Barrelyzer: Design and Implementation of a Low-Cost Tank Barrel Inspection System
摘要
Maintaining confined spaces such as 10 cm barrels and pipes is a challenging task to perform without the use of intelligent systems. However, these devices are high-priced and require special training for the operators. Hence, the present paper proposes the design of a low-cost cannon barrel inspection robot. An important factor to consider in the design is the rifling of the cannon barrels. As a result, the robot was implemented and able to navigate on a simulated environment through rifling, measuring the position of the robot with an error of 4% with a confidence level of 99% and detecting defects using Convolutional Neural Networks with an F1-Score of 82.70% using AdamW optimizer.