摘要
采用通常方法处理暗背景下运动发光目标,因监测摄像机所拍摄二维图像背景的复杂性,导致对目标数量统计的误差,为此,提出了综合形状检测法。该方法在对原始图像提取封闭边缘后,用形状因子和边界在两个正交方向的对称性作为判据,根据发光体目标本身及其光线投影的几何形状差异,将发光体和投影区域分割开,只对反映发光体本身的光亮区域进行统计计数。试验测试表明,该方法的相对误差为4.2%左右。
Because of the complexity of the 2-D image background captured with a surveillance camera, general methods applied to deal with a moving luminous object in dark background will induce a count error for objects. Therefore, a method being called synthetic shape detection is described. The method first extracts the enclosed boundary of the original image, and then regards both shape-factor and symmetry at two orthorhombic directions as the criterion, furthermore differentiates between the illuminant object itself and its projection according to the difference of their geometrical shapes. Only is the lighting zone of the illuminant object itself statistically calculated. The experiments show that the relative error of the method is about 4. 2%.
出处
《吉林大学学报（信息科学版）》
CAS
2006年第1期94-98,共5页
Journal of Jilin University（Information Science Edition）
关键词
图像处理
运动枪测
目标识别
计算机视觉
image processing
motion detecting
object identification
computer vision
