Read Pointer Meters in complex environments based on a Human-like Alignment and Recognition Algorithm

TL;DR

This paper tackles the practical problem of reading pointer meters in complex industrial scenes, where traditional pipelines struggle with speed and robustness to image degradation. It introduces a unified end-to-end framework consisting of a YOLO-based meter detector (YDM), a Spatial Transformed Module (STM) for implicit, fast front-view alignment, and a Value Acquisition Module (VAM) that jointly retrieves meter components and recognizes the numeric readings. The approach is validated on a new Meter Challenge MC1296 dataset, demonstrating high detection accuracy, faster alignment, and accurate readings (notably with around 25 FPS) even under challenging conditions, and shows that end-to-end training of VAM outperforms two-stage baselines. The results support practical deployment in robotic inspection, with potential for real-time video meter reading and broader industrial adoption.

Abstract

Recently, developing an automatic reading system for analog measuring instruments has gained increased attention, as it enables the collection of numerous state of equipment. Nonetheless, two major obstacles still obstruct its deployment to real-world applications. The first issue is that they rarely take the entire pipeline's speed into account. The second is that they are incapable of dealing with some low-quality images (i.e., meter breakage, blur, and uneven scale). In this paper, we propose a human-like alignment and recognition algorithm to overcome these problems. More specifically, a Spatial Transformed Module(STM) is proposed to obtain the front view of images in a self-autonomous way based on an improved Spatial Transformer Networks(STN). Meanwhile, a Value Acquisition Module(VAM) is proposed to infer accurate meter values by an end-to-end trained framework. In contrast to previous research, our model aligns and recognizes meters totally implemented by learnable processing, which mimics human's behaviours and thus achieves higher performances. Extensive results verify the good robustness of the proposed model in terms of the accuracy and efficiency.

Figures (7)

• Figure 1: (a) shows the efficiency of our STM (2) for meter alignment, which is 5 times faster than the conventional perspective transform method (1). (b) shows that our VAM (bottom line) can read more accurate values in some low-quality images than prior methods (top line).
• Figure 2: Overview of previous pointer meter reading pipeline (a) compared to ours(b). "Det", "PM", "PT" and "CR" represent meter detection, point matching, perspective transform and component retrieval. "STM" and "VAM" are our spatial transformed module and value acquisition module.
• Figure 3: The proposed framework of the pointer meter recognition. YDM can detect meter targets and crop meter regions into STM, where aligned views can be obtained. VAM can output meter values accurately and efficiently.
• Figure 4: Visualization results of one sample in the dataset.
• Figure 5: Qualitative results of the meter detection, where the yellow bounding box indicates the pointer meter and the green bounding box indicates the digital meter. "ID-num" is the detection confidence.