Environmental Adaptability of Visual Perception System from Vibration-blurred Image Perspective

Abstract

Vision-based perception systems are widely used in unmanned platforms and have an important role in tasks such as object classification, object detection, and instance segmentation. The function and performance of visual perception system is not only related to the sensing hardware and recognition algorithm, but also influenced by the external environment. In this paper, the detection performance of the artificial intelligence model is used as a metric to study the impact of vibration-blurred images on the environmental adaptability of visual perception system. The structure and function of AI-baesd visual perception system are analyzed, and two vibration-blurred image generation methods, vibration test and vibration simulation, are introduced. The experiment quantitatively verified the effect of mechanical vibration stress on the performance of AI model during the operation of visual perception system.