论文部分内容阅读
本文利用最小二乘法导出了图像局部灰值分布的二元二次多项式系数的最佳估计量，并且把剩余成份看成零均值白噪声，由此建立了二次型模型，它是对局部灰值分布的最好逼近．为了从图像序列中检测帧间变化，基于二次型模型导出了服从F（n＋3，2n－12）分布的统计假设检验模型，其中n为检测窗内的象素数目．因为只有F检验一般还不能够从变化检测中选择出运动目标，本文利用专门设计的聚类分析和补空洞技术，能使目标损失和虚警趋近于零．对于可见光和红外图像序列的实验结果充分说明了所建立的技术路线及其关键技术的有效性和兼容性.
In this paper, the least squares method is used to derive the best estimate of the bivariate quadratic polynomial coefficients of local gray value distribution in the image, and the remaining components are regarded as zero-mean white noise. Thus, a quadratic model is established, The best approximation of the distribution of values. In order to detect the inter-frame variation from the image sequence, a statistical hypothesis testing model subject to F (n + 3, 2n-12) distribution was derived based on the quadratic model, where n is the number of pixels in the detection window. Because only the F test can not select the moving target from the change detection in general, this paper makes use of the specially designed clustering analysis and hole filling technology to make the target loss and the false alarm approach to zero. The experimental results of visible and infrared image sequences fully demonstrate the validity and compatibility of the established technical route and its key technologies.