AVID – AI Visual Inspection Data Fusion System for Car Damage Inspection
AVID is an AI vision-based system to detect car damage using precise 3D LiDAR data fused with multimodal 2D images. Combining different technologies (sensor fusion) ensures robustness, reliability and accuracy; reduces the false alarm ratio and enhances the detection confidence. The AVID solution is based on the L3CAM, the commercial name of Beamagine’s product that combines high-resolution 3D images with 2D. Those are the main sensors included:
3D Data: A plurality of Light Detection And Ranging (LiDAR) sensors providing precise high-density 3D data covering the entire inspection area, enabling AVID to accurately measure detailed geometric/volumetric information of the car.
2D Data: A set of camera sensors providing 2D images of different complementary modes (usually colour and thermal). This 2D information offers rich semantic data for AI-based inspection. Seamlessly integrated using 3D data, these modes enhance the detection capabilities covering a wider range of car damages.
Beamagine accumulates expertise and knowledge about optical systems, sensor integration, data fusion and AI-based perception with a cross-functional team of experts from more than 10 years, who have developed our L3CAM device to surpass other market products that present inconsistencies in the data fusion and does not exploit all the benefits from all sensors. Hence, these unique capabilities make it feasible to turn AVID into an integrated parallax-free (congruent data fusion) solution.
The main added value of AVID is that end-users receive high-confidence and precise data analytics under any operational condition, indoors or outdoors and in a static or dynamic scenario. Based on our expertise, we propose a two-stage inspection system (go-no go) to meet efficiency and precision requirements, to be discussed with the end-user.
The first stage is a fast inspection done by a gantry with the car driving through at a speed of up to 50 km/h. The gantry, equipped with sensors and an OCPU (On-site Control and Processing Unit), integrates data, processes it (AI-based video analytics) and communicates results via Ethernet. We already installed and verified a similar system for the construction sector.
If any damage is detected and a more detailed record is required for potential client reclamations, the second stage inspection uses a hand-held device for a stationary car inspection. This AI-assisted check takes under 1 minute and achieves millimetre accuracy, as we proved in large and complex scenarios on construction sites.
All detections and evidence are transferred to the car rental firms' back-end IT system via standard protocols. AVID integrates seamlessly into a network or cloud infrastructure thanks to the dedicated OCPU. For example, it may send a defined message containing an identifier (the car’s plate), metadata (date and time) and the inspection results. Handling and storing of personal data will be done in strict accordance with Directive 95/46/EC (GDPR).
AI systems need large datasets to generalize and achieve optimal performance, making AI deployment high-resource-consuming (time and human). Beamagine capabilities about Digital Twins and AI training boost the deployment using hybrid AI training based on synthetic datasets that pre-train the deployed AI and an agile training methodology to enable continuous learning to achieve faster its optimal performance.
In summary, the superior advantages of AVID offer the following capabilities and functionalities to meet the requirements:
A dedicated processing unit to gather, compute and send the data to the client's IT architecture
An easy-to-install visual inspection device based on high-fidelity parallax-free 3D+2D data to automatically detect damage on static or moving vehicles up to 50 km/h
Two possible inspection stages to fast damage evaluation below a minute and data transfer to the operator
GDPR-compliant system because any personal or biometric data will be instantly removed from raw images with a dedicated AI propriety of Beamagine
Customizable data analytics and damage reports to provide all documentary evidence required by the car rental companies
Robust and reliable AI-based system under any operational condition
Solution based on products already available in the market with a clear scalability path
Continuous learning of the AI through an agile methodology
Intuitive UX/UI with customizable dashboards
The fast inspection stage can be installed in the parking entry, turning AVID into a non-invasive and unnoticed solution for car drivers, saving them time when in the rush to take the flight
This is the key value for car rental companies, enabling them to build strong customer relations based on customer satisfaction and loyalty and an opportunity for Frankfurt Airport to stand out as a high-tech-driven airport with advanced facilities.
