摘要
果形是实现柚子质量检测与分级的重要特征,针对柚子果形自动检测的迫切需求,采用图像处理技术,提出了基于轮廓方向特征的柚子果形检测方法。文章首先引入纹理梯度算子,定义方向熵函数获取柚子轮廓点上的方向角度并对柚子的轮廓进行编码;建立轮廓方向角度直方图以反应轮廓像素点在不同角度级的分布情况,并作为描述柚子果形特征的方法;在检测工作中,通过计算柚子图像之间的轮廓方向角度直方图相关系数实现对柚子果形优劣的判别。经实验证明:本文算法不仅具有较高的准确率,达到95.4%,而且具有检测效率高(每幅柚子图像的平均检测时间为100 ms),操作简便,算法通用性强的优点,满足实际的生产检测要求。
The shape is one of the most important features in pomelo quality test and classification. Aiming at pomelo＇s shape inspectlon,an inspection algorithm for pomelo＇s shape based on contour directional feature is presented using image processing. The texture gradient operator is firstly introduced, directional entropy function is defined to obtain the directional angle of pomelo＇s profile points, which is used to extract and encode pomelo＇s contour. And then contour directional histogram is established to describe the distribution on shape＇s directionality as pomelo＇s shape feature. During inspecting process,pomelo shape＇s inspection can be realized by means of calculating correlation coeffi cient between contour directional histograms. The experiment proves its effectiveness, the inspecting accuracy can reach 95.4%. And every inspecting image＇s processing time is 100 ms on average. So the inspecting algorithm has the advantage of applicability, generality and convenience. That is very fit for practical application.
出处
《激光与红外》
CAS
CSCD
北大核心
2013年第6期712-716,共5页
Laser & Infrared
基金
广东省自然科学基金项目(No.S2012010010368)
2011年梅州市产业技术研究与开发资金计划项目资助
关键词
柚子果形
纹理梯度
方向熵函数
轮廓
相关系数
pomelo＇ s shape
texture gradient
directional entropy function
contour directional histogram
correlation coefficient
