Method for automatic quantitative statistical distribution characterization of dendrite structures in a full view field of metal materials

A method for automatic quantitative statistical distribution characterization of dendrite structures in a full view field of metal materials, comprising the following steps:

• (1) establishment of an object detection model based on deep learning
• performing metallographic sample preparation, polishing and chemical corrosion on a standard metal material sample with the same material as a metal material to be detected, so that the surface of the sample shows a clear and complete dendrite structure;
• acquiring dendrite structure feature maps of the standard metal material sample after the metallographic chemical corrosion by using a fully automatic metallographic microscope, and establishing a dendrite structure feature map data set;
• labeling images in the dendrite structure feature map data set by using dendrite centers as objects to obtain images containing marker boxes, with the dendrite centers being inside the marker boxes, wherein marked information comprises a category and regional information of the object, the category indicating whether the marker boxes contains the dendrite center, and the regional information comprising the position of the dendrite center;
• and generating an object detection sample library from the marked dendrite structure feature map data set;
• and
• establishing an object detection network, and training with the obtained object detection sample library to obtain an object detection model;
• (2) automatic acquisition of dendrite structure feature maps in a full view field of the metal material to be detected
• performing metallographic sample preparation and metallographic chemical corrosion on the metal material to be detected, by using the same metallographic sample preparation and metallographic chemical corrosion process as in step (1), and then automatically acquiring dendrite structure feature maps in a full view field of the surface of the metal material to be detected after the metallographic chemical corrosion, by using a fully automatic metallographic microscope;
• (3) automatic identification and rapid positioning of dendrite structures
• inputting all images to be detected acquired in step (2) into the object detection model established in step (1), to directly obtain dendrite structure feature maps labeled with the dendrite centers;
• (4) automatic stitching and fusion of the dendrite structure feature maps in a full view field
• identifying and stitching overlapping regions of neighboring ones of all the dendrite structure feature maps labeled with the dendrite centers obtained in step (3) by using an image registration algorithm that is based on mutual information, to obtain a dendrite structure feature map containing the marked dendrite centers in a full view field;
• and
• (5) Quantitative Distribution of the Number and Spacing of Dendrite Structures in a Full View Field within a Wide Range
• performing statistical distribution analysis on the number and spacing of dendrites in the dendrite structure feature map in a full view field obtained in step (4), to obtain quantitative statistical distribution data such as the position of each dendrite, the number of the dendrites and the spacing between two neighboring dendrites in a full view field within a wide range.