Coastal pollution analysis for environmental health and ecological safety using deep learning technique. (May 2023)

• Record Type: Journal Article
• Authors: Sathish, T.; Maheswari, S. Uma; Balaji, V.; Nirupama, P.; Panchal, Hitesh; Li, Zhixiong; Tlili, Iskander
• Is Part Of: Advances in engineering software. Volume 179(2023)
• Publication Date: 2023-05
• Subjects: Coastal pollution analysis -- Environmental health -- Ecological safety -- Anisotropic diffusion -- Global histogram equalization -- Deep CNN -- Reptile search optimization; ANN Artificial Neural Network -- BN Bayesian Network -- CNN Convolutional Neural Network -- DCNN Deep Convolutional Neural Network -- FC Fully Connected -- KNN K-Nearest Neighbour -- LSTM Long Short-Term Memory -- MLP Multi-Layer Perceptron -- NB Naive Bayes -- PM Particulate Matter -- RF Random Forest -- RNN Recurrent Neural Network; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2023.103441
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Coastal pollution analysis using CNN for environmental health and ecological safety. Image pre-processing to process the input data related to coastal pollution. Anisotropic diffusion and global histogram equalization are utilized in preprocess. Anisotropic for noise removal and histogram equalization to enhance quality. DCNN is to detect coastal pollution for environmental health and ecological safety. Abstract: Environmental degradation and loss of biodiversity occur widely in marine and coastal regions. The coastal ecosystems have diverse components, including mammals, invertebrates, and plants, which are now the most densely populated zones. Because of its localized interface, it is highly vulnerable to anthropogenic pollutants such as plastic debris, metal debris, volatile methyl siloxanes, and oil spills. Conventional algorithms related to coastal pollution classification face low accuracy and time consumption issues. A deep learning-based Deep Convolutional Neural Network (DCNN) is developed to overcome these issues. Initially, coastal images are collected and pre-processed using anisotropic diffusion and global histogram equalization for further processing. Anisotropic diffusion is utilized to remove the unwanted noise present in the image. Global histogram equalization is utilized to enhance the contrast level of the image. In sequence with pre-processing phase, the processed image is classified using deep learning-based DCNN to detect and analyze coastal pollution in coastal areas. In order to optimize, the weights present in the DCNN are optimally selected by using a reptile search algorithm to improve the classification performance. According to the experimental study, the proposed approach achieves 97.2% correctness, 2.8% error, 96% precision, 95.4% recall, 96% specificity, and 96.5% f1-score. Therefore, the developed approach attains better performance compared to other existing approaches. This prediction model helps detect coastal pollution in coastal areas to protect ecological safety and environmental health.