Research On Reading Recognition Algorithm Of Industrial Pointer Instrument

Industrial pointer-type meters are widely used in industrial manufacturing for their high reliability and low maintenance cost,and most of them are read manually,which not only causes low productivity,but also brings errors to the readings due to human fatigue level and other factors.Although scholars at home and abroad have made great research on pointer meter reading recognition,there are still many problems and challenges due to the vulnerability of pointer meters to external factors such as light.In this paper,an industrial pointer meter reading recognition algorithm is studied as follows.(1)For the problem of tilt in the captured images due to the complex installation environment of pointer-type meters,an improved method of tilt correction of pointer-type meter images is studied.Firstly,the locations of the feature points determined by the scale-invariant feature transformation method are used as candidate feature points,then the candidate feature points are filtered based on the changes of pixel values in the circular neighborhood of each candidate feature point,and finally the eligible ones are retained as feature points and used for tilt correction of the meter images.The results show that the efficiency of the improved tilt correction method increases by 4.14%,and the instrument dials can be corrected to a positive view using the method in this chapter.(2)To address the problem that the rich information of dials in pointer meters leads to the difficulty of extracting the scale lines and pointer regions,an improved method of segmenting the reading regions of pointer meters based on Mask R-CNN is studied.Firstly,the pointer region of the Mask R-CNN segmented pointer is used as the basis,and the edge point at the pointer tip is used as the center point on the original map to construct a rectangle,then the contour tracking algorithm is used to detect the pointer region inside the rectangle,and finally the two parts of the pointer region are combined to get the final pointer segmentation result.The results show that the improved segmentation method improves the segmentation accuracy by 0.81%,reaching99.32%.(3)A method of reading recognition and calculation is studied for the problem of low accuracy of pointer meter reading recognition.Firstly,the polar coordinate transformation is used to adjust the scale area,and then the projection is used to complete the positioning of the scale;then the binary mask of the pointer extracts the center line of its contour,and the pointing of the pointer is judged according to the distance relationship between the center of rotation of the pointer,the center of mass of the contour and the two end points of the center line of the pointer;finally,the overall scale of the dial and the corresponding scale value are combined,and the corresponding reading of the pointer is calculated according to the distance relationship between the two scale lines adjacent to the pointer and the Finally,the reading corresponding to the pointer is calculated according to the distance relationship between the pointer and the two adjacent scale lines.The results show that the reading error calculated by this method can reach 1/4 of the index value.