Research On The Key Technologies Of Automatic Reading Recognition Based On Machine Vision For The Circular Pointer Instrument

Pointer instruments have been widely used in power industry,metallurgy and other industrial production area owing to the simple structures,conveniently reading and less affected by external environment.The automatically reading recognition technologies based on computer vision are researched in this paper which improves the reading accuracy and efficiency.First of all,the original image of circular pointer instruments is picked under the natural light condition using the camera.Because there are the camera distortion when the original image of the circular pointer instruments is collected,camera calibration is performed using the calibration method of the two-dimensional planar target and then the camera's internal and external parameters are picked.The distortion correction of the original images is performed using these calibration parameters.Then,after distortion correction step,the images are preprocessed based on the technology of digital image processing,such as filtering and denoising,threshold binarization and Laplacian edge detection,and so on.The noise interference is eliminated in the instrument image and the useful dial information is enhanced.Because of the complexity of the environment of the pointer meters,the installation position of the pointer meters is not perpendicular to the ground accurately.There's a degree of rotation tilt when acquiring the images of the circular pointer instruments.The feature extraction algorithm of SIFT(Scale Invariant Feature Transform)is used to extract the key points of the test images and template images.Then the rotation tilt errors are corrected using the method of template matching.So,the angle of the pointer is calculated more accurately and the reading recognition error can be eliminated or be decreased due to the tilt rotation of the dial.For the pointer instrument images after rotation tilt correction,we calculate the angle of the pointers after detecting the circular instruments' contours and the pointers,and then,we calculate the current reading using the method of pointer angle calculation.Because the low efficiency of the classic Hough transformation to detect circles which must transform the points from Cartesian coordinate space to parameter space,the maximum connectivity area algorithm is proposed to detect the pointer instruments' contours.This algorithm analyses the connectivity between each referenced pixel and its eight surrounded pixels to get the maximum connectivity area,then the contour of the pointer instrument is located.The morphological dilation and erosion operations are used to extract the skeleton of the pointer.The algorithm of the Hough Transform is used to detect the pointer internal the dashboard.We also compare the detection efficiency and accuracy of the algorithms such as the least squares,region growth and the Hough Transform method.Last,the accuracy and efficiency of the automatically reading recognition method based on the method of camera calibration and maximum connectivity area is verified through experiments.The experimental results show that this method can improve the recognition accuracy and efficiency to a large extent and it has a certain theoretical and practical significance.