论文部分内容阅读
提出一种使用人工神经网络技术来估计红外图像背景的快速算法,并利用红外图像中弱小目标的特性来构建目标模型,采用中心重合的大、小两个窗口,用大窗口的外层来估计目标周围的背景,即隐含层第一个结点的输出值,大窗口内的小窗口则是用来估计中心像素的特性,即隐含层第二个结点的输出值,用隐含层第二个结点减去第一个结点的差的大小来判断中心像素是属于目标还是背景,差值越大输出值越大。采用该思想训练网络权值,可以更好地检测真目标,剔除虚假目标。
A fast algorithm for estimating the background of infrared images using artificial neural network technology is proposed. The target model is constructed by using the characteristics of weak targets in infrared images. Large and small windows overlapped by centers are used to estimate the outer windows of large infrared windows. The background around the target, that is, the output of the first node of the hidden layer, the small window in the large window is used to estimate the characteristics of the central pixel, that is, the output of the second node of the hidden layer, Layer second node minus the size of the difference between the first node to determine whether the center pixel belongs to the target or the background, the greater the difference the greater the output value. Using this idea to train network weights can better detect true targets and eliminate false targets.