为了增强机器人单目视觉系统在黑暗环境下的环境感知能力,提高其在视觉图像序列处理过程中光流场计算的准确率,对由红外摄像机所采集的红外图像序列的预处理方法进行了研究,建立了一种空域与变换域相结合的处理方法.为平衡红外图像序列对之间的亮度差异并提高图像亮度,在空域中采用直方图均衡化方法进行处理；然后,对图像进行非下采样Contourlet变换,在变换域中利用图像的强边缘、弱边缘和噪声在不同分解尺度和分解方向上具有不同几何流的性质进行区分,分别采取保留、增强以及去除处理；最后利用调整后的系数进行图像重构.该预处理方法在平衡图像亮度的同时增强了图像纹理并减少了噪声.实验结果表明,预处理程序有效地提高了红外图像对的光流有效点识别率,该方法增强了机器人单目视觉系统在黑暗环境中对环境的感知能力.%In order to enhance the environment perception ability of the monocular vision system in the darkness and to improve the correct rate of the optical flow calculation, this paper investigated infrared image pair pretreatment method for optical flow analysis and proposed an algorithm based on processing in both in space domain and transform domain. First, histogram e-qualization algorithm was used to balance the brightness difference between the image pair in space domain. This also improved their brightness. Then was applied the nonsubsampled Contourlet transform to the images. In the transform domain, it categorized the strong edges, weak edges and the noise points according to their property of geometric manifolds. And it kept strong edges,enhanced weak edges and eliminated the noises. Finally, it implemented the images by reconstructing the new coefficients. This method not only balances the brightness but also enhances the image texture and de-noises the image. The experimental results show that the pretreatment improves the optical flow correct rate. So this method can improves the perception a-bility of monocular vision system for a mobile robot.
展开▼
