MICROPLASTIC DETECTION IN BEACH SAND USING IMAGE PROCESSING AND MORPHOLOGICAL SEGMENTATION

Abstract

The presence of microplastic waste in coastal areas is an extreme ecological hazard due to its persistence and microscopic size, making it indistinguishable from natural particles, such as grains of sand. The paper presents a powerful image processing and deep learning environment for precisely identifying and categorizing microplastics in beach sand. The system utilizes a microplastic-specific dataset specifically designed for use in computer vision. Due to microplastic particle segmentation, the preprocessing phase utilizes binary image thresholding and morphology to segment the microplastic particles effectively. To enhance feature selection, a saliency attention mechanism is combined, enabling the model to focus on the most critical areas of the image. Data on amplitude and phase holography are merged to enhance the ability to differentiate microplastics from other materials with similar appearances. The output of improved features is fed into the Guided Convolutional Neural Network (AG-CNN) to make gradable predictions. Standard performance measures, accuracy 92.3%, precision 91.0%, recall 90.5%, and F1-score 90.7% are used to evaluate the proposed approach and show better detection rates and fewer false positives than traditional types of methods. Such a mixed solution promises the potential of a robust, automated track of microplastics that can aid in the coastal management of the world and help with the reduction of microplastic pollution within marine environments.