Super-resolution Mapping Approach Research Articles

Generating annual high resolution land cover products for 28 metropolises in China based on a deep super-resolution mapping network using Landsat imagery

ABSTRACT High resolution of global land cover dynamic is indicative for understanding the influence of anthropogenic activity on environmental change. However, most of the land cover products are based on Landsat image that only has 30 m resolution, which is insufficient to distinguish the heterogenous urban structure; while very high spatial resolution image usually has low temporal resolution, which is difficult to monitor the urban dynamic. Deep-learning-driven super-resolution mapping is a prevailing way of achieving very-high-resolution land cover dynamic products in aspect of alleviating the mixed pixel problem of Landsat image. However, two limitations are obvious: 1) the fixed grid of kernel during the upsampling process favors spatial homogeneity and suppresses the learning of spatial heterogeneity of urban composition and 2) geometric or radiation variation over large spatial and long temporal extent in remote sensing images makes the super-resolution mapping approach difficult to transfer for application. Here, we attempt to solve these two limitations: 1) a progressive edge-guided super-resolution architecture is designed to allow nonuniformed kernel specific at the low-confidence edge region and intensify the learning of heterogenous compositions’ patterns and 2) an alternating optimization strategy is designed to minimize the resultant entropy and modulate the classification hyperplane to accommodate to the manifold of the discrepant region. Validation experiments are investigated based on a fine-grained and large-extent super-resolution (FLAS) dataset constructed in this study, and it is found that our approach remarkably enhances rich detailed patterns of heterogenous region and outperforms other state-of-the-art algorithms. Besides, we applied DETNet to the large spatial extent of 28 metropolises in China (>40,000 km2) and the large temporal extent of continuous 21-year (2000–2020) in Wuhan city to examine transferability. From the land cover areas variation, we find that the expansion rate of cropland is faster than the urban expansion over the past 10 years, which are gradually becoming the principal source for the encroachment of forest and lakes. From detailed urban dynamic reflected by the 21-year products, we find that urban-villages between the old city zone and the outer high-tech development zone are gradually disappeared. The captured dynamic is consistence with the urban-village renovation policy during this period, which is meant to redistribute the spatial configuration of the city for a more sustainable urban structure. We believe that the proposed method can facilitate a seamless and fine-grained observation system that can fill the weakness of the existing land cover activities and provide a brand-new insight into the urban dynamic and its underlying mechanism.

Super‐resolution mapping of hyperspectral satellite images using hybrid genetic algorithm

To assess the rate of sedimentation and the consequent reduction in the storage capacity, periodical capacity surveys of multi-purpose reservoirs is essential. Hydrographic surveys and acoustic surveys are time-consuming and expensive. The limited availability and high cost of the high-resolution images require a different methodology to accurately estimate the water-spread area of the reservoir. In this study, 30 m resolution hyperspectral image (hyperion) and multi-spectral image (The Earth Observing One (EO-1) advanced land imager) are used to estimate the water-spread area of the Peechi Reservoir, South India. A hybrid genetic algorithm (GA)-based super-resolution mapping approach is developed and demonstrated, which incorporates the multi-objective GA and Hopfield neural network (HNN). The hybrid GA-based super-resolution mapping approach gives a global optimum solution in half of the original computation time. Furthermore, mapping approach gives an error of 6.38% for the multi-spectral image and a lesser error of 3.86% for the hyperspectral image, while the HNN-based super-resolution mapping approach gives an error of 8.23% for the multi-spectral image and 5.71% for the hyperspectral image. Thus, in this work, an efficient technique based on hybrid GA is presented, which is a useful tool for accurate mapping of water bodies at the sub-pixel scale using hyperspectral imagery.

Super resolution land cover mapping of hyperspectral images using the deep image prior-based approach

ABSTRACTDue to the instantaneous field-of-view (IFOV) of the sensor and diversity of land cover types, some pixels, usually named mixed pixels, contain more than one land cover type. Soft classification can predict the portion of each land cover type in mixed pixels in the absence of spatial distribution. The spatial distribution information in mixed pixels can be solved by super resolution mapping (SRM). Typically, SRM involves two steps: soft class value estimation, which is similar to the image super resolution of image restoration, and land cover allocation. A new SRM approach utilizes a deep image prior (DIP) strategy combined with a super resolution convolutional neural network (SRCNN) to estimate fine resolution fraction images for each land cover type; then, a simple and efficient classifier is used to allocate subpixel land cover types under the constraint of the generated fine fraction images. The proposed approach can use prior information of input images to update network parameters and no longer require training data. Experiments on three different cases demonstrate that the subpixel classification accuracy of the proposed DIP-based SRM approach is significantly better than the three conventional SRM approaches and a transfer learning-based neural network SRM approach. In addition, the DIP-SRM approach performs very robustly about small-area objects within multiple land cover types and significantly reduces soft classification uncertainty. The results of this paper provide an extension for utilizing SRCNN to address SRM issues in hyperspectral images.

A genetic algorithm based approach to estimate the volume of a drinking water reservoir in Chennai city, South India, using multi-spectral satellite images

This paper is concerned with the use of satellite image processing and remote sensing approaches to determine the water-spread area and capacity of a drinking water reservoir in Poondi, near Chennai City, south India. Estimation of water-spread area and volume of the reservoirs using traditional methods like field surveys, acoustic surveys and hydrographic surveys is time-consuming, laborious and expensive. To overcome these problems, many researchers have used satellite images and the per-pixel (hard classification), the sub-pixel (soft classification) and the super resolution mapping approaches. The conventional hard and soft classification approaches do not give accurate results due to the presence of mixed pixels in the image scene of the periphery of the reservoir. In this work, a per-pixel algorithm (Maximum Likelihood), a sub-pixel algorithm (Fuzzy C Means) and a super resolution mapping algorithm (Genetic algorithm) are developed and the water-spread area is estimated for the reservoir using multi-date Landsat8 OLI images. The volume of the reservoir at different water levels are estimated using the water-spread area and the Trapezoidal formula. The capacity estimated from satellite image-derived area is compared with the capacity data obtained from the Poondi reservoir authority. The error in estimation due to the per-pixel approach is 16.74%, while it is 9.06% for the sub-pixel approach and a mere 3.25% for the super-resolution approach. Thus, the super-resolution mapping approach results in minimum error when compared to the sub-pixel approach which in turn gives lesser error result when comparing with the per-pixel approach.

Object-Based Superresolution Land-Cover Mapping From Remotely Sensed Imagery

Superresolution mapping (SRM) is a widely used technique to address the mixed pixel problem in pixel-based classification. Advanced object-based classification will face a similar mixed phenomenon—a mixed object that contains different land-cover classes. Currently, most SRM approaches focus on estimating the spatial location of classes within mixed pixels in pixel-based classification. Little if any consideration has been given to predicting where classes spatially distribute within mixed objects. This paper, therefore, proposes a new object-based SRM strategy (OSRM) to deal with mixed objects in object-based classification. First, it uses the deconvolution technique to estimate the semivariograms at target subpixel scale from the class proportions of irregular objects. Then, an area-to-point kriging method is applied to predict the soft class values of subpixels within each object according to the estimated semivariograms and the class proportions of objects. Finally, a linear optimization model at object level is built to determine the optimal class labels of subpixels within each object. Two synthetic images and a real remote sensing image were used to evaluate the performance of OSRM. The experimental results demonstrated that OSRM generated more land-cover details within mixed objects than did the traditional object-based hard classification and performed better than an existing pixel-based SRM method. Hence, OSRM provides a valuable solution to mixed objects in object-based classification.

Estimation of Water-Spread Area of Singoor Reservoir, Southern India by Super Resolution Mapping of Multispectral Satellite Images

Estimation of reservoir water-spread area is often carried out by field surveys which are cumbersome, time consuming, expensive and involves more man power. Hence, such surveys cannot be carried out periodically. To overcome this issue, satellite images are used, wherein the reservoir water-spread is estimated by conventional per-pixel classification such as the maximum likelihood and minimum distance to mean approaches that often results in inaccurate estimate of water-spread area due to the presence of mixed pixels. High cost and non-availability of high resolution images demands the use of an alternative approach that can give accurate information about the reservoir water-spread area. In this work, IRS, LISS III images of moderate (24 m) resolution were used for accurate estimation of the water-spread area of Singoor reservoir, southern India. The reservoir water-spread areas were extracted using per-pixel classification, sub-pixel classification and super resolution mapping approaches. These results were validated with the water-spread areas obtained from field data of the same dates. The error produced by the per-pixel approach was 6.66%, while it was 4.37% for the sub-pixel approach and a meagre 1.71% for the super-resolution approach. Fairly less error produced by the super resolution mapping technique implies that it is an efficient approach for accurate quantification of reservoir water-spread area. The estimated water-spread can be used in a simple volume estimation formula to estimate the capacity of the reservoir.

Learning-Based Spatial–Temporal Superresolution Mapping of Forest Cover With MODIS Images

Forest mapping from satellite sensor imagery provides important information for the timely monitoring of forest growth and deforestation, bioenergy potential assessment, and modeling of carbon flux, among others. Due to the daily global revisit rate and wide swath width, MODerate-resolution Imaging Spectroradiometer (MODIS) images are used commonly for satellite-derived forest mapping at both regional and global scales. However, the spatial resolution of MODIS images is too coarse to observe fine spatial variation in forest cover. The last few decades have seen the production of several fine-spatial-resolution satellite-derived global forest cover maps, such as Hansen’s global tree canopy cover map of 2000, which includes abundant spectral, temporal, and spatial prior information about forest cover at a fine spatial resolution. In this paper, a novel learning-based spatial–temporal superresolution mapping approach is proposed to integrate both current MODIS images and prior maps of Hansen’s tree canopy cover, to map present forest cover with a fine spatial resolution. The novel approach is composed of three main stages: 1) automatic generation of 240-m forest proportion images from both 240- and 480-m MODIS images using a nonlinear learning-based spectral unmixing method; 2) downscaling the 240-m forest proportion images to 30 m to predict the class possibilities at the subpixel scale using a temporal-example learning-based downscaling method; and 3) final production of the fine-spatial-resolution forest map by solving a regularization-based optimization problem. The novel approach produced more accurate fine-spatial-resolution forest cover maps in terms of both visual and quantitative evaluation than traditional pixel-based classification and the latest subpixel based superresolution mapping methods. The results show the great efficiency and potential of the novel approach for producing fine-spatial-resolution forest maps from MODIS images.

Delineation of sub-pixel level sedimentary litho-contacts by super resolution mapping of Landsat image

Abstract. To delineate the geological formation at the surface, satellite image classification approaches are often preferred. This study aims to produce a super resolved map with better delineation of the litho-contacts from the medium resolution Landsat image. Conventionally used per-pixel classification provides an output map at the same resolution of the satellite image, while the super resolved map provides the high resolution output map using the medium resolution image. In this study, four test sites are considered for delineating different litho-contacts using super resolution mapping approach in Cuddalore district, southern India. The test sites consists of charnockite, fissile hornblende-biotite gneiss, marine sandstone and sandstone with clay, limestone with calcareous shale and clay, clay with limestone bands/lenses, mio-pliocene and quaternary argillaceous and calcareous sandstone, fluvial and fluviomarine formations. This work compares the per-pixel, super resolved output derived from linear spectral unmixing (LSU) based HNN and spectral angle mapper (SAM) based HNN approaches. The super resolution mapping approach was performed on the medium resolution (30 m) Landsat image to obtain the litho-contact maps and the results are compared with the existing maps and observations from field visits. The results showed improved accuracy (90.92%) of the map prepared by the SAM based super resolution approach compared to the LSU based super resolution approach (90.14%) and the maximum likelihood classification approach (83.74%). Such an improved accuracy of the super resolved map (6 m resolution) is due to the fact that the lithological mapping is done not merely at the resolution of the image, but at the sub-pixel level. Hence, it is inferred that super resolution mapping applied to multispectral images may be preferred for mapping lithounits and litho-contacts than the conventional per-pixel and sub-pixel image classification methods.

Super-resolution mapping of hyperspectral images for estimating the water-spread area of Peechi reservoir, southern India

Though the estimation of the water-spread area in reservoirs is often carried out by field surveys, it is time-consuming and tedious, and cannot be done periodically. To overcome this issue, satellite images are often used where the estimation is made through density slicing or conventional per-pixel classification. This results in an inaccurate estimation of reservoir capacity. The high cost and nonavailability of high-resolution images demands the use of an alternative approach that can give accurate information about the reservoir water-spread area. A hyperspectral image (Hyperion) of moderate resolution is used for the accurate estimation of the water-spread area of Peechi reservoir, southern India. The reservoir water-spread area obtained from per-pixel classification, subpixel classification, and super-resolution mapping approaches are compared with the water-spread area obtained from the ground truth hydrographic survey data. It is observed that the water-spread area estimated from the hyperspectral image by the per-pixel approach is 7.66 sq km, that by the subpixel approach is 6.34 sq km, and that by the super-resolution approach is 5.69 sq km compared to the actual area of 5.95 sq km. The classification accuracy estimated for the Hopfield neural network based super-resolution technique is 92.97%, whereas that for the conventional classifier (maximum likelihood) is 86.72%. This improved accuracy in classification resulted in an accurate estimation of water-spread area. Hence, it is inferred that super-resolution mapping applied to hyperspectral images is a computationally efficient approach for the accurate quantification of reservoir water-spread area.

Interpolation-based super-resolution land cover mapping

Super-resolution mapping (SRM) is a technique to produce a land cover map with finer spatial resolution by using fractional proportion images as input. A two-step SRM approach has been widely used. First, a fine-resolution indicator map is estimated for each class from the coarse-resolution fractional image. All indicator maps are then combined to create the final fine-resolution land cover map. In this letter, three popular interpolation methods, Inverse Distance Weighted (IDW), Spline and Kriging, as well as four indicator map combination strategies, including the maximal value strategy and the sequential assignment strategy with and without normalization, were assessed. Based on the application to two simulated images, the performance of all SRM algorithms was assessed. The results show that the two-step SRM approach can obtain smoother land cover maps than hard classification. An increase in zoom factor results in the appearance of numerous small patches and linear artefacts in the SRM results. The accuracies of Spline and Kriging are similar and are both higher than that of IDW. The maximal value strategy can generate a smoother land cover map than the sequential assignment strategy in most cases, and a normalizing indicator value has a mixed effect on the result.

A super-resolution mapping method using local indicator variograms

Super-resolution mapping (SRM) is a recently developed research task in the field of remotely sensed information processing. It provides the ability to obtain land-cover maps at a finer scale using relatively low-resolution images. Existing algorithms based on indicator geostatistics and downscaling cokriging offer an SRM approach using spatial structure models derived from real data. In this article, a novel SRM method is developed based on a sequentially produced with local indicator variogram (SLIV) SRM model. In the SLIV method, indicator variograms extracted from target-resolution classification are produced from a representative local area as opposed to using the entire image. This simplifies the application of the method since limited target-resolution reference data are required. Our investigation on three diverse case studies shows that the local window (approximately 2% of the entire study area) selection process offers comparable accuracy results to those using globally derived spatial structures, indicating our methodology to be a promising practice. Furthermore, comparison of the proposed method with random realizations indicates an improvement of 7–12% in terms of overall accuracy and 15–18% in terms of the kappa coefficient. The evaluation focused on a 270–30 m pixel size reconstruction as a potential popular application, for example moving from Moderate Resolution Imaging Spectroradiometer (MODIS) to Landsat-type resolutions.