基于正交投影特性的带状目标检测

Detection of band object based on the characteristics of cross-projection

图像中带状目标在行列上具有规则的几何特性,其边缘映射图在笛卡儿坐标系下的正交投影具有显著性特征。由此提出了一种基于正交投影特性的带状目标检测方法。基于Hough变换的思想,引出一种投影变换,利用投影宽度谱的局部峰值定位边缘映射图中的带状目标。为减小噪声点影响,给出一种距离约束准则下的序贯最小二乘直线拟合方法。对两组图像数据进行了算法验证,实验结果表明,算法识别精度高、实时性好,具有很强的鲁棒性。

The band objects in image have inerratic geometric characteristics on the rows and columns. Thus, the cross-projection of the edge-mapped image presents prominent characteristics in the Cartesian reference frame. A method to detect the band object based on the characteristics of the crossprojection is presented. First, a project-transform which is based on the idea of the Hough transform is discussed. Then, the width at the local salient peak of the projection width spectrum is used to extract the band object from the edge-mapped image. A sequent least square method with distance constraint is used to compute the line parameters from the image to reduce the influence of the noise. Experimental results indicate that the algorithm has high recognition precision, good time efficiency and strong robustness.