摘要
随着城市交通环境日益复杂,全景影像系统在汽车上的集成已成为解决碰撞风险的主流方案。然而,该技术应用也面临挑战,在视角重叠区域存在形变断层和色彩跳变问题。为此,文章围绕深度学习驱动下的全景影像生成技术展开探讨,深入分析多摄像头图像的空间映射关系,设计并开发基于嵌入式平台的模型。实车验证表明,该模型在典型泊车场景中的应用能够有效规避车身剐蹭风险,同时优化了极端环境适应性与硬件兼容性,显著提高行车安全性与技术普及性。
With the increasingly complex urban traffic environment,the integration of a panoramic image system in cars has become the mainstream solution to solve the collision risk.However,there are also challenges in the application of this technology,such as deformation fault and color jump in the overlapping area of view angles.Therefore,this paper discusses the panoramic image generation technology driven by deep learning,deeply analyzes the spatial mapping relationship of multi-camera images,and designs and develops a model based on an embedded platform.The real vehicle verification shows that the application of this model in typical parking scenarios can effectively avoid the risk of vehicle body scratches,while improving the adaptability to extreme environments and hardware compatibility,significantly enhancing driving safety and technical popularity.
作者
吴晓宇
孙佳慧
Wu Xiaoyu;Sun Jiahui(Guangzhou Automobile Group Co.,Ltd.,Guangzhou 510000,China)
出处
《汽车电器》
2026年第2期82-84,共3页
Auto Electric Parts
关键词
全景影像拼接
深度学习
性能验证
智能驾驶与辅助
panoramic image mosaic
deep learning
performance verification
intelligent driving and assistance
