In order to enhance performance of visual perception system object detection model in complex environments such as lighting variations, blur, noise, and occlusion, this study investigates the performance variations of object detection models under different complex scenarios using YOLO 11x and data augmentation techniques. The results indicate that various environmental disturbances have distinct impacts in model perfprmance, with Gaussian noise and motion blur having more pronounced impacts, while brightness adjustments and random occlusions are relatively less disruptive.

Object detection  /  Complex environments  /  Data augmentation  /  Autonomous driving