Detection of Lead Contamination in Surface Water using Remote Sensing and Computer Vision

Abstract

Water is the Essene of life, Lead contamination in surface water bodies is a cause of various diseases. Detection of lead contamination using traditional approaches is in efficient and time consuming. This article introduces a remote sensing and computer vision based approach for detecting the Lead contamination in various locations of In dus River. This approach highlights the water bodies areas where lead contamination exists using the satellite imagery. It also performs the comparative analysis of different deep learning approaches for detection of Lead contamination. Transfer Learning based VGG16UNET out performed other models such as DeepLabv3. VGG16UNET model achieved mIoU of 0.70. We also used machine learning approaches for detection of lead contamination just from water quality parameters. The results showed that Random Forest based approach achieved the highest accuracy of 73% followed by SVM at 71%. This approach can be used for real-time monitoring of water bodies and it covers a waste area just from single image. It can help the policy makers in identifying the areas where water is contaminated.