A Meter Inspection Method Based on Machine Vision

Abstract

The invention discloses an instrument detection method based on machine vision. The instrument detection specifically includes: (1) for any input image, first pass through the cascaded Adaboost coarse detector to detect the target instrument candidate area; (2) for the step (1) For each candidate area obtained, use a double-cascade parametric regressor to regress the affine transformation matrix between the standard image and the image to be recognized, and then affine transform the image to be recognized to the pose of the standard image to achieve instrumentation Normalization of attitude; (3) The image after the attitude normalization obtained in step (2) is confirmed through the posterior verifier to confirm whether there is a target instrument, and finally the detection result is output. The instrument detection method proposed by the invention solves the problem of posture and scale in the instrument detection, has the advantages of high positive detection rate and low false detection rate, and at the same time, the invention has a fast processing speed and realizes the real-time detection function of the instrument.

Description

technical field [0001] The invention belongs to the field of digital image processing and instrument detection, and in particular relates to a machine vision-based instrument detection method. Background technique [0002] In industrial fields such as chemical plants, substations, and oil refineries, a large number of on-site indicating instruments are installed. These instruments do not have remote transmission functions and require on-site readings to monitor the operating conditions of industrial sites. Among them, instrument inspection is an important part of it. It is time-consuming and labor-intensive to detect on-site instruments by manual inspection. Therefore, the automatic instrument inspection technology based on machine vision has broad application prospects. [0003] In the existing research, common instrument detection methods include template matching, Hough transform, feature point matching and other methods. Dai Yawen proposed a method based on multi-featur...

AI Technical Summary

Problems solved by technology

[0004] The above instrument detection method uses template matching to have extremely high requirements on the attitude of the instrument in the image, and is sensitive to illumination and image noise

The method of using Hough transform relies on the edge detection algorithm. Under complex industrial sites and different imaging conditions, the circular instrument panel is not necessarily strictly circular, and the edges of the circular panel and the pointer may not be able to be detected, so Once the Hough circle detection or straight line detection fails, the correct detection result cannot be obtained

Using the method of feature point matching, in the actual use process, there is a problem of partial occlusion of the dial, and even some dial panel images have relatively few texture features, and features such as ORB and SIFT cannot be extracted by themselves, resulting in correctly matched feature points. Very few, and it is also possible that due to the interference of the complex image background, the corresponding features are extracted on the complex background, resulting in a large number of mismatches. Due to the above reasons, the method of feature point matching is also relatively easy to fail.

[0005] To sum up, the existing research instrument detection methods are very sensitive to the working conditions of instrument images under different lighting conditions, attitudes, scales, partial occlusions, blurred images, etc., and it is difficult to meet the actual use requirements.

Images