Abstract
Pointer-type meters can suffer from inefficiencies in using manual data reading due to the lack of digital interfaces. Traditional pointer gauge reading recognition algorithms can only operate in specific environments or fixed locations without high reliability. In this paper, by introducing the traditional dial extraction algorithm Hoff circle detection, we propose the improved Mask R-CNN deep learning algorithm for automatic recognition of pointer-type dashboard based on Mask R-CNN algorithm, and introduce the maximum pooling algorithm for image feature extraction and the improved method of using PrROIPooling pooling technique. Experiments show that the improved Mask R-CNN algorithm improves the target detection accuracy by 2.1% and the instance segmentation accuracy by 1.9%. Compared with the traditional dial localization algorithm, this algorithm has the features of hih accuracy and robustness.