摘要
研究的是数码相机在交通管制中定位物体特征点的优化设计问题.将数码相机成像问题简化为针孔成像问题后,研究了靶标图像与数码相机像平面上图像间的旋转和平移关系,为了使二维图像获取三维信息,建立起图像间的透视投影模型.利用该模型,通过确立标定点与相应的二维图像坐标,运用最小二乘法求解投影矩阵,实现对相机内、外部参数的标定.标定过程中,运用数字图像处理技术,对图像进行二值化处理,提高了数据的准确度和科学可信性,由此解得较为精确的内、外部参数,获得基于此模型的圆心在像平面的坐标表示,分别为A(321,189)、B(422,196)、C(639,212)、D(582,502)、E(284,501).使用BP神经网络方法对已建立的模型进行检测,通过对比BP神经网络与透视投影求得圆心值与实际圆心值的均方根误差,论证了该模型有较高的精度和稳定性.并基于已有工作,建立了双目视觉定位模型,得到两部相机间的相对位置关系.在模型的进一步讨论中,研究了椭圆中心偏移问题,对像平面上椭圆的中心坐标进行了修定,得到实际的中心为:A(325,191)、B(427,201)、C(642,216)、D(584,504)、E(288,505).以及讨论了图形的畸变问题,确定了相应的关系.
This paper is an optimal design of locating objects with digital camera in traffic control. Digital camera imaging process is simplified as a pinhole imaging model. We establish a series of transformation target image between camera image perspective projection model and receive relevant data the parameters of the model. During this process, we use digital image processing technology, such as an image binarization processing to improve the accuracy of the results and scientific credibility. Based on this model we obtain the center coordinates compared with the BP neural network model in the image plane, namely A （321,189）, B （422,196）, C （639,212）, D （582,502）, E （284,501）. Based on existing, established the location model of the relationship between the relative position in binocular vision system.
出处
《西南民族大学学报（自然科学版）》
CAS
2009年第6期1183-1188,共6页
Journal of Southwest Minzu University(Natural Science Edition)
关键词
透视投影
图像二值化
BP神经网络
双目视觉系统
perspective projection
image binarization
BP neural network
binocular vision system
