基于减法直方图算法的土壤图像阴影检测

Shadow Detection of Soil Image Based on Subtraction Histogram Algorithm

自然环境下机器视觉采集的土壤图像存在阴影,土壤图像阴影检测将消除或者减弱阴影对后一步的子图分割及土种识别的影响.为了提高土壤图像阴影和非阴影的区分度,实现阴影检测,重构了比率(α)特征,并通过高斯平滑分别获得亮度(I)和比率(α)特征直方图的2个主峰值点,缩小分割阈值搜索区间以减少减法直方图求高保留率点(F)的次数;再引入拉伸因子对2个特征的保留直方图拉伸,增大阴影和非阴影保留率差异,以获取其高保留率点F;最后,在搜索区间构建亮度(I)和比率(α)特征的64等份中位点及其对应F的2条折线的交点,即亮度(I)和比率(α)特征的F值相等点,获得交点对应的亮度(I)和比率(α)特征值为阴影检测阈值,分割土壤图像阴影和非阴影.仿真实验显示:本文算法相较于对比实验具有更好的分割精度,能自适应检测土壤图像阴影,算法是有效的.

Shadows exist on the soil images collected by machine vision in the natural environment,and their detection is a necessary preprocessing work to eliminate the influence of shadows on sub-image segmentation and further soil species identification.In order to improve the discrimination between shadows and non-shadows of the soil image,the Ratio(α)feature is reconstructed in this paper.The two main peak points of Brightness(I)and Ratio(α)feature histogram are obtained through Gaussian smoothing,and the search interval of segmentation threshold is narrowed by the main peak points and the times of computing FHRRP(F)of subtraction histograms are reduced.Then,a stretch factor is introduced to stretch the retention histograms of the two features to increase the difference of the retention rate between the shadows and the non-shadows so as to obtain the FHRRP(F).Finally,the order points in the search interval and their corresponding FHRRP(F)in the 64-division midpoints of the Brightness(I)and Ratio(α)features are used to form the intersection of the two feature polylines,that is,the point where the FHRRP(F)values of the Brightness(I)and Ratio(α)features are equal,and the corresponding Brightness(I)and Ratio(α)feature value are obtained as the detection threshold to segment the shadows and non-shadows of the soil image.The results of a simulation experiment show that the shadows and non-shadows are segmented by the improved algorithm,and their segmentation accuracy is better than the contrast algorithms.The algorithms can adaptively detect the shadow of the soil image.