中文摘要：

工业生产中对介质阻挡放电（dielectric barrier discharge，DBD）均匀性的要求与实验室存在明显不同，较大时间尺度上实现的均匀放电在实际应用中也具有重要价值。基于高斯混合模型，提出一种能够较为准确地识别丝状和均匀DBD的灰度概率模型，通过采用信赖域算法对模型参数进行估计，并通过模型检验及应用实例对模型的有效性和实用性进行验证。研究结果表明：丝状放电的灰度直方图（gray level histogram，GLH）从双高斯概率分布，而均匀放电的GLH则服从单高斯分布。随着曝光时间的逐渐增大，由双高斯概率模型计算得到的丝状放电中放电区域和背景区域的灰度平均值和标准差均逐渐增大，由单高斯概率模型计算得到的均匀放电的灰度平均值也逐渐增大，但标准差基本保持不变。空气气隙间距的减小使GLH由均值逐渐变大的双高斯分布转化为均值逐渐变小的单高斯分布，表明放电产生由丝状放电向均匀放电的转化，丝状放电强度逐渐增大，而均匀放电强度逐渐减小。以上结果表明，所建模型可以有效地对丝状和均匀放电进行定量识别。

英文摘要：

The requirement of the uniformity of dielectric barrier discharge （DBD） used in the industry is much different from that in the laboratory, the uniform discharge in a long time-scale may meet the needs of some industrial applications. A gray probability model based on the Gaussian mixture model was proposed to identify the filamentary and uniform DBD accurately, the parameters of the model were estimated by a trust region algorithm, and its effectiveness and practicability were verified by validation of the model and application examples. The gray level histogram （GLH） obeys a double gaussian probability distribution in a filamentary discharge, while the GLH obeys a normal Gaussian probability distribution in a uniform discharge. With an increase of the exposure time, the mean gray level and the gray level standard deviation of the discharge and the background region in a filamentary discharge calculated by＇ the double Gaussian probability model become larger, and the mean gray level calculated by the normal Gaussian probability model also become larger, while the gray level standard deviation remain almost constant in a uniform discharge. With a decrease of the air gap spacing, the GLH of the corresponding discharge image transits from a double Gaussian probability distribution with its mean gradually increasing into a normal Gaussian probability distribution with its mean gradually decreasing. It indicates that the transition of the filamentary discharge into the uniform discharge, the filamentary discharge becomes more intense, while the uniform discharge gets less intense. These results show that the model can quantitatively identify the filamentary and uniform discharge.