基于双目视觉与目标识别拟合的三维屏幕定位方法

A Three-Dimensional Screen Positioning Method Based on Binocular Vision and Target Recognition Fitting

为解决针对三维立体屏幕的定位问题,提出了基于双目视觉与目标识别、拟合的屏幕定位算法。首先利用边缘检测算法提取屏幕轮廓,获取屏幕在二维图像中的位置;其次利用双目视觉算法对识别出的屏幕进行深度数据采集,利用最小二乘法对采集的屏幕空间数据进行空间拟合得到屏幕的空间方程;最后计算屏幕顶点在三维空间中的位置从而确定屏幕范围,至此完成在三维空间中对屏幕的定位。实验结果表明,该方法计算得到的屏幕位置具有较高准确度。

In order to solve the 3 D-space screen position localization problem, a localization algorithm based on binocular vision, screen object detection, and spatial fitting is proposed. It first leverages the edge and contour detection to obtain the area of the subject screen in the 2 D image. The algorithm then uses binocular vision technique to collect depth data from the identified screen area and uses least square method to fit the collected spatial data to obtain the spatial equation of the screen, and finally calculates the position of the four screen corner vertices in 3 D space to locate the position of the screen. Experimental results show that the screen positions calculated by this method have a high degree of accuracy.