Instantly measuring quality on production lines.

Context

We automated the extraction of quality indicators from the production line of a large industrial group using video streams.

The Challenge

The production line of our client spans over 500 meters in length. Detecting quality issues as early as possible is crucial for improving operator responsiveness and corrective actions, thereby minimizing losses of manufactured products. Additionally, the production parameters at each site are numerous, and their optimization relies on a rigorous and precise quantification of product quality at strategic points along the line.

The Solution

We developed real-time processing algorithms for the video streams from the various cameras positioned along the production line. The proposed solutions combine convolutional neural networks to detect key components of the line and signal processing methods to identify edges and automatically extract the dimensions of the manufactured products.

Tech Stack

Results

The initial versions of our algorithms have been deployed in production at one of our client’s industrial sites, providing operators with real-time access to the measurements in the control room.