[工学]摄影测量学-2-2012 1.ppt

Roberts Prewitt Canny Laplacian of Gaussian Sobel Roberts Prewitt Canny Laplacian of Gaussian Original image But we haven’t found edge segments, only edge points How can we find and describe more complex features? The Hough transform is a common approach to finding parameterised line segments (here straight lines) Hough transform 用于检测图像中直线、圆、抛物线、椭圆等 Image space Hough transform Hough transform steps 对影像进行预处理提取特征并计算其梯度方向. 将(?，?)参数平面量化，设置二维累计矩阵H(?i，?j). 边缘细化，即在边缘点的梯度方向上保留极值点而剔除那些非极值点. 对每一边缘点，以其梯度方向?为中心，设置一小区间[?-?o，?+?o]. （?，?） 取累计矩阵中备选点中的极大值点为所需的峰值点，即所检测直线的参数。 对累计矩阵进行阈值检测，将大于阈值的点作为备选点. Hough transform steps 习 题 2 1. 名词解释 Pixel, IFOV, Nyquist rate, Sinc sampling function , Resampling， Convolution kernel 2. 怎么确定数字影像的采样间隔? 3. 怎么根据已知的数字影像离散灰度值精确计算其任意一点上的灰度值？ 4. 什么是数字影像？其频域表达有什么作用？ 习 题 3 名词解释 Coordinate transformation， Epipolar line rearrangement ，Moravec operator ， LOG，Mutual correlation function 2. 什么是线特征，有哪些梯度算子可用于线特征的提取？ 3. 叙说利用Hough变换提取直线的基本思路？ 4. 相关函数产生多峰值的原因是什么？它会给相关结果带来什么影响？ * * * Parameter determination Relatively oriented double-image Contents 5.1 Image Sampling and Resampling 5.2 Point Feature Extraction 5.3 Line Feature Extraction Point Feature Extraction Point features such as definite points, corner points have high accuracy. The operator for point feature extraction is usually called the interest operator. Point feature with grey value differences Moravec operator Moravec于1977年提出利用灰度方差提取点特征的算子 r c （1）Get the interest value IV （2）If the value M is larger than a threshold given, then the point is taken as a candidate point Moravec operator take a window of 5x5 pixel with the pixel as the centre and the quadratic sums of grey value differences along the four major directions. 确定窗口大小 （3）Feature point is taken from the extremum point of candidate point Forstner operator 计算各像素的Robert’s梯度和像素（c，r）为中心的一个窗口的灰度协方差矩阵，在影像中寻找具有尽可能小而接近圆的误差椭圆的点作为特征点。 （l）C