Pointer Meter Identification In Complex Environments

Pointer meters play a very important role in our daily life and industrial production.Compared with digital meters,they have the advantages of low cost and high reliability.The interpretation of pointer meters is mainly done by manual method,which is a waste of manpower and inefficiency.In view of the shortage of manual reading and the rapid development of digital image processing technology,many automatic interpretation methods of pointer meters have been proposed one after another.At present,most of the automatic interpretation methods of pointer meters are based on indoor single meter,and the experimental objects and environment are simple.In view of the shortcomings and limitations of the existing methods,an automatic interpretation method in two complex environments,indoor multi meter and outdoor single meter is proposed in this thesis.For indoor multi meter,four square pointer meters of factory switch cabinet are taken as the research object,and a monitor system of switch cabinet meter is designed.The main contents and innovations of this part are as follows:(1)Fisheye image capture and correction.In order to solve the problem that the ordinary lens with narrow angle can not monitor multiple meters in close range,fisheye lens are used instead of ordinary lens.The image taken by fisheye lens contains severe distortion,which is usually corrected by Zhang Zhengyou’s calibration method.In view of the problem that only radial distortion is considered and tangential distortion is ignored in Zhang Zhengyou’s calibration method.tangential distortion is considered to improve the rationality of the calibration algorithm in this thesis.(2)Multiple meters positioning.In order to locate each meter in a multi-meter image,the Canny operator is used to get the image edge,and the area where the meter is located based on the outline characteristics of the image edge.In view of the problem that Canny operator needs to set global high and low thresholds artificially,the adaptability and threshold effect are poor.A local gray statistics method is proposed to compute the thresholds in this thesis,which can avoid manual troubles and improve the threshold effect.(3)Improvement of image matching algorithm.In the process of scale number recognition,the MAD algorithm is prone to errors because of lighting.A MAD algorithm based on FT significance detection is proposed to focus on the target object and weaken the impact of the background.For outdoor single meter,a single elliptical pointer meter in substation is taken as the research object,and a substation pointer meter recognition system is designed.The main contents and innovations of this part are as follows:(1)Light processing.In view of the influence of complex illumination on the recognition results in substation.A Retinex algorithm with edge preservation based on Sobel edge detection is proposed in this thesis,and combines the improved Retinex algorithm with histogram equalization to improve image contrast while adjusting illumination.(2)Motion blur processing.The image of pointer meter in substation is taken by patrol robot which is always moving,and some images may be motion blurred.Getting the fuzzy direction and scale by using the spectral and cepstrum characteristics of the image in this thesis,and using Lucy-Richardson algorithm to restore the image.(3)Raindrop processing.In view of the influence of raindrop on the recognition results in substation.Combining with guided filter and anisotropic diffusion filter to remove rain in this thesis which is better than using the guided filter alone.(4)Detecting pointer.The Hough transform with the center constraint can detect the pointer quickly,but only if the center position is accurate,otherwise the result is likely to be wrong.In this thesis,the center constraint is extended to the center area constraint to improve the anti-interference of pointer detection.(5)A pointer meter recognition system for substation design and Implementation.It includes three steps: system analysis and overall design,system detailed design and implementation,and system testing.