A vision-based automatic recognition method and system for instrument pointer readings

Abstract

The invention discloses a vision-based automatic recognition method and system for instrument pointer readings. The method is as follows: firstly collect a watch image, use a neural network to segment and extract the pointer and each area of ​​the watch, and then use edge detection to screen out the centroid of the scale, Fit the inner circle of the dial, determine the center of the dial, and then detect the center line of the pointer, take the center point of the dial as the reference point, take the center of mass of the reference point and 12 o’clock as the reference line, calculate the rotation angle of the center line relative to the reference line, and then The calculated angle is converted into a watch reading, and the current watch reading is stored in the database.

Description

technical field [0001] The invention relates to a vision-based automatic recognition method and system for instrument pointer readings. Background technique [0002] As a timing tool with a balance spring oscillating system, a mechanical watch’s travel time accuracy is changing every moment. According to the detailed rules for the evaluation of mechanical watches, high-precision mechanical watches must meet the requirements of “High-precision Mechanical Watches SQL / HSTU009 -2017" standard to obtain certification. In order to ensure the travel time accuracy of the watch, each watch needs to be tested for 16 days before leaving the factory, that is, during the 16 consecutive days of travel time, the accuracy of the watch is calculated by reading the travel time difference between the watch and the standard time every day. The different accuracy errors of the travel time at different positions can correct the speed trend of the watch. At present, this step is mainly operated ...

AI Technical Summary

Problems solved by technology

At present, this step is mainly operated manually. This method can only detect 200-300 finished watches per day, which has low efficiency and high labor costs. At the same time, the accuracy of detection is easily affected by the proficiency and working status of the staff.

The existing automatic reading technology is mainly applied to the instrument, and it is difficult to solve the problem of overlapping and mutual occlusion of multiple pointers of the watch. At the same time, due to the different styles and designs of the watch, there are a large number of patterns on the dial, which will greatly affect the recognition of the pointer. The diversity of pointers also greatly increases the difficulty of detecting the central axis of the pointer

Images