基于完全局部二值模式的多光谱法识别损伤苹果

目的 建立一种基于完全局部二值模式的多光谱法识别损伤苹果。方法实验搭建苹果的多光谱数据采集平台，采集了558组苹果多光谱数据。使用完全局部二值模式算法提取苹果的特征向量，再将特征向量送入支持向量机中，比较分类结果。结果通过准确率、特异度和召回率三种平均指标，在完全局部二值模式结合支持向量机分类模型下，苹果多光谱图像的25 个波段对表皮有损苹果和表皮无损苹果有很好的识别效果。并在第20 波段的识别准确率达到最高为99.63%。多光谱25个波段的平均分类准确率达到了99.110%，第20波段分类准确率最高，达到了99.632%，准确率越高，分类效果越好。结论所建立的方法可以实现有损苹果和无损苹果的高效识别，对苹果的储运和分选都有一定的意义。

Multispectral damage identification of apple based on complete local binary pattern

Objective To establish a multispectral method for identifying damaged apples based on complete local binary mode. Methods The apple multispectral data acquisition platform was built and 558 groups of apple multispectral data were collected. The full local binary pattern algorithm was used to extract the feature vectors of apples, and then the feature vectors were fed into the support vector machine to compare the classification results. Results Through the three average indicators of accuracy, specificity and recall, Under the complete local binary pattern combined with support vector machine classification model, the 25 bands of apple multispectral image have a good recognition effect on damaged apples and lossless apples. The recognition accuracy in the 20th band is 99.63%.Conclusion The established method can realize the efficient identification of damaged apples and lossless apples, and has certain significance for the storage, transportation and sorting of apples.