Efficient Tire Tread Depth Estimation with ESP32CAM and U-Net Based Segmentation

• Andre dos Santos Edge Innovation Center / Vertex Institute of Innovation and Technology
• Derek N. A. Alves Edge Innovation Center / Vertex Institute of Innovation and Technology / UFAL
• Bruno G. Ferreira Edge Innovation Center / Vertex Institute of Innovation and Technology / University of Porto
• Tiago F. Vieira Edge Innovation Center / Vertex Institute of Innovation and Technology / UFAL

Resumo

Tire inspection is essential for ensuring the safety and efficiency of land vehicles, such as cars and trucks. Reg-ularly checking tire tread depth helps prevent accidents like aquaplaning and reduced braking performance, and also lowers fuel consumption. Traditional estimation methods, like depth gauges, are time-consuming, especially for vehicles with multiple tires, and often imprecise due to the rounded shape of some tire treads. To address these challenges, a portable device is proposed, consisting of cameras, microcontrollers and lasers, capable of quickly capture the tire tread shape. The proposed device connects to an Android application via WiFi, offering the user a graphical interface to conduct tire tread inspections on vehicles. The device retrieves the images from the cameras and uses an artificial intelligence model to estimate the tire treads shape, which are highlighted by lasers. At the end, the device provides the tire tread depth estimation in millimeters, allowing the user to determine if the tire is still functional or not. The proposed device significantly reduces the inspection time per tire by 69.46% compared to traditional depth gauges, with a tread depth estimation error of 0.112 millimeters. The proposed device can be integrated into tire management and maintenance companies to enhance the efficiency of tread inspection processes. This innovation offers significant benefits in both scientific and economic contexts, improving overall quality control and productivity in tire inspection process.

Palavras-chave: Tire Tread Depth Estimation, Tire State Analysis, Embedded System, Tire Inspection, Artificial Intelligence, Deep Learning, ESP32CAM, U-Net model, Attention-based model, Android device