摘要
在港口智能集装箱号识别系统中,为准确识别集装箱编号,需对采集的倾斜箱号图像校正。针对传统校正中,由于干扰字符的存在导致图像校正精度低、算法复杂导致处理时间长等问题,提出一种基于分块质心和旋转投影的倾斜校正方法。水平倾斜角检测采用分块质心法,即对图像分块得到无干扰区,计算每列像素质心并用最小二乘拟合直线得到水平倾斜角,逐行扫描去除干扰字符,垂直倾斜角检测采用旋转投影法,依据设定的步长旋转并投影,统计最大投影值,对应的角度为垂直倾斜角。通过仿真分析得出:水平和垂直角度检测都能精确在2°以内且运行时间较传统方法提高了22.8%。与其它方法相比在校正精度上和校正是时间上均有优势,是一种有效可行的集装箱号校正方法。
An image tilt eorrection method of container number is proposed based on block center of mass and ro- tation projection. The block center of mass is used to detect level tilt angle, and the non-interference region of image block is obtained. The center of mass of pixel in each column is calculated, and least square fit line is used to obtain the level tilt angle, then progressive scan is carried out to eliminate interference character. The rotation projection is used to detect vertical tilt angle. The rotation and projection is carried out according to the given step length and counted maximum projection value. Corresponding angle is the vertical tih angle. Simulation results show that the de- tection precision of horizontal and vertical angle is within 2°. The method improves operation time by 22.8% compared with traditional method and has advantage in correction precision and correction time compared with other methods. It is an effective and feasible method for correction of container number.
出处
《计算机仿真》
北大核心
2017年第9期230-234,253,共6页
Computer Simulation
关键词
倾斜校正
分块质心
最小二乘法
行扫描
旋转投影
Tilt correction
Pixel center of mass
Least squares method
Line scan
Rotation projection
