Sub-pixel mapping with point constraints

Abstract

Remote sensing images contain abundant land cover information. Due to the complex nature of land cover, however, mixed pixels exist widely in remote sensing images. Sub-pixel mapping (SPM) is a technique for predicting the spatial distribution of land cover classes within mixed pixels. As an ill-posed inverse problem, the uncertainty of prediction cannot be eliminated and hinders the production of accurate sub-pixel maps. In contrast to conventional methods that use continuous geospatial information (e.g., images) to enhance SPM, in this paper, a SPM method with point constraints into SPM is proposed. The method of fusing point constraints is implemented based on the pixel swapping algorithm (PSA) and utilizes the auxiliary point information to reduce the uncertainty in the SPM process and increase map accuracy. The point data are incorporated into both the initialization and optimization processes of PSA. Experiments were performed on three images to validate the proposed method. The influences of the performances were also investigated under different numbers of point data, different spatial characters of land cover and different zoom factors. The results show that by using the point data, the proposed SPM method can separate more small-sized targets from aggregated 2 artifacts and the accuracies are increased obviously. The proposed method is also more accurate than the advanced radial basis function interpolation-based method. The advantage of using point data is more evident when the point data size and scale factor are large and the spatial autocorrelation of the land cover is small. As the amount of point data increases, however, the increase in accuracy becomes less noticeable. Furthermore, the SPM accuracy can still be increased even if the point data and coarse proportions contain errors.

Aerial Scene Parsing: From Tile-level Scene Classification to Pixel-wise Semantic Labeling

“…Even with a great improvement of classification accuracy, sub-pixel class composition estimated by fuzzy classification and spectral mixture analysis cannot provide the spatial distribution of land cover classes within pixels. To address this issue, the sub-pixel mapping approaches are developed [88], [95], [107], [108]. In this scheme, each pixel is divided into subpixels which are predicted to get single semantic labels.…”

General solution to reduce the point spread function effect in subpixel mapping

“…As an ill-posed problem, however, uncertainty exists inevitably in SPM and effective usage of supplementary information has generally been a focus of SPM. Various sources of additional information have been used to tackle the uncertainty issue in SPM over the past decades, such as digital elevation models (DEM) (Ling et al, 2008), multiple shifted images (Ling et al, 2010;, point constraints (Wang et al, 2020), height information from Light Detection And Ranging (LiDAR) elevation (Nguyen et al, 2005), training images (Jia et al, 2019;Ling and Foody, 2019), and panchromatic images (Nguyen et al, 2011).…”

Fast and Slow Changes Constrained Spatio-Temporal Subpixel Mapping

“…SPM is a typical ill-posed problem: multiple land cover maps at the target fine spatial resolution can fulfill the coherence constraint exerted by the coarse proportions [29]. To address this issue, various auxiliary data have been considered to decrease the uncertainty in SPM, such as vector data [30], color images [31], panchromatic images [32], training images [33] and subpixel shifted images [34], [35] and point data [36]. Learning-based SPM methods have been also developed by using fine spatial resolution training data [37][38][39][40].…”