苹果果梗和表面缺陷的计算机视觉检测方法研究

研究了苹果果梗与果体的识别方法和果面缺陷的查找方法.根据苹果果梗的特性,提出用分块扫描判断果梗是否存在;分析了苹果的坏损表面与非坏损表面的不同反射特性,以及不同灰度值象素点数的统计特性,找出坏损区域,并从中区分出果梗和果萼.对1 5幅无果梗的图象判断准确率为100%,对90幅果梗完好图象的准确率为88%.试验证明该方法对坏损的检测是有效的.

Study on detecting methods for apple stem and defected surface with computer vision ,

The characteristics of apple stem were studied, the existence of apple stem was detected by scanning. The different reflectance properties of defected surface and non defected surface of apples were analyzed, the statistical properties of gray value of different pixel were analyzed too. To 15 apple pictures without stem, the classification accuracy is 100%, to 90 pictures whose stems are in good condition the accuracy is 88%. This classification lasted about 1 second. the R (Red) value was used to find the suspected defected pixel, the defected area was found by region growing method and the non defected pixel (including the stem and the calyx) was discarded. Finally, the total defected area of the whole apple was worked out. The tests proved that the method for detecting defected surface was efficient.