Association analysis between floc image information and coagulation efficiency based on partial least squares

Floc image analysis shows promise for monitoring coagulation in water treatment, yet a systematic understanding of the relationship between image features and coagulation efficiency remains lacking. In this study, floc images were captured using a non-invasive image acquisition system, and multiple morphological and textural features were extracted via Python-OpenCV. The associations among floc images, settling behaviour, and effluent turbidity were systematically investigated using partial least squares regression. The results demonstrated that image features can effectively characterise floc population behaviour. Floc number removal was jointly influenced by the three-dimensional fractal dimension (D₃), the proportion of 100-150 μm particles, and textural correlation. The fractal dimension and textural correlation reflect floc compactness and internal uniformity, while the small-particle proportion indicates flocculation completeness. Effluent turbidity was primarily governed by the initial floc number, D₃ and small-particle proportion, and showed a strong positive correlation with the residual floc count. This study provides a theoretical basis for optimising the operational management of drinking water treatment processes, thereby supporting SDG 6 (Clean Water and Sanitation) through improved coagulation monitoring and process control.