<![CDATA[The increasing volume of municipal solid waste demands low-cost, real-time sorting solutions to improve recycling efficiency and reduce landfill burden. Objective: This study develops and evaluates a low-cost, real-time computer vision system to classify plastic waste and organic leaf waste for automated sorting. Methodology: The system uses a standard RGB camera (640×480, 30 fps) and OpenCV-based processing, including Gaussian blurring, HSV color-space conversion, morphological operations, contour detection, and geometric feature extraction (circularity, solidity, aspect ratio, and extent). Classification is performed using a hierarchical rule-based logic that combines HSV color masks with a proposed overlap ratio to quantify the spatial correspondence between object contours and leaf-color regions. Findings: Experimental testing under controlled illumination (500–1000 lux) achieved 89% overall accuracy with an average processing time of 45 ms/frame and an operational throughput of approximately 7 objects/min. The system correctly classified 8 plastic items and 7 leaf samples in the initial test set. Implications: The proposed approach supports practical deployment in small-scale or resource-constrained waste management facilities by enabling real-time sorting without large, labeled datasets or GPU hardware. Originality: This work introduces an interpretable hybrid decision framework that integrates a mask-based overlap ratio with multiple geometric shape descriptors, improving discrimination between plastic and leaf waste while maintaining computational efficiency.]]>
Copyrights © 2025
