基于Hough变换圆检测的颅脑穿刺靶标信息提取算法对比研究

Comparative study of algorithms for information retrieval of brain puncture targets based on Hough transform circle detection

目的 对比3种Hough变换算法(标准Hough变换、基于梯度的Hough变换、随机Hough变换)的原理与性能,为颅脑穿刺机器人精确、快速识别靶标并获取靶心坐标提供合适的控制基础。方法 在MATLAB软件中搭建仿真环境,研究分析图像特征识别、滤波、边缘检测、累加投票等处理工程,实现多个场景下目标圆的轮廓识别、拟合及其圆心坐标获取,定量化比对各算法的识别、拟合性能,最终根据颅脑穿刺机器人的实际应用环境确定更优的检测算法。结果 标准Hough变换算法拟合的标记圆与目标圆误差最大,同时因计算较大致使算法运行时间最长。随机Hough变换算法检测速度低于基于梯度的Hough变换算法的速度,但拟合准确度比标准Hough变换算法略好。基于梯度的Hough变换算法的圆拟合速度与准确度相比前两者而言具有明显优势。结论 基于梯度的Hough变换算法更适用于颅脑穿刺机器人系统的靶心坐标获取。

Objective To compare the principles and performance of three Hough transform algorithms(standard Hough transform,gradient based Hough transform,and random Hough transform)in order to establish a suitable control basis for precise and rapid recognition of targets and acquisition of target center coordinates for craniocerebral puncture robots.Methods A simulation environment in MATLAB software was built to study and analyze image feature recognition,filtering,edge detection,cumulative voting and other processing engineering.Contour recognition and fitting of target circles were achieved in multiple scenarios before their center coordinates were obtained.The recognition and fitting performance of these algorithms was quantitatively compared.Finally,a better detection algorithm based on the actual environment of the craniocerebral puncture robot was determined.Results The standard Hough transform algorithm had the largest error between the mark circle and the target circle,and the running time of this algorithm was the longest due to large computation.The detection speed of the random Hough transform algorithm was lower than that of the gradient‐based Hough transform algorithm,but the fitting accuracy was slightly better than that of the standard Hough transform algorithm.The speed and accuracy of circle fitting based on the gradient Hough transform algorithm had significant advantages over the other two.Conclusion The gradient based Hough transform algorithm is more suitable for obtaining the target center coordinates of the craniocerebral puncture robot system.