Landsat 7 thermal-IR image sharpening using an artificial neural network and sensor model

G.P. Lemeshewsky; R.A. Schowengerdt

doi:10.1117/12.438256

Home/Archive/USGS

USGS · 70023672

Landsat 7 thermal-IR image sharpening using an artificial neural network and sensor model

G.P. Lemeshewsky·R.A. Schowengerdt

View original source ↗Export metadata ↓

Abstract

The enhanced thematic mapper (plus) (ETM+) instrument on Landsat 7 shares the same basic design as the TM sensors on Landsats 4 and 5, with some significant improvements. In common are six multispectral bands with a 30-m ground-projected instantaneous field of view (GIFOV). However, the thermaL-IR (TIR) band now has a 60-m GIFOV, instead of 120-m. Also, a 15-m panchromatic band has been added. The artificial neural network (NN) image sharpening method described here uses data from the higher spatial resolution ETM+ bands to enhance (sharpen) the spatial resolution of the TIR imagery. It is based on an assumed correlation over multiple scales of resolution, between image edge contrast patterns in the TIR band and several other spectral bands. A multilayer, feedforward NN is trained to approximate TIR data at 60m, given degraded (from 30-m to 60-m) spatial resolution input from spectral bands 7, 5, and 2. After training, the NN output for full-resolution input generates an approximation of a TIR image at 30-m resolution. Two methods are used to degrade the spatial resolution of the imagery used for NN training, and the corresponding sharpening results are compared. One degradation method uses a published sensor transfer function (TF) for Landsat 5 to simulate sensor coarser resolution imagery from higher resolution imagery. For comparison, the second degradation method is simply Gaussian low pass filtering and subsampling, wherein the Gaussian filter approximates the full width at half maximum amplitude characteristics of the TF-based spatial filter. Two fixed-size NNs (that is, number of weights and processing elements) were trained separately with the degraded resolution data, and the sharpening results compared. The comparison evaluates the relative influence of the degradation technique employed and whether or not it is desirable to incorporate a sensor TF model. Preliminary results indicate some improvements for the sensor model-based technique. Further evaluation using a higher resolution reference image and strict application of sensor model to data is recommended.

Explore related subjects

Conference Paper

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

G.P. Lemeshewsky, R.A. Schowengerdt. 2001. Landsat 7 thermal-IR image sharpening using an artificial neural network and sensor model. https://doi.org/10.1117/12.438256

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Time-to-depth conversion of seismic-reflection data from eastern Lake Superior and implications for the eastern arm of the Midcontinent Rift

Seismic-reflection data were acquired in the mid 1980s along several lines across eastern Lake Superior by industry and the Great Lakes International Multidisciplinary Program on Crustal Evolution (GLIMPCE) (Fig. 1). The lines form part of a larger network of crossing lines over the entire lake, which can be used to develop three-dimensional geologic models of the Mesoproterozoic Midcontinent Rift that lies below. To better interpret these lines, we developed velocity models to convert seismic reflections versus two-way travel time (TWTT) to reflections versus depth. In addition, the velocity models themselves provide insights into the structure of the Midcontinent Rift by recognizing common velocity ranges for certain rock types (Grauch, 2023).

eastern Lake Superior↗

USGS2026

Revisiting the utility of regional-scale, high-quality geophysical data in mineral exploration - A case study featuring the Mammoth Magnetic Anomaly, Pinal County, Arizona

Regional aeromagnetic surveys passively measure the total magnetic intensity (TMI) and are a foundational tool used in mineral exploration (Airo, 2015). With the increased global demand and the number of critical mineral resources required for manufacturing high-tech devices, developing high-quality, regional-scale geophysical surveys could aid critical mineral exploration efforts and geologic mapping. In 2019, the U. S. Geological Survey launched the Earth Mapping Resources Initiative (Earth MRI) to modernize the geologic and geophysical mapping of regions that have the potential to contain critical mineral resources within the United States. In support of planning Earth MRI geophysical surveys, Drenth and Grauch (2019) defined five aeromagnetic data quality rankings (rank 1 through rank 5) applying them to the airborne geophysical survey inventory of the United States (Johnson et al., 2021). Rank 1 aeromagnetic surveys are of the highest quality, meeting modern standards and allowing best practices for qualitative and quantitative interpretation; whereas rank 5 aeromagnetic surveys are of the lowest quality, being useful only for qualitative interpretation of broad features. Through the Earth MRI effort, 48 high-quality, regional-scale rank 1 and 2 airborne magnetic and radiometric geophysical surveys have been planned, collected, or publicly release through May 2025 (U. S. Geological Survey, 2025). Here, a portion of a rank 1 Earth MRI aeromagnetic survey in southeast Arizona is presented and compared to a legacy rank 5 aeromagnetic survey over the Mammoth Magnetic Anomaly (MMA), demonstrating how modern, high-quality aeromagnetic data improves our view of crustal geology, aiding mineral exploration.

Arizona↗

USGS2026

Early Miocene volcanic rocks and associated tectonics, Lava Hills and southern Bristol Mountains, California

Volcanic rocks of latest Oligocene to early Miocene age form an east-west belt across part of the central eastern Mojave Desert from the Whipple Mountains on the east to the Rosamond Hills on the west. We term this the central belt because it is separated from northern and southern belts by swaths with no volcanic rocks. Limited geochronologic data indicate that much of the belt is latest Oligocene and early Miocene in age, about 24 to 19 Ma, a finding that is consistent with these rocks being overlain by the 18.8 Ma Peach Spring Tuff in many places. We describe Miocene geology in a central area of the belt, in the Lava Hills, southern Bristol Mountains, and southern Old Dad Mountains. Sedimentary basins formed coeval with early andesite to rhyolite volcanism, progressing from fluvial and lacustrine tuffaceous sandstone to volcanic lavas, tuffs, and breccias, indicating that early basins formed proximal to volcanic edifices. Higher strata are fluvial and lacustrine with lavas punctuating the sequence. Although basins may partly have been formed within topographic lows bounded by volcanic domes, plateaus, and stratovolcanoes, consistent stratigraphic sections over wide areas indicate that tectonic basin evolution affected broad areas. The volcanic section is capped by local basalt flows and the regional Peach Spring Tuff. Limited data on normal faults support interpretations of early extensional basin development caused by northeast-southwest oriented stretching. Later extension caused stratal rotations, tilting early deposits down to the southwest. This tilted and subsequently beveled basin architecture was overlain by the youngest volcanic deposits, primarily rhyolite and basalt. The Peach Spring Tuff, 18.8 Ma, lies within this upper unit. Similar stratigraphic and structural relations are exposed in the nearby Marble Mountains and Van Winkle Mountain sections, reinforcing that a broad area underwent similar volcanism and tectonism. In our study area the upper unit is only very gently tilted except near dextral strike-slip faults of the eastern California shear zone. These late Miocene to Recent faults are represented as four main fault zones spaced about 5 km apart, representing distributed shear north of the Bristol Lake basin.

California↗