DexiNed-based Aluminum Alloy Grain Boundary Detection Algorithm

Grain size measurement is an important link in metallographic analysis, which determines the mechanical properties of aluminum alloy materials such as strength, ductility, plasticity, and fatigue resistance. Accurate and efficient detection of grain boundaries is essential for grain size measurement. This paper proposes an edge detection algorithm based on DexiNed to detect the grain boundaries of aluminum alloy metallographic images. First we build an aluminum alloy grain boundary detection data set to train DexiNed. Then we use the trained DexiNed to perform grain boundary detection on the metallographic image of the aluminum alloy, and report the detection performance on the test set. In addition, using the detected grain boundary images, we further explored the distribution of grain areas in the images. Through comparison experiments with traditional edge detection algorithms, we find that DexiNed can accurately and efficiently detect the grain boundaries of aluminum alloys.