(Discussion) AI/ML Approaches for Microwave Link Configuration Inspection

Hi,

I’m currently conducting research for my master’s thesis on automating the inspection of microwave links. In my role at a regulatory authority, we regularly inspect microwave link configurations to ensure they align with the licenses held by telecom operators in our country and to address interference complaints.

Since operators are unable to send configuration data directly from their NOC, they instead provide screenshots of the requested microwave links for verification. Due to the high number of links that must be checked manually, I’m exploring ways to automatically extract configuration data from these screenshots using large language models (LLMs) or generative AI. Given the limitations in resources and budget, I’ve also tested simpler OCR-based solutions. So far, PaddleOCR has delivered the best results for extracting data from NEC and E******* (E///) screenshots.

We also have access to historical inspection records and license data from the past few years. I’d like to ask:

• Do you know of any similar solutions being used in other countries or by other regulatory authorities?
• Could you suggest additional research ideas related to this topic?
• In your opinion, is this topic valuable enough to pursue further, or should I consider shifting to a different research direction?