Danube Dynamics and Aspöck Systems have introduced an AI-based inspection system that analyzes images of multi-chamber vehicle lamps in real time to verify that each light segment activates correctly at the end of the production line.
Aspöck Systems manufactures rear lamps with multiple functional chambers, such as indicators, brake lights and fog lights. During final inspection, each chamber must be activated individually and checked to confirm that only the intended segment lights up and that brightness and position correspond to specifications. Errors can include a chamber failing to illuminate, the wrong chamber activating, multiple chambers lighting simultaneously, or insufficient light intensity. Variations in orientation, ambient light, dust and product variants further complicate inspection.
To address these requirements, Aspöck Systems deployed the “auros for quality” software developed by Danube Dynamics. The system uses adaptive AI algorithms to evaluate image data captured during the activation sequence of each lamp. Two uEye XCP industrial cameras from IDS Imaging Development Systems capture high-resolution images of the lamp while specific chambers are switched on by the control system.
The AI processes these images on an industrial PC installed directly on the production line. It identifies which light segments are illuminated, determines their spatial position within the housing and compares the observed activation pattern against the expected configuration for the specific product variant. The software also assesses brightness distribution and detects irregularities such as weak illumination, unintended reflections or cross-activation between chambers.
Unlike rule-based inspection systems that rely on predefined thresholds and manually configured regions of interest, the AI model is trained on image data from acceptable and defective lamps. This enables it to recognize complex or previously unseen defect patterns without extensive reprogramming. When a new product variant is introduced, the system can be adjusted via the human-machine interface by loading the corresponding configuration and training data.
The cameras used in the installation, the IDS U3-3680XCP Rev.1.2 models, are equipped with 5.04-megapixel rolling shutter CMOS sensors. Their light sensitivity and compact form factor support stable image acquisition under varying production conditions. The IDS peak software environment is used to integrate the cameras into the inspection setup.
The two cameras operate in parallel. While one lamp is being analyzed, the next can be positioned on the test bench, allowing overlapping inspection cycles. Image evaluation is performed locally without cloud connectivity, reducing latency and ensuring that inspection decisions are available immediately.
Test results are displayed in real time on a touchscreen interface. Each inspection outcome is logged and can be transmitted to higher-level production systems via an interface, enabling traceability at the individual product level. Lamps that fail the AI evaluation are automatically classified as not compliant and can be removed from the production flow.
According to the companies, the system enables automated verification not only of whether a lamp lights up, but also whether each chamber performs its intended function under real production conditions. As inspection requirements in the automotive sector become more detailed and variant diversity increases, AI-based image analysis is being integrated into end-of-line quality control to manage complexity and reduce manual inspection effort.
