采用机器人实现水果的自动化精确采摘是现代水果采摘技术发展的基本趋势之一。但由于成像条件以及机器人自身电子电路等因素的影响, 使得所获取的图像或多或少存在大量噪声。将小波变换与双边滤波算法有机结合, 提出了一种针对水果图像中颗粒噪声的滤波算法。该算法首先对噪声图像实现小波分解, 对于各高频子图像, 给出了相似像素的基本定义并分别设计出尺寸为 5的多方向的相似像素检测窗口；然后根据检测窗口中相似像素的个数与检测窗口全部像素点个数的比值大小, 分别采用自适应双边滤波算法和改进加权中值滤波算法加以处理。最后实现小波系数重构。实验仿真结果表明, 该滤波算法性能优于双边滤波算法以及小波域双边滤波算法, 对于农产品视觉图像、基于流媒体技术设施现场采集的图像处理等领域也具有一定的参考价值。

Using automatic picking robot to pick up agricultural products is one of the basic trends of modern agricultural development. Because of the imaging conditions and the influence of such factors as robot electronic circuit, such factors make the image be polluted by some kinds of noise. Wavelet transform and the adaptive bilateral filtering algorithm are combined, and a new adaptive filtering algorithm against the noise in apple images is put forward. Firstly, the noise image is conducted wavelet decomposition, the basic definition of similar pixels is given and the similar pixels detection windows are designed so as to deal with the high frequency wavelet coefficients. Then according to the number of similar pixels in detection windows and the ratio of total number of pixels size in detection windows, the adaptive bilateral filtering algorithm and the improved weighted median filer algorithm are used to deal with the high-frequency wavelet coefficients. The simulation results show that the performance of the filtering algorithm in this paper is better than that of bilateral filtering algorithm and Bilateral filtering algorithm in wavelet transform domain. This filtering algorithm has some certain reference value for the visual images of agriculture, facilities on site collection of image processing on streaming media technology and other fields.