<p>The rise of Industry 4.0 has intensified the demand for advanced automation solutions in industrial environments. However, existing algorithms poses a significant challenge in these settings, where conventional approaches rely heavily on manually labelled datasets. This paper introduces a new framework for pointer meter reading recognition (PMRR), which substantially reduces human intervention through the integration of two key components: a Segment Anything Model (SAM)-based labelling system and a Multimodal Large Language Model (MLLM)-driven character recognition approach. A frame sequence mechanism in SAM is utilized to produce high-quality labelled datasets, effectively eliminating the need for extensive manual annotation typically required in semantic segmentation tasks. Additionally, the framework employs an MLLM, which is fine-tuned with Low-Rank Adaptation (LoRA), to achieve precise character recognition, even in demanding scenarios involving occluded or distorted text. Lastly, a correction methodology is improved to mitigate image truncation artifacts that arise during coordinate transformation. The performance of the proposed method is demonstrated by its consistent segmentation performance, which deviates by less than 3% from that of the human intervention approach, even under challenging conditions such as extreme brightness, blur, and occlusion. Additionally, the PMRR achieves a quoted error of only 0.23%, effectively matching the detection accuracy of similar algorithms.</p>