融合改进Canny与RANSAC的圆形孔位识别方法

A Circular Hole Detection Method Based on Improved Canny Edge Detection and RANSAC

摘要: 为解决酶联斑点(ELISPOT)图像中孔位轮廓在不同因素干扰下难以准确识别的问题，提出了基于改进Canny边缘检测、弧段筛选与RANSAC圆拟合相结合的孔位检测方法。在边缘检测阶段，引入结构感知自适应高斯滤波与基于梯度幅值的动态双阈值策略，以适应不同复杂场景并防止真实边缘被抑制；随后引入弧段长度与方向一致性约束对候选弧段进行筛选，有效抑制斑点及噪声干扰；在圆拟合阶段，建立半径聚类约束的RANSAC模型，并结合半径一致性筛选与内圆优先选择策略，实现同心圆结构条件下的鲁棒拟合。以524张不同场景下的酶联斑点孔位图像为测试数据，设置圆心与半径误差阈值分别为5px和10px。实验结果表明，所提方法的孔位圆心检测准确率为95.42%，半径检测准确率为88.93%，在复杂成像条件下能够保持稳定的检测性能。

Abstract: To address the challenge of accurately identifying well contours in Enzyme-Linked Immunospot (ELISPOT) images under various interfering factors, a well detection method is proposed by integrating an improved Canny edge detection algorithm, arc segment screening, and RANSAC-based circle fitting. In the edge detection stage, a structure-aware adaptive Gaussian filtering strategy is introduced, together with a gradient magnitude-based dynamic double-threshold scheme, which enables the method to adapt to complex imaging conditions while preventing true edges from being suppressed; Subsequently, candidate arc segments are filtered by imposing constraints on arc length and directional consistency. This step effectively suppresses interference caused by spots and noise, thereby improving the reliability of contour extraction. In the circle fitting stage, a RANSAC model with radius clustering constraints is established. By integrating radius consistency filtering and an inner-circle priority selection strategy, the method achieves robust fitting under concentric circle conditions. Using 524 images of ELISPOT well positions from various scenarios as test data, and set the center and radius error thresholds to 5px and 10px, respectively. Experimental results show that the proposed method achieves a detection accuracy of 95.42% for well center detection and 88.93% for radius detection, maintaining stable detection performance under complex imaging conditions.