Manual truck height verification in manufacturing plants like Volvo Trucks is a critical but time-consuming and error-prone process. This thesis addresses the need for automation by developing and evaluating two distinct computer vision and machine learning-based systems for accurate truck height measurement.

The first system employs a 2D vision approach using a single industrial camera. All processing, including truck segmentation via a fine-tuned YOLOv11l-seg model, distance prediction using a linear regression model, and subsequent height calculation based on photogrammetric principles, is performed on a central server.

The second system utilizes an Intel® RealSense™ Depth Camera D456 for direct 3D depth sensing. This camera is connected to a Mini PC, which is solely responsible for capturing and streaming synchronized color and depth data to the central server.

The server then performs all subsequent operations: segmenting the truck from the streamed color images (using the same YOLOv11l-seg model), processing the corresponding depth data, and calculating the truck’s height.

Empirical evaluation demonstrated the viability of both server-centric approaches.The 2D system achieved a Mean Absolute Error (MAE) of 1.8 cm. The 3D system, with all processing of streamed data executed on the server, exhibited superior accuracy with an MAE of 1.2 cm. Both systems met operational processing time requirements.

This research concludes that while both developed systems offer feasible automated solutions, the 3D stereo vision approach, leveraging server-side processing of streamed depth data, provides higher measurement accuracy.

The findings offer Volvo Trucks a data-driven basis for enhancing quality control processes through automated dimensional inspection, highlighting the practical application of advanced vision technologies in industrial environments.