An Enhanced YOLOv8n-Seg Algorithm for Concrete Flowability Measuring

The slump spread of concrete is a critical indicator for evaluating how easily concrete flows without external force except gravity, directly influencing the workability and final quality of construction. Therefore, accurately and rapidly measuring the slump spread of concrete is of significant research issue. Precise instance segmentation of concrete is a key prerequisite for measuring the slump spread using image-based methods. In this paper, we propose an enhanced YOLOV8n-seg algorithm to measure the slump spread of concrete. First, MobileViT module is employed to replace C2f modules in the backbone, substituting local modeling in the original model's convolutions with global modeling so as to encode more effective information. Second, Squeeze-and-Excitation attention mechanism is embedded in the backbone to enhance the feature maps, thereby improving detection accuracy. Finally, WIoU loss function is introduced to accelerate model convergence. Extensive experimental validation shows that the optimized algorithm improves the mean average precision by 2.2 % compared to the original YOLOv8n-seg. The model contains only 3.7 M parameters, enabling high-precision deployment while maintaining low computational requirements. It is appropriate for the intelligent measurement of concrete slump spread.