Principles and Processes of Remote Sensing Satellite Image Processing

What is remote sensing image processing?

Remote sensing images contain a wealth of information, which can only be effectively analyzed and extracted after digitization (sampling and quantization by the imaging system, and digital storage). Further processing of the image data, such as correcting graphic alignment coordinates and enhancing feature outlines, can greatly improve the accuracy of image processing and the efficiency of information extraction. This process can be called "remote sensing digital image processing."

As a fundamental and important component of the "Earth observation" process, remote sensing image processing occupies a crucial position in the mid-to-downstream of the satellite application industry chain, serving as a bridge between upstream and downstream sectors. It connects with satellite ground facilities at the front end and provides "ready" data services or tools for specific business applications in industries such as agriculture, forestry, meteorology, and natural resources at the back end.

Remote sensing image processing workflow

Data storage and management

Large-scale remote sensing data aggregation, management, storage, and distribution system

The raw information received by the ground station is photographically processed, transformed, and digitized into positive film or computer-compatible magnetic tape. The resulting photos are then bound into albums and cataloged for users to choose from.

Image preprocessing

Processing equipment is used to perform geometric and radiometric corrections on systematic errors in remote sensing images, such as geometric shape and position errors and image radiometric intensity information errors.

Geometric correction principle diagram

Geometric correction

During remote sensing imaging, due to the combined effects of factors such as deformation of photographic materials, lens distortion, atmospheric refraction, Earth curvature, Earth rotation, and topographic relief, the geometric position, shape, size, dimensions, orientation, and other characteristics of ground features in the original image are often inconsistent with the characteristics of the corresponding ground features. This inconsistency is called geometric deformation, also known as geometric distortion.

Geometric correction is performed to address the causes of geometric distortion. Before providing data to users, the ground receiving station uses conventional processing methods and simultaneously receives information on the operating attitude, sensor performance indicators, atmospheric conditions, and solar altitude angle from the images to correct and eliminate this geometric distortion through a series of mathematical models, ensuring accurate positioning.

Before geometric correction - After geometric correction

Radiation correction

It refers to the correction of systematic and random radiation distortion or aberration caused by external factors in data acquisition and transmission systems, and the process of eliminating or correcting image distortion caused by radiation errors.

Radiometric correction refers to the process of eliminating various distortions in image data that are attached to radiance. The processing station receives the raw data from the receiving station, reads it into the image processing system, and first decomposes the data to create separate files: raw remote sensing image data and telemetry auxiliary information data. Then, based on the remote sensing image radiometric error correction model derived from the radiative transfer equation, and with the support of the image processing system's hardware and software, system radiometric correction is performed.

In short, it involves removing sensor or atmospheric "noise" to more accurately represent ground conditions and improve image "fidelity," primarily by restoring missing data, removing haze, or preparing for mosaicking and change monitoring.

Before and after radiation correction

Image enhancement

Histogram before and after stretching

To enhance the readability of ground feature information contained in remote sensing images and make targets of interest more prominent, image enhancement processing is required. This mainly includes steps such as image contrast enhancement, color compositing, histogram transformation, density segmentation, grayscale inversion, inter-image operations, image fusion, image cropping, image stitching, and mosaicking with color balancing.

Information Extraction

The characteristics of target features in remote sensing images are the reflection of differences in electromagnetic radiation of these features on the remote sensing image. The process of identifying the type, properties, spatial location, shape, size, and other attributes of features based on the features in the remote sensing image is called remote sensing information extraction.

Currently, information extraction methods include visual interpretation and computer classification. Among them, visual interpretation is the most commonly used method.

• Visual interpretation

Also known as manual interpretation, it involves interpreting remote sensing images manually, sketching the extent of target features on the images, and extracting information.

• Image classification

It is the process of classifying all pixels in an image into several categories based on their properties, according to the spectral characteristics of ground objects, determining the discrimination function and corresponding discrimination criteria. The main methods are supervised classification and unsupervised classification.

As an important component of Earth observation and remote sensing industrialization, remote sensing data processing, located in the midstream and downstream of the industry, has also been impacted by the era of big data. It is responding to this trend and undergoing transformation, moving towards real-time, standardization, large-scale, and automation.

Today, remote sensing information processing methods and models are becoming increasingly scientific. Information models and technologies such as neural networks, wavelets, fractals, cognitive models, geoscientific expert knowledge, and the integration of image processing systems will greatly improve the accuracy and reliability of multi-source remote sensing technology fusion, classification, and extraction. The organic combination of statistical classification, fuzzy logic, expert knowledge, and neural network classification constitutes a composite classifier, significantly improving classification accuracy and the number of classes. The fusion and composite application of multiple platforms, multiple layers, multiple sensors, multiple temporal phases, multiple spectra, multiple angles, and multiple spatial resolutions is an important development direction for remote sensing technology. The development and application of uncertain remote sensing information models and artificial intelligence decision support systems also require further research.

Some content in this article is sourced from...

GIS Frontiers, Super-Powered Spacetime