is no need for preprocessing images. Parameters used in the intersecting cortical model (ICM) are difficult to set for each image separately. This is solved by our model. Time sequences are extracted from each image based on new parameters. Finally, a neural network is preowned to segregate the species of leaf images. This new feature evaluation model is tested on leaf snap database and results are compared with traditional Pulse Coupled neural network (PCNN), simplified Intersecting Cortical Model (ICM).This model achieves a higher accuracy than the existing methods.