On the Relation between Optical Aperture and Automotive Object Detection

TL;DR

This work investigates how optical aperture size and shape influence automotive camera-based object detection using a PSF-based filtering pipeline. It employs Zemax ray tracing to model PSFs across distances, depth-guided PSF filtering, and AWGN-based noise balancing to create synthetic images for four aperture geometries, evaluated with YOLOv8n on 100 classes. Across traffic signs, traffic lights, and speed signs, the study finds no statistically significant advantage for non-circular or varying f-number apertures on detection performance, though very high gain degrades accuracy due to noise. The results imply potential cost savings from smaller-diameter optics and underscore the importance of PSF-aware synthetic data for reducing domain gaps in autonomous driving perception.

Abstract

We explore the impact of aperture size and shape on automotive camera systems for deep-learning-based tasks like traffic sign recognition and light state detection. A method is proposed to simulate optical effects using the point spread function (PSF), enhancing realism and reducing the domain gap between synthetic and real-world images. Computer-generated scenes are refined with this technique to model optical distortions and improve simulation accuracy.

Figures (20)

• Figure 1: Imaging & Optical System Model
• Figure 2: 'Plus' & Circular-Shaped Apertures
• Figure 3: MTF Comparison of 'Plus'-shaped Aperture vs. Circular Aperture
• Figure 4: PSF Array Example
• Figure 5: PSF Array - Zoom-In Example