Computational Intelligence in Remote Sensing

With the development of Earth observation techniques, vast amounts of remote sensing data with a high spectral–spatial–temporal resolution are captured all the time, and remote sensing data processing and analysis have been successfully used in numerous fields, including geography, environmental monitoring, land survey, disaster management, mineral exploration and more [...]


Introduction
With the development of Earth observation techniques, vast amounts of remote sensing data with a high spectral-spatial-temporal resolution are captured all the time, and remote sensing data processing and analysis have been successfully used in numerous fields, including geography, environmental monitoring, land survey, disaster management, mineral exploration and more.For the processing, analysis and application of remote sensing data, there are many challenges, such as the vast amount of data, complex data structures, small labeled samples and nonconvex optimization.In recent years, the convergence of computational intelligence (CI) and remote sensing has ushered in a new era of possibilities for understanding and harnessing the wealth of information that Earth observation satellites provide.Computational intelligence methods, such as deep neural networks, evolutionary optimization and swarm intelligence, have demonstrated remarkable capabilities in unveiling intricate patterns within satellite images, time series data and multispectral/hyperspectral information.In the future, CI will produce effective solutions to the challenges in remote sensing.

Recent Research and Progress
This Topic series aims to highlight the latest research and advances in the application of computational intelligence in the field of remote sensing.In total, this Topic series contains 12 papers written by research experts on topics of interest.Based on the synthesis of these latest achievements, they can be categorized into four sections: computational intelligence methods in hyperspectral remote sensing images; object detection techniques in remote sensing images; deep learning approaches in remote sensing image classification and intelligent optimization and control in satellite image applications.

Computational Intelligence Methods in Hyperspectral Remote Sensing Images
This section consists of three papers.The first paper is written by A.C.P. Silva, K.T.Z.Coimbra, L.W.R. Filho, G. Pessin and R.E.Correa-Pabón.They mainly explore the possibility of applying machine learning models to monitor the quality of iron ore [1].The second paper, written by W. Shuai, F. Jiang, H. Zheng and J. Li, mainly proposes a new method with high processing efficiency for change detection in remote sensing images, called MSGATN [2].The last work studies SAR image segmentation based on fuzzy c-means and is by J. Zhu, F. Wang and H. You.Experiments show that the framework can achieve more than 97% segmentation accuracy [3].

Object Detection Techniques in Remote Sensing Images
The following three papers mainly utilize deep learning techniques to solve practical problems in the field of remote sensing image object detection.The first paper, by R. Chen and S. Liu et al., proposes an effective infrared object detection method based on source model guidance [4].They show two explicit examples based on Cen-terNet and YOLOv3, respectively, and experimentally demonstrate that the method can achieve powerful performance with limited samples.The second paper, by L. Yu and X. Zhou et al., proposes a method for boundary-aware salient object detection in optical remote sensing images [5]

Deep Learning Approaches in Remote Sensing Image Classification
This section includes three papers.The first paper is authored by H. Toriya and A. Dewan et al., who primarily explore the key point matching problem in image features.They propose using a deep neural network (DNN) to construct an image translator and introduce a new edge enhancement filter methodology within the conditional generative adversarial network (cGAN) structure to tackle this issue [7].The second paper, written by Z. Wei and Z. Zhang, describes a network built on multi-level strip pooling and a feature enhancement module (MSPFE-Net).Here, deep learning is effectively applied to address the challenge of road extraction [8].In the third paper, L. Zeng and Y. Huo et al. develop the high-quality seed instance mining (HSIM) module, alongside the dynamic pseudo-instance label assignment (DPILA), to address the issue of weakly supervised detection in remote sensing images [9].

Intelligent Optimization and Control in Satellite Image Applications
This section includes three state-of-the-art papers for reference focusing on different research directions in satellite images.The first paper is authored by T. Zheng, Y. Dai, C. Xue and L. Zhou.They propose a method for solving near-lossless hyperspectral data compression using recursive least squares.They use the linear combination of previous pixels to predict the target pixel values while using a recursive least squares filter to iteratively update the weight matrix for prediction, which effectively removes spatial and spectral redundancy information [10].The second paper is written by N. Andrijević, V. Urošević, B. Arsić, D. Herceg and B. Savić.This paper designs a time prediction model for bee influx and outflow in a bee colony ecosystem with a large number of sensors by simulating the correlation between the environment and bee colony activity to simulate the bee colony ecosystem [11].L. Li, D. Yin, Q. Li, Q. Zhang and Z. Mao propose a verification method for ultraviolet imagers using the seeker optimization algorithm.This method can effectively use ultraviolet imagers to conduct authenticity check studies on ocean surface radiation data [12].

Discussion
The papers provide an exchange platform for researchers in the field of remote sensing images, covering topics such as hyperspectral remote sensing image processing, remote sensing image classification, segmentation, object detection and intelligent optimization and control in satellite image applications.These themes represent a series of key issues in the field of remote sensing images.The research papers in this journal not only delve into these issues, but also propose new methods and ideas, providing strong support for future research directions.
In this issue of the journal, we have seen a series of important developments in the field of hyperspectral remote sensing image processing.Researchers have utilized the rich information of hyperspectral data to not only improve the performance of segmentation, but also provide new tools for application fields such as resource management and environmental monitoring.In addition, remote sensing image classification, segmentation and object detection have always been research hotspots.Research in this journal shows that deep learning technology has made significant progress in the application of these tasks.
The papers in this research Topic showcase the innovative and influential contributions of researchers in this field.Researchers have not only delved into various issues, but also proposed many new methods and technologies, demonstrating the potential of computational intelligence in advancing our understanding of remote sensing images and providing strong support for future research directions.In the future, we can look forward to more interdisciplinary cooperation, combining remote sensing image research with application fields such as environmental science, agriculture and urban planning to solve complex real-world problems.We encourage readers to further explore the cutting-edge research and novel applications presented in these papers to provide new impetus for scientific and technological innovation.
. The method uses a graph convolutional network-based feature extraction module and a boundary-aware attention-based module to improve the accuracy and robustness of boundary-aware salient object detection.The third paper, by F. Zhou and H. Deng et al., studies deep learning-based aircraft detection [6].The paper proposes an enhanced YOLOv5 model in which a ConvNext-based feature extraction module and a Transformer-based feature fusion module are used to improve the detection performance.