Method for recognition of indication values of a pointer meter

Pointer meters are commonly used in various fields, including energy, transportation, and manufacturing. Accurately recognizing the values of pointer meters is crucial for operating the systems and equipment. Manual reading methods for recognizing a large number of pointer meter indication values are time-consuming and prone to errors and omissions. Compared to manual reading, the use of machine vision technology for recognizing pointer meter indication values can effectively reduce reading time and the probability of errors, providing clear advantages. This paper proposes a method for recognizing pointer meter indication values by first using image processing algorithms to obtain the pointing angle of the pointer on the dashboard. After that, a deep learning algorithm is used to obtain the keywords and corresponding coordinates on the dashboard, and the coordinates of the keywords are transformed into the angles of the keywords. Then, the angle method is used to calculate the relationship between the pointer's pointing angle and the keyword angle to obtain the final test results. The test results show that compared to the manual direct reading, the proposed reading method has a relative error within 2.7% when used under ideal conditions or conditions with a certain degree of interference, such as interfering objects on the dashboard, offset in the observation angle, and low light conditions.