Abstract
In the field of mining engineering, ensuring the safe operation of mines is of utmost importance, and the stability of the backfill materials plays a pivotal role. This research comprehensively analyzes the strain field evolution and crack development in cemented paste backfill (CPB) specimens made from whole tailings under various backfill mix designs by using uniaxial compressive strength (UCS) testing, digital image correlation, and computer vision recognition (CVR) technology. The experimental outcomes reveal that the UCS of the CPB decreases with reductions in cement-to-tailings ratio, filling concentration, and curing age, while the rate of principal strain field evolution significantly accelerates. The developed computer vision recognition model (HSV-CVR), based on hue, saturation, and value color patterns, processes strain field data to quantify the proportions of various strain regions. By applying the first derivative of these proportions, the model enables early crack prediction. This approach overcomes the limitations and subjectivity of traditional artificial vision methods for crack identification, providing precise quantification of CPB strain evolution. The research enhances understanding of mining backfill materials behavior and provides a strong scientific basis for design, monitoring, and risk management, crucial for improving mining safety and efficiency.