A Simulated Annealing Algorithm based on Variance Consensus for Logistics Cargo Sorting

Image segmentation plays a vital role in logistics for precise target extraction and cargo sorting. Traditional methods struggle with complex images, particularly those with uneven lighting. Our paper introduces a novel approach using a simulated annealing algorithm enhanced by inter-class variance consensus criteria. This technique calculates an optimal segmentation threshold for each pixel, enabling adaptive segmentation. The simulated annealing algorithm, guided by the variance criteria, effectively finds optimal solutions for each pixel and ensures quick convergence. We tested our method using the OSU Thermal Pedestrian Database, showing notable reductions in misclassification errors and processing time, crucial for logistics applications like cargo sorting. Our results demonstrate that this method has the potential to significantly improve image segmentation in logistics, boosting both accuracy and efficiency.