基于计算机视觉的自动驾驶环境感知算法优化

打开文本图片集

中图分类号：U463.6 文献标识码：A 文章编号：1003-8639（2026）03-0079-03

Optimization of Autonomous Driving Environmental Perception Algorithms Based on Computer Vision

Li Yongsi, Li Shuaijing (Shangqiu Institute of Technology, Shangqiu 476ooo, China)

【Abstract】In the inteligent transformation of automobiles,autonomous driving environmental perception (relying on computer vision) is the core of vehicle safetydecision-making.This paper optimizes computer vision perception algorithms for complex driving scenarios: a dynamic feature extraction strategy is proposed to address missed detection of irregular targets,a lightweight network is designed tobalance theaccuracyandreal-time performanceof semantic segmentation,and a "vision + radar" fusion framework is constructed to compensate for perception deficiencies in extreme scenarios.Experimental tests demonstrate thattheoptimized algorithm outperforms mainstream methods suchas YOLOv5 and DeepLabv3 + ,enabling the engineering application of autonomous driving perception systems. It enhances driving safety,accelerates the commercializationof autonomous driving,and provides practical technicalsolutions foradvancing the industrialization of autonomous driving.

【Key words】computer vision;autonomous driving of automobiles;environmental perception；algorithm optimization

1汽车自动驾驶环境感知与计算机视觉技术

计算机视觉技术凭借低成本、高信息密度的优势，成为汽车自动驾驶环境感知系统这一保障车辆安全行驶的“感知中枢”的核心支撑。（剩余5161字）