Weed Classification Analysis Using Localized Multiple Kernel Learning (LMKL)

Abstract

Weed classification is a need in the agricultural research to improve the weed control system. There are many kernel-based learning algorithms to identify weed images proposed in the literature; however, most of the weed classification technique proposed a single kernel-based algorithm. Recently, the Localized Multiple Kernel Learning (LMKL) instead of using a single kernel has been proposed for the classification technique that can enhance the interpretability of the decision function and improve performances. LMKL is composed of a kernel-based learning algorithm and a parametric gating model to assign local weights to kernel functions. These two components are trained in a coupled manner using a two-step alternating optimization algorithm. The learning algorithm is derived from three different gating models (softmax, sigmoid, and Gaussian), which applies the LMKL framework on the machine learning problems of binary classification. Therefore, in this work, feature vectors of weed images extracted using the Gabor Wavelet and the Fast Fourier Transform (FFT) were employed to analyze weed pattern images using LMKL algorithms. The result with the aid of gating model are visualized and discussed to prove the performance of LMKL classifier. The results showed the visualization using six types of combinations kernels for all set feature vectors are different for each weed dataset.