Automatic indentation measurement method based on full convolutional neural network

Abstract

The invention discloses an automatic indentation measurement method based on a full convolutional neural network, and the method comprises the following steps: 1, carrying out the preprocessing of an indentation image, and obtaining a binary image of the indentation image; 2, searching a maximum connected domain in the binary image, and determining an external frame of the maximum connected domain; expanding the external frame, and cutting an expanded external frame area on the binary image; and step 3, inputting the cut image into the trained full convolutional neural network, and outputting a binarized indentation pixel mask picture, thereby obtaining accurate indentation shape and position information on the input picture. The invention is high in robustness, and the specific shape of the indentation can be extracted.

Description

technical field [0001] The invention relates to a hardness indentation measurement method, in particular to an automatic indentation measurement method based on a fully convolutional neural network. Background technique [0002] In the Brinell and Vickers hardness tests, in order to detect the hardness of the material to be tested, it is necessary to use the corresponding indenter according to the national standard, apply a specified pressure on the surface of the material to be tested, and then calculate the indentation by measuring the diagonal length of the indentation. Measure the hardness of the material. In most measurements, the diagonal length of the indentation on the surface of the material being tested is measured in microns. The method of manually adjusting the scale plate through the microscope to align the edge of the indentation is more complicated and difficult to eliminate measurement errors. At the same time, in the process of measuring the same hardness ...

AI Technical Summary

Problems solved by technology

However, the existing automatic indentation calculation methods often use mathematical or image analysis methods, which are often difficult to work on some more complex indented surfaces

On some indentation images that are deformed due to the characteristics of the material itself at some edges and sharp corners, large errors often occur, and the robustness of the algorithm is poor

At the same time, the existing technical methods can only obtain the indentation diagonal length data for the indentation image, and cannot extract the specific shape of the indentation in the image, which is not convenient for later analysis of material rupture and collapse

Images