Lintao Han , Hengyi Lv , Chengshan Han , Yuchen Zhao , Qing Han , Hailong Liu
{"title":"Atmospheric scattering model and dark channel prior constraint network for environmental monitoring under hazy conditions","authors":"Lintao Han , Hengyi Lv , Chengshan Han , Yuchen Zhao , Qing Han , Hailong Liu","doi":"10.1016/j.jes.2024.04.037","DOIUrl":null,"url":null,"abstract":"<div><p>Environmental monitoring systems based on remote sensing technology have a wider monitoring range and longer timeliness, which makes them widely used in the detection and management of pollution sources. However, haze weather conditions degrade image quality and reduce the precision of environmental monitoring systems. To address this problem, this research proposes a remote sensing image dehazing method based on the atmospheric scattering model and a dark channel prior constrained network. The method consists of a dehazing network, a dark channel information injection network (DCIIN), and a transmission map network. Within the dehazing network, the branch fusion module optimizes feature weights to enhance the dehazing effect. By leveraging dark channel information, the DCIIN enables high-quality estimation of the atmospheric veil. To ensure the output of the deep learning model aligns with physical laws, we reconstruct the haze image using the prediction results from the three networks. Subsequently, we apply the traditional loss function and dark channel loss function between the reconstructed haze image and the original haze image. This approach enhances interpretability and reliability while maintaining adherence to physical principles. Furthermore, the network is trained on a synthesized non-homogeneous haze remote sensing dataset using dark channel information from cloud maps. The experimental results show that the proposed network can achieve better image dehazing on both synthetic and real remote sensing images with non-homogeneous haze distribution. This research provides a new idea for solving the problem of decreased accuracy of environmental monitoring systems under haze weather conditions and has strong practicability.</p></div>","PeriodicalId":15788,"journal":{"name":"Journal of Environmental Sciences-china","volume":null,"pages":null},"PeriodicalIF":5.9000,"publicationDate":"2024-04-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Environmental Sciences-china","FirstCategoryId":"93","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1001074224002249","RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENVIRONMENTAL SCIENCES","Score":null,"Total":0}
引用次数: 0
Abstract
Environmental monitoring systems based on remote sensing technology have a wider monitoring range and longer timeliness, which makes them widely used in the detection and management of pollution sources. However, haze weather conditions degrade image quality and reduce the precision of environmental monitoring systems. To address this problem, this research proposes a remote sensing image dehazing method based on the atmospheric scattering model and a dark channel prior constrained network. The method consists of a dehazing network, a dark channel information injection network (DCIIN), and a transmission map network. Within the dehazing network, the branch fusion module optimizes feature weights to enhance the dehazing effect. By leveraging dark channel information, the DCIIN enables high-quality estimation of the atmospheric veil. To ensure the output of the deep learning model aligns with physical laws, we reconstruct the haze image using the prediction results from the three networks. Subsequently, we apply the traditional loss function and dark channel loss function between the reconstructed haze image and the original haze image. This approach enhances interpretability and reliability while maintaining adherence to physical principles. Furthermore, the network is trained on a synthesized non-homogeneous haze remote sensing dataset using dark channel information from cloud maps. The experimental results show that the proposed network can achieve better image dehazing on both synthetic and real remote sensing images with non-homogeneous haze distribution. This research provides a new idea for solving the problem of decreased accuracy of environmental monitoring systems under haze weather conditions and has strong practicability.
期刊介绍:
The Journal of Environmental Sciences is an international journal started in 1989. The journal is devoted to publish original, peer-reviewed research papers on main aspects of environmental sciences, such as environmental chemistry, environmental biology, ecology, geosciences and environmental physics. Appropriate subjects include basic and applied research on atmospheric, terrestrial and aquatic environments, pollution control and abatement technology, conservation of natural resources, environmental health and toxicology. Announcements of international environmental science meetings and other recent information are also included.