论文部分内容阅读
基于摄像机的航空器识别是机场场面监视的重要工具。针对多摄像机场面航空器识别算法存在的计算效率低等缺点,提出基于GPU CUDA的加速算法。利用CUDA线程并行处理能力与GPU计算能力,对算法进行了重新设计与优化。通过实地对多路场面视频监视数据进行了多次实验,验证了在NVIDIA Geforce 8800GTS显卡上可实现10倍以上的加速性能,提高了航空器目标识别效率,可以满足机场场面监视中对航空器识别与跟踪的实时性要求。
Camera-based aircraft identification is an important tool for airport scene monitoring. Aiming at the disadvantage of low computational efficiency of aircraft recognition algorithm in multi-camera scene, an acceleration algorithm based on GPU CUDA is proposed. CUDA thread parallel processing capabilities and GPU computing power, the algorithm has been redesigned and optimized. Through field experiments on multi-channel video surveillance data, we verified that the NVIDIA Geforce 8800GTS graphics card can achieve more than 10 times acceleration performance and improve the efficiency of aircraft target recognition to meet the needs of aircraft identification and tracking in airport scene surveillance Real-time requirements.