| In recent years,due to the development of industrialization and urbanization,the occurrence of sudden pollution incidents in the water environment has led to changes in the color of water bodies resulting in abnormal water color.Water color is an important indicator of water quality,and remote sensing is a crucial method for detecting water color anomalies caused by pollution.Traditional remote sensing methods for water color anomalies involve analyzing water absorption and backscatter coefficients and establishing spectral indices to identify and extract water color.However,pollutants causing abnormal color in water bodies are complex mixtures of multiple pollutants,lacking uniform spectral reflection peaks and absorption valleys,and without fixed wavelengths.As a result,it is difficult to identify and extract them using a unified spectral feature index.This study utilized domestic GF2-PMS data as the data source,focusing on the water area outside Meidong Dazha in Beilun District,Ningbo City,and constructed a discriminant function for normal and abnormal water color using the spectral characteristics of remote sensing images.Based on GF2-PMS image data,this study set a threshold for determining water color anomalies in the water area outside the Meidong Gate.Then,using the GF2-PMS image from June 9,2022 as the extracted image,this study compared and analyzed the extraction results based on single image features and the water color anomaly detection threshold set based on the multi-temporal image features from 2018 to 2022.The results were compared and analyzed with the on-site verification sampling results and the extraction results based on different band combinations.The research results indicate that:(1)By utilizing the spatial distribution information of water color anomalies extracted from a single remote sensing image on June 9,2022,not only was a relatively comprehensive area of water color anomalies detected,but also water bodies with high NDWI gray values and suspended sediment water bodies were preserved.The accuracy of the water color anomaly extraction results was improved through setting the judgement function twice.(2)Compared to the feature extraction process based on single-scene image features that require setting the discrimination function again to extract the water color abnormality,the spatial distribution information of water color abnormality extracted based on multi-temporal image features only needs to be extracted once,achieving similar results to those obtained by extracting single-scene image features again.(3)After comparing and analyzing the results of water color anomaly detection using single and multi-scene images,it was concluded that the judgment function based on long-term,multi-scene image features is more accurate and faster in extraction speed than the judgment function based on single scene image features.(4)This study applied mean filtering on NDWI grayscale images before constructing a discriminant function to distinguish between normal and abnormal water color,and to extract water color anomalies present in the study area.The water color anomaly extraction results using different band combinations showed a reduction in small patches,in comparison to those based on single scene and multi-temporal image features.Thus,it was concluded that mean filtering is effective in preventing the refinement of the extraction results.(5)This article employs the spectral properties of an image to detect water color anomalies resulting from unidentified pollutant types,enabling the acquisition of spatially-distributed information.It addresses the challenge of recognizing water color anomalies caused by pollutants lacking typical spectral features or those composed of mixed pollutants with several typical spectral features.This approach provides vital information and technical support for the management of water environment pollution,facilitating prompt detection and remediation of water pollution,and promoting water environment and ecological security. |
PDF Full Text Request