Spectral-Spatial Active Learning With Structure Density for Hyperspectral Classification

Abstract

In this paper, a spectral-spatial active learning (AL) method is proposed based on an upto-date unlabeled samples sampling strategy concentrated on the structure density supported by breaking ties. The proposed sampling criterion in AL is used for hyperspectral image classification, which involves several steps: First, superpixel segmentation algorithm is conducted on the HSI to cluster pixels with similar spectral-spatial signature into the same superpixel block. Then, density peak clustering technique is performed on the each superpixel block to obtain structure density of the pixels. Meanwhile, probabilitybased classifier is employed to achieve the probability distributions of pixel. Next, breaking ties (BT) score of each pixel can be calculated by exploiting the probabilities. Additionally, a fusion mechanism is introduced to select the unlabeled samples with representativeness and informativeness advantages by employing the BT-assisted structure density(SD sampling criterion) of each pixel. Finally, the samples with manual labeled class labels are put into the training set to retrain the classifier. Experimental results manifest that the proposed SD-based sampling criterion in active learning can significantly improve the classification accuracy in few labor costs. Thus, it has certain feasibility in practical application.INDEX TERMS Hyperspectral imagery, active learning, structure density, breaking ties, sampling criterion.

Gabor Filter-Based Multi-Scale Dense Network Hyperspectral Remote Sensing Image Classification Technique

“…Imaging wavelength range 0.4-2.5µm, spectral resolution of 10µm, spatial resolution of 20m, is in the continuous 220 band continuous imaging of the ground, but the use of 20 cannot be water reflection of the band, leaving 200 bands as the object of study. There are 16 types of features in the dataset, and the samples are unevenly distributed, and the sample distribution is shown by Table The Pavia University dataset (abbreviation: PU) was acquired by the Airborne Optical Spectral Imager (ROSIS) in 2003 at the University of Pavel, Italy, in the wavelength range 0.43-0.86 µm [19]. was used for hyperspectral image classification.…”

Multi-View Structural Feature Extraction for Hyperspectral Image Classification

“…Duan et al modeled hyperspectral image as a linear combination of structural profile and texture information, in which the structural profile was used as the spatial features [15]. After that, many improved approaches have been also studied for classification of HSIs, such as ensemble learning [16,17], semisupervised learning [18,19], and active learning [20,21]. For instance, in [17], a random feature ensemble method was proposed by using ICA and edge-preserving filtering to boost the classification accuracy of HSIs.…”

Features extraction from multi-spectral remote sensing images based on multi-threshold binarization

“…From the perspective of processing and feature extraction of multispectral images, decomposition methods have proven to be effective, as they enable the detailing of potential hidden features 26 . In particular, approaches utilizing processing in the spectral domain 27 and methods based on the use of subpixel data 28 for multispectral images can be distinguished. Feature formation technology based on threshold approaches works well for multichannel images 29 , making it possible to apply it to multispectral images.…”