基于正交投影宽度谱的带状目标检测

Detection of ribbon objects based on the cross-projection width spectrum

针对带状目标在行列上具有规则的几何特性,提出了一种基于正交投影宽度谱的带状目标检测方法,将边缘映射图在笛卡儿正交方向上投影,建立边缘点对的投影宽度谱。采用D-S证据理论融合边缘点对的亮度、梯度方向、亮度方差和宽度特征,抑制背景边缘点对的干扰,使带状目标的宽度谱呈现为突峰。为消除噪声点对的影响,利用距离约束的序贯最小二乘拟合方法定位边缘映射图中的带状目标。实验结果表明:该算法识别精度高、实时性好,具有很好的鲁棒性。

In accordance with inerratic geometric characteristic of ribbon objects in row and column direction, a method to detect ribbon objects in image based on the cross-projection width spectrum （CPWS）was presented. Firstly, the original image including ribbon objects was transformed into edgemapped image using edge detectors. Then, the edge-mapped image was projected in the direction of Cartesian coordinates to establish CPWS of edge point pair. Multi-characteristics including luminance, grads direction, variance of luminance and transcendental width of edge pair were fused to suppress clutter of background edge point pair based on D-S evidence theory, which made width spectrum of ribbon objects present prominence in CPWS. A sequential least square method with distance constraint was used to locate ribbon objects to reduce noise effect. Using airport images to test the performance of this algorithm, the experimental results show that the algorithm can detect ribbon objects in image correctly, has high recognizable precision, good real-time and strong robust.