摘要
研究了水果在复杂背景下准确检测问题。针对水果在果树等复杂背景下难于进行自动化检测的困境,为了提高水果自动化采摘系统的准确率,提出了一种基于图像处理技术的水果自动检测系统。首先利用将果树彩色图像进行变换,求得红色通道图像,并依据相应的灰度图像计算图像中各像素的信息熵,对比度以及能量函数等特性。通过对这些特征反复进行筛选,提取初始的水果所在的位置,然后结合颜色特征进行最终的判别。利用贝叶斯分类器,只需少量的训练数据进行数据统计,无需复杂的训练过程。采用计算机视觉和图像处理技术,只需简单的图像采集设备即可对复杂背景下的水果进行定位操作。通过试验证明,利用水果颜色和纹理特征的结合,可以有效的实现水果的自动检测,并且具有较高的检测准确率。
A method for locating fruits in image was developed to process real-time images.The concepts of background modeling in RGB color were used,which is a novel approach to the fruit segmentation problem.In background modeling,the distributions of background colors,entropies,energies and contrast were approximated from real data.The algorithm developed for this task,a co-occurrence matrix based features was used for detection applications.The algorithm correctly identified above 90% of both red and yellow fruits.
出处
《计算机仿真》
CSCD
北大核心
2012年第4期279-281,400,共4页
Computer Simulation
关键词
水果
检测
信息熵
Fruit
Detection
Entropy
