光学定位中近红外目标实时检测系统设计

Design of near-infrared target real-time detection system for optical positioning

为了提高基于立体视觉的光学定位中近红外(NIR)发光标记的定位精度,同时满足高速率高分辨率NIR图像的实时处理性能需求,基于Zynq平台设计了NIR目标的实时检测系统。利用平台集成的ARM+现场可编程门阵列(FPGA)架构,实现了3台工业相机的图像同步采集与传输、畸变校正、分割等功能,通过灰度重心法提取目标中心并传输给上位机进行目标的立体匹配及实时显示。部署了Linux系统对各模块进行控制及调度,采用高层次综合方法设计算法,实现算法的流水线并行运行。实验结果表明:系统能够对3台相机的高分辨率图像内目标以60帧/s的速率完成实时同步检测与定位,定位精度高,满足实际手术导航应用场景需求。

To improve positioning precision of near-infrared(NIR)luminous markers in optical positioning based on stereo vision and meet the real-time processing demand of high rate and high resolution NIR images, a real-time detection system for NIR target is designed based on Zynq platform.By utilizing the integrated architecture combining ARM and field programmable gate array(FPGA),functions including image synchronous acquisition and transmission, distortion correction and segmentation of 3 cameras are realized.Extract target center by gray centroid method and transmit to upper computer for stereo matching and real-time display of target.Linux operating system is developed for controlling and scheduling each module.High level synthesis method is used to design algorithm, realize pipelined parallel running.Experimental results show that the system can realize real-time synchronous detection and positioning of targets in high-resolution images of three cameras at a rate of 60 frames per second with high positioning precision, and satisfy the requirements in practical surgical navigation scenes.