Automatic Recognition Technique Research And Implementation Of The Analog Display Instrument

Analog Measuring Instruments are widely used in metrology, mines industry and power plants to monitor various parameter indexes due to their simple structure, convenient maintenance, anti- electromagnetic interference ability, as well as durableness, especially widely used to monitor operating status of equipment and pressure of gas in power switch when power plants are pursuing problems of devices should be predicted, located and solved quickly, while artificial patrol, reading and recording cannot meet the demands of efficiency, Recognizing analog dial automatically and monitoring dials distributed have been related to power plants’ automatization and informatization. the former item is prerequisite of extracting, fusing, mining information of power plants’ devices’ ope rating parameters and providing references and guiding opinion for production management. Currently computer vision is mostly adopted, whose key technologies involve locating dial’s rotation center, extracting dial’s pointer centerline, recognition of multi-pointer dials, and the algorithm should be Self-adaptive to illumination variation. These technologies have gotten progress to some degree, while neither mature nor deep, this paper takes the substation renovation subprojects as application background, aims to recognize five kinds of analog dial instrument automatically based on computer vision. Through comparing with common algorithm, designs and proposes a feasible point meter recognition scheme, the research contents and conclusions are as follows:1 This thesis analyzes and summarizes the process of automatic reading of dial image, combined with whose design structure. While current method need grab multi images or calculate too much when locating dial’s rotation center. This thesis proposes a new method based on improved Canny edge detection operator, which can acquire rotation radius and center quickly based on single dial image. also avoid disturbance of dials’ complex background by selecting effective displaying region.2 Considering dial’s pointer centerline passing by rotation center, this thesis redesigned Hough transform passing by a specified point, Time consumed reduced from O(N3) to O(N2) when locating pointer centerline. Which can be applied to dial’s binary image without adopting thinning operation before, Improved the robustness of the automatic meter reading scheme.3 Given that illumination of dial’s image may variate a little to some degree, By the analysis and contrast with familiar image illumination processing algorithms, this thesis adopt local thresholding algorithm based on gaussian operator. Which can obtain better segmentation effect at a lower time complexity cost.4 This thesis further study dial plate with multi pointers, extracts red pointer based on different Color proportion, remove interferential graduation line and reserve pointers’ region at the same time through flitering algorithm based on area threshold of connected-region, solving the problem that dial’s pointer would be difficult to detect when it’s been truncated by liquid lever. we designed experimental platform based on scheme proposed and concluded that deviation is no more than 1/3 of dial’s minimum graduation line, whose accuracy appears higher than estimate reading, which verified the scheme’s feasibility as well as possibility in engineering.