摘要
结合货车转向架安全链脱落及丢失的照片形态特征,提出了一种基于先验统计信息的故障计算机视觉识别算法;算法流程包括图像增强,自适应阈值变换滤除无关特征,应用Hough变换提取主特征线条,依据线型特征图像水平梯度沿垂直方向的积分投影定位主特征线,基于极角约束Hough变换推出两条主特征线交点,最后根据安全链和支架的几何形态特征判读故障;算法设计应用积累大量先验统计信息作为视觉不变量,极大地提高了检测速度和可靠性。
According to the characters of images of the fault of shedding and losing of bogie safety--chain, a computer vision fault--recognition algorithm based on transcendental statistics information was proposed. The process of the algorithm include： image enhancement, using self--adapting threshold value reduce redundance, using Hough transformation extract first feature line, according to the vertical projection of the horizontal gradient perform characters line, using the improving Hough transformation with angle constraints acquire the intersection of the two feature lines, according the characters of geometry position to judge the fault. Because of analyzing plenty of transcendental statistics information to establish vision--invariant, the speed and reliability of detection was greatly enhanced.
出处
《计算机测量与控制》
CSCD
2008年第8期1169-1170,1183,共3页
Computer Measurement &Control
关键词
先验统计信息
HOUGH变换
自适应阈值
极角约束
transcendental statistics information
Hough transformation
self-- adapting threshold value
angle constraints
