Research on Online Detection and Segmentation Method of Coarse Aggregate Based on Deep Learning
When using the image method to detect aggregate particle size,image segmentation quality is an essential factor affecting the detection of aggregate particle size.At present,aggregate image segmentation has developed from the traditional wa-tershed and threshold segmentation algorithms to the instance segmentation algorithms to segment stacked coarse aggregates.Ai-ming at the problem of many undivided aggregates in the ISTR(end-to-end instance segmentation with transformers)network model,an improved algorithm and an evaluation index were proposed for evaluating the network model.Finally,a comparative ex-periment was conducted on the network model before and after optimization.The experimental results show that compared with the original network model,the MIoU(Mean Intersection over Union)of the optimized algorithm has increased by 3.4%,reaching 82.6%,the proportion of unsegmented aggregates has decreased by 8.2%,reaching 9.4%,the detection and segmentation ability improves significantly,which proves that the feasibility and effectiveness of the proposed method in aggregate detection and seg-mentation tasks.
NETL
NSTL
万方数据
