苹果图像的背景分割与目标提取

Background Segmentation and Object Extraction of Apples Images

摘要:

水果的缺陷、大小和颜色差异以及光照等因素影响图像背景分割与目标提取精度。以苹果为研究对象，针对4种不同光照强度条件下采集的280幅不同姿态的苹果图像，将彩色图像的R、G、B分量进行算术运算，然后采用形态学开运算进行消噪处理，采用线性空间滤波消除锯齿状边界，采用自动阈值分割方法进行背景分割与目标提取。结果显示，203幅图像的分割偏差小于1%，占总量的72.5%；70幅图像的分割偏差大于1%而小于2%，占总量的25%；偏差大于2%的有7幅，占总量的2.5%。最大分割偏差为2.83%。

Abstract:

The defects, size, color of fruit and the lighting influence the accuracy of segmentation. In order to improve the segmentation accuracy, a combinational method was presented based on apple images processing. The R, G, B components were calculated by arithmetic operations at first. Then the arithmetic result was processed for noise cancellation by morphological opening and for smooth boundary by linear spatial filtering. After these operations, the automatic threshold method was used for background segmentation. This combinational method shows good performance to process 280 images of apples with different attitudes, size, color and defects. And these images were gained in 4 types of illumination conditions. The segmentation deviations of 203 images which are 72.5% of total images are less than 1%. The segmentation deviations of 70 images which are 25% of total images are larger than 1% but less than 2%. Only 7 images’ deviations are larger than 2%, and the maximum segmentation deviation is 2.83%.