A low-cost multi-channel real-time digital instrument panel visual identification method and system

Abstract

The invention provides a low-cost multi-channel real-time digital instrument panel visual identification method and system. Includes digital tube area configuration and digital tube decimal point information configuration S2: performing sub-area ROI preprocessing; S3:performing Sub-region identification; S4: performing Sub-region post-processing; 5, train, updating and teste that digital recognition model; S6: Publishing completion model. The invention solves the identification difficulties of the instrument panel one by one so as to be better applied to industrial production monitoring.

Description

technical field [0001] The invention belongs to the field of industrial production monitoring, and in particular relates to a low-cost multi-channel real-time digital instrument panel visual recognition method and system. Background technique [0002] The task of identifying various instrument representations is different from the traditional optical character recognition (OCR) problem, and is closer to the text recognition problem of natural scenes. The research in this area can be divided into detection and recognition as a whole. Different from the OCR problem of documents, the detection of numbers in various instrument representations is a very difficult thing itself, and many scholars have made a lot of attempts. [0003] Before the emergence of deep learning methods, methods based on traditional handcraft features, including connected regions and HOG-based detection frame descriptions, were more mainstream; Regions) to obtain character candidates, and regard these cha...

AI Technical Summary

Problems solved by technology

[0008] 1. Most of the instrument display tasks are displayed in the form of digital diodes. Digital diodes have a fixed refresh rate, which will cause more pictures to show that the diodes that should have been lit should be refreshed when shooting a video continuously. It becomes darker at this frequency, which increases the difficulty of identifying stability;

[0009] 2. The noise of instruments and meters in actual shooting is very large, and the difference between the front and back of each frame of image is very high, such as figure 1 As shown, although the human eye cannot detect it, it is still very difficult for the computer to ensure that the value in the dashboard can be stably recognized in a noisy environment;

[0010] 3. Real-time data update has speed requirements. In the actual production data recording, the frequency of digital changes is relatively fast. How to ensure high-speed and real-time identification of shooting data is still difficult for the current more comprehensive OCR algorithm;

[0011] 4. Resolution requirements under cost constraints, there is a large demand for the monitoring of industrial digital dashboards, but if a high-resolution camera is used, although the accuracy rate can be improved, the speed will decrease and the cost increase will not be acceptable for practical applications. Therefore, how to recognize numbers at low resolution is also an innovation of this algorithm compared with other OCR algorithms;

[0012] 5. In the actual production line, multiple dashboards need to be monitored at the same time, so it is inevitable that most of the dashboards are not completely facing each other in the shooting, and there are various distortions. How to ensure the recognition of the dashboard numbers under various distortions is the task one of the difficulties

Images