Method for automatically identifying number of steel billet

Abstract

The invention discloses a method for automatically identifying the number of a steel billet in the technical field of character identification, which comprises the following steps: (1) preprocessing an image; (2) intercepting the area of the number of the steel billet and carrying out perspective collineation; (3) performing gray scale processing on the image; (4) processing by binarization; (5) cutting a numerical symbol; (6) standardizing a character; (7) calculating the degree of similarity and identifying the character; and (8) judging whether the number of the steel billet is mistaken or not. The invention can realize the overall process of the automatic identification and control of the number of the steel billet in an on-line way and has the following obvious technical advantages: (1) the cutting of the character is accurate and effective, and the cutting error rate is smaller than 0.1%; and (2) the identification of the character is accurate and quick, the identifying precision ratio reaches 100%, and the operation period of the identification of the number of the entire steel billet does not exceed 100ms.

Description

technical field [0001] The invention relates to a character recognition method in the detection technology field, in particular to a billet number automatic recognition method. Background technique [0002] In the process of steel billet production in steel mills, symbols such as specific number sequences will be sprayed on one side of the billet to indicate a specific batch, and different production procedures are stipulated according to different batches, but due to the inevitable Filling errors can occur, causing billets from other batches to mix into the production line, resulting in product quality defects. Existing iron and steel plants generally monitor and check billet numbers manually. However, the manual method is not conducive to the automatic production of the assembly line. It is not only slow to rely on humans to judge whether the steel billets belong to the same batch, but also cause judgment errors due to human eye fatigue. In order to improve the accuracy,...

AI Technical Summary

Problems solved by technology

[0004] (1) The pre-processing of the image is too complicated, so it takes up a large amount of computer computing time, which will inevitably affect the recognition computing speed;

[0005] (2) It is necessary to combine the fixed positional relationship of the license plate character sequence to complete the segmentation of the license plate characters, which will inevitably increase the error rate of character segmentation;

[0006] (3) The BP neural network algorithm includes a multi-layer perceptron, which will inevitably increase the time-consuming calculation and cannot meet the technical requirements of fast recognition

[0008] (1) Since the actual two-dimensional image pattern is usually stored in the computer in a discrete form, all calculations must be performed through difference equations. The important point is to ensure that the dimension of the observation data space must be large enough, otherwise the binary There will be many pixels in the image that do not produce a central projection, resulting in insufficient statistical information;

[0009] (2) The test result of the example proves that the correct rate of character recognition of this method only reaches 93.4%, so there is still a big gap from the practical technology

Images