A Method of Establishing Neural Network Model for Classification of Scrap Steel

Abstract

The invention discloses a method for establishing a neural network model for grade classification of scrap steel. The model is used for grade classification detection of scrap steel collection and storage. Process to remove invalid watermarks, improve image contrast, perform image data feature extraction on image data, and perform convolutional neural network learning on extracted image data features of different levels to form a hierarchical division neural network model with hierarchical classification output; the image data features Extraction is the extraction of the set of convolutional neural network convolution calculations for the pixel matrix data of the image screen, including: the calculation of the objects in the image composed of multiple lines of convolutional layers or convolutional layers plus pooling layer calculations output by the set The extraction of color, edge features and texture features, as well as the extraction of the associated features between object edges and textures in the image.

Description

technical field [0001] The invention relates to a method for establishing a neural network model for classifying scrap steel. Background technique [0002] Steel mills collect and store a large amount of steel scrap every year. Except for waste machinery and equipment, it is mainly a large amount of scrap steel fed in by cars. Shredded steel includes pipes, blocks and plates of various shapes, with different sizes and shapes. , which is also mixed with other impurities. The imported shredded steel needs to be graded according to the thickness and paid according to the grade. Therefore, the grade approval is very important. At present, the quality inspection process is to manually receive the material, and manually fill in the receipt notice In the process of quality inspection, quality inspectors are required to climb onto the scrap truck, manually measure the size of scrap steel, manually inspect the quality of scrap steel, and judge the grade of scrap steel, and then manua...

AI Technical Summary

Problems solved by technology

This approach has the following disadvantages: 1) Scrap steel has different shapes and piles up messily, and it is impossible to check the overall scrap steel situation in detail

2) Since the inspectors need to get on and off the car frequently, there will be potential safety hazards

3) The production pace of the steel mill is fast, and the quality inspectors frequently get on and off the car for inspection, which affects the unloading progress of the crane and the production process

4) It is necessary to set up a special person to do this work, and there is a labor cost problem

5) Manual transcription and inspection are all based on visual inspection. The standards vary from person to person and are not uniform. There are misjudgments and misjudgments, and the accuracy rate is not high

6) There is a risk of private communication between suppliers and quality inspectors

[0003] As an image recognition technology, convolutional neural network has been widely used in face recognition. By establishing a recognition model and inputting face images into the model for recognition, can this method also be used to classify broken steel grades? However, in The process of establishing a face recognition model is usually learned by extracting the edge features of the face, and the crushed steel materials shipped as cars are overlapped and extruded together, and some small crushed steel materials are mixed together It is impossible to distinguish its shape when it is laid in the carriage, so it is impossible to build a model by using the conventional method of extracting edge features

Images