摘要： 在识别图像中L型角点及计算其角度大小时，会检测出较多错误角点且角度计算误差较大。为此，提出一种融合图像边缘特征和灰度特征的Harris角度计算方法。利用高斯窗口内边缘线权重和，以边缘梯度的平均值替代边缘上的梯度，通过比值影响函数筛选出L型角点。采用加权平面近似灰度表面，使用有关像元的主曲率幅度值拟合加权，以这2种方式的角点响应函数值相等为条件，推导角度计算公式。实验结果表明，该方法可使角点误检率从24.6%降为3.3%，L型角点角度计算平均误差率从10.21%降为3.82%。
关键词:
角度计算,
Harris算法,
L型角点,
边缘梯度,
高斯滤波,
曲率
Abstract: For the problem of corner detection at L-junction in image and calculate the corner angle at L-junction, this paper proposes a new algorithm based on improved Harris to detect the corner at L-junction and calculate the corner angle at L-junction. The algorithm calculates the sum of the weight on each edge in differential coefficients window, and takes the average value of all the differential coefficients as the representative differential coefficient, and measures the principal curvature on the detected corners and through the ratio influence function identify the corner at l-junction. According to the calculated values of the response function from the above two steps, it calculates the corner angle at L-junction. Experimental results show that angle calculation algorithm based on Harris corner detection improves recognition accuracy rate from 24.6% to 3.3%, and calculates the corner angle at L-junction rate from 10.21% to 3.82%.
Key words:
angle calculation,
Harris algorithm,
L-junction,
edge gradien,
Gaussian filtering,
curvature
中图分类号: