In the rapidly evolving landscape of automotive technology, vision sensing has emerged as a crucial component for intelligent driving systems. A recent article published in Engineering, titled “Vision Sensing for Intelligent Driving: Technical Challenges and Innovative Solutions,” offers an in-depth analysis of the current state of vision sensing technology and explores potential solutions to enhance its performance.

The authors, Xinle Gong from the School of Mechanical Engineering at Beijing Institute of Technology and the School of Vehicle and Mobility at Tsinghua University, along with Zhihua Zhong from the School of Automotive Studies & College of Transportation at Tongji University, delve into the intricacies of vision sensing systems, highlighting their significance in providing critical road condition data and supporting autonomous driving functions. Compared to other in-vehicle perception sensors, vision sensors offer more detailed and comprehensive environmental information, which is essential for intelligent driving decision-making and control.

However, the article underscores that automotive-grade cameras face more complex and stringent technical requirements compared to industrial and consumer cameras. These cameras must adapt to variable driving environments, maintain high performance in image processing, hardware robustness, and system integration under different lighting conditions. They also need specialized materials and advanced packaging technologies to ensure good imaging quality even in adverse weather conditions.

The authors identify several key challenges faced by vision sensors. For instance, optical lenses are constrained by physical factors such as depth of field, aperture diffraction, and lens size, which affect image quality under varying lighting conditions. Current lenses often fail to preserve critical details in high-contrast scenes, hindering accurate recognition of road conditions and traffic signs. Additionally, wide-angle lenses introduce optical distortion, leading to errors in object recognition and distance estimation. The sharpness of lenses is also limited by the diffraction limit, making it difficult to improve resolution while maintaining miniaturization.

CMOS image sensors face challenges related to the trade-off between resolution and frame rate. High spatial resolution increases the number of pixels and the amount of image data collected, but this lengthens data processing time and reduces temporal resolution. Current automotive-grade CMOS sensors typically achieve a dynamic range of only 120–140 dB, which is insufficient to accommodate variations in lighting intensity across all scenarios encountered in intelligent driving. Moreover, increasing resolution reduces pixel size, decreasing saturation charge capacity and further limiting the dynamic range of CMOS image sensors.

Image signal processors are also limited in their on-chip and parallel processing capabilities, affecting the speed of image processing, especially for high-resolution and high-frame-rate image data. This limitation directly impacts the real-time response capability of intelligent vehicles. In low-light conditions, sensor noise increases, and the noise reduction and signal enhancement algorithms of current image signal processors require improvement.

To address these challenges, the authors propose several innovative solutions. They suggest exploring new high-efficiency photosensitive materials, such as quantum dots and perovskites, to enhance sensing performance. Architectural evolutions inspired by the human visual system can achieve higher dynamic range and ultra-high-speed imaging. Additionally, developing intelligent image processing algorithms based on neuromorphic and quantum computing paradigms can enable real-time, energy-efficient, and robust perception in complex environments.

The article provides a comprehensive analysis of the technical challenges faced by in-vehicle vision sensors and offers targeted ideas and recommendations for future research and development. By leveraging new materials, advanced architectural designs, and intelligent algorithms, vision sensing technology can continue to evolve towards greater intelligence, reliability, and integration, ultimately enabling safer, more efficient, and fully autonomous intelligent driving.

The paper “Vision Sensing for Intelligent Driving: Technical Challenges and Innovative Solutions” is authored by Xinle Gong, Zhihua Zhong. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.06.038. For more information about Engineering, visit the website at https://www.sciencedirect.com/journal/engineering.