目的 实现锂电池薄膜表面缺陷特征的有效提取。 方法 采用稀疏分解算法实现表面去噪,即通过选取合适的原子函数,在过完备字典中对含有点噪声、高斯噪声、椒盐噪声和加乘噪声背景下的缺陷图像进行稀疏分解迭代,通过观察法得到终止迭代值作为经验值,并将该经验值用于特定噪声背景下的稀疏分解终止迭代条件,得到去噪后的缺陷图像。 最后将该方法与中值滤波技术进行比较。 结果 稀疏分解的去噪性能远优于中值滤波,对锂电池薄膜缺陷有很好的还原性。 结论 稀疏分解算法能够较好地去除锂电池薄膜图像中的噪声,从而识别出锂电池薄膜缺陷。

Objective To effectively extract the defect features on the surface of lithium battery film. Methods Surface de-noising was realized by sparse decomposition algorithm, i. e. , the best atomic function was selected, and sparse decomposition iteration was conducted for defect images with point noise, gaussian noise, salt and pepper noise, as well as additive and multiplicative noise in the over-complete dictionary. The terminating iteration value was got by observation and used as the experience value as the sparse decomposition iteration termination condition for denoising under specific background noise, in order to obtain the denoised defect image. Finally, this method was compared with the median filtering technology. Results Sparse decomposition denoising showed much better performance than the median filter, and had a good recovery for defects in lithium battery film. Conclusion Sparse decomposition algorithm could well remove the noises in lithium battery film image to identify the defects of lithium battery film.