痤疮是属于丙酸杆菌皮肤病脉的一种慢性炎症, 它危害着人体健康。虽然市面上存在痤疮识别手段, 但其仪器较大且费用昂贵, 目前尚无民用级别的痤疮识别系统投入使用。本文提出一种基于多光谱成像技术的面部痤疮识别方案, 即利用多光谱相机设备, 分别对面部正常与不同严重程度痤疮皮肤进行多光谱图像信息采集, 通过图像处理方法对采集的信息进行多光谱图像分析, 并通过光谱反演算法获取光谱信息。然后将反演出的正常和不同严重程度痤疮皮肤的反射率谱线, 与高精度光谱仪在同等实验条件下探测的谱线趋势进行对比。最后建立支持向量机(support vector machine, SVM)面部痤疮三度四级分类模型, 准确率为90%, 验证了基于多光谱成像技术对面部痤疮无创识别与分类的可行性。

Acne is a chronic inflammation of the skin veins of Propionibacterium, which endangers human health. Although there are acne identification methods on the market, their instruments are large and expensive, and there is currently no civilian-level acne identification system in use. This paper proposes a facial acne recognition scheme based on multispectral imaging technology, that is, using multispectral camera equipment to collect multispectral image information for normal and acne skin of different severity on the face, multispectral image analysis of the collected information by image processing method, and obtain spectral information through spectral inversion algorithm. The reflectance lines of normal and acne of different severities are then compared with the trend of the lines detected by the high-precision spectrometer under the same experimental conditions. Finally, a three-degree and four-level classification model of facial acne in the support vector machine was established with an accuracy rate of 90%, which verified the feasibility of non-invasive identification and classification of facial acne based on multispectral imaging technology.