Mengkun Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Mengkun Zhang has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Hui Zhang
Henan University of Technology
1 shared publication - Jilong Wang
College of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, China
1 shared publication - Shijie Zhang
College of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, China
1 shared publication - Xiang Wu
School of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, China
1 shared publication - Yu’Ang Qin
School of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, China
1 shared publication - Zhanyuan Cao
School of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, China
1 shared publication
Latest Publications
- Open Access
- 21 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (20) , 4875
Automatic Reading of Analog Pointer Meters Based on YOLO Localization and Template-Calibrated Angle-to-Value Mapping
Automatic reading of analog pointer meters is affected by complex backgrounds, illumination changes, and camera viewpoints. This study implements and evaluates a template-calibrated reading procedure that integrates established detection and image-processing components for fixed-layout analog instruments with instrument-specific pointer segmentation. YOLO localizes the meter, and the detected region is aligned with a manually calibrated template. Illumination-adaptive color segmentation and geometric processing estimate the pointer angle, which is converted to a reading by interpolation between adjacent calibrated angles. Compared with endpoint-based linear mapping, this procedure uses intermediate scale information to represent local angle–value relationships. Relative to approaches requiring dedicated pointer-segmentation or reading-regression networks, its reading stage avoids pixel-level pointer annotations and dedicated network training, while requiring manual template calibration and suitable pointer-segmentation settings. On the 58-image independent real-scene detection test set, YOLO achieved an mAP@0.5 of 99.50% and an mAP@0.5:0.95 of 77.72%. On 58 still images of a single 0–1.6 MPa pressure gauge, the method achieved a mean absolute error (MAE) of 0.0095 MPa, a root-mean-square error of 0.0135 MPa, and a maximum absolute error of 0.0408 MPa relative to manual reference readings. Image-level predictions were matched to the means of 16 supplied manual readings per image. These errors quantify agreement with those means; the variability and unvalidated uncertainty of the manual references preclude interpreting the MAE as equivalent absolute pressure-measurement accuracy. Supplementary evaluation on 26 images of a second pressure gauge and 28 images of a thermometer yielded scale-normalized MAEs of 0.1870% and 0.0787%, respectively, after instrument-specific calibration and segmentation configuration. These evaluation images informed segmentation-parameter selection; the results therefore illustrate reuse after configuration, without establishing generalization to unseen instruments or images.
- 21 October 2026
- Published by MDPI AG in Electronics
- Vol. 15 (20) , 4875
Open AccessAutomatic Reading of Analog Pointer Meters Based on YOLO Localization and Template-Calibrated Angle-to-Value Mapping
Automatic reading of analog pointer meters is affected by complex backgrounds, illumination changes, and camera viewpoints. This study implements and evaluates a template-calibrated reading procedure that integrates established detection and image-processing components for fixed-layout analog instruments with instrument-specific pointer segmentation. YOLO localizes the meter, and the detected region is aligned with a manually calibrated template. Illumination-adaptive color segmentation and geometric processing estimate the pointer angle, which is converted to a reading by interpolation between adjacent calibrated angles. Compared with endpoint-based linear mapping, this procedure uses intermediate scale information to represent local angle–value relationships. Relative to approaches requiring dedicated pointer-segmentation or reading-regression networks, its reading stage avoids pixel-level pointer annotations and dedicated network training, while requiring manual template calibration and suitable pointer-segmentation settings. On the 58-image independent real-scene detection test set, YOLO achieved an mAP@0.5 of 99.50% and an mAP@0.5:0.95 of 77.72%. On 58 still images of a single 0–1.6 MPa pressure gauge, the method achieved a mean absolute error (MAE) of 0.0095 MPa, a root-mean-square error of 0.0135 MPa, and a maximum absolute error of 0.0408 MPa relative to manual reference readings. Image-level predictions were matched to the means of 16 supplied manual readings per image. These errors quantify agreement with those means; the variability and unvalidated uncertainty of the manual references preclude interpreting the MAE as equivalent absolute pressure-measurement accuracy. Supplementary evaluation on 26 images of a second pressure gauge and 28 images of a thermometer yielded scale-normalized MAEs of 0.1870% and 0.0787%, respectively, after instrument-specific calibration and segmentation configuration. These evaluation images informed segmentation-parameter selection; the results therefore illustrate reuse after configuration, without establishing generalization to unseen instruments or images.