面向机柜表面缺陷检测的不均匀光照和低亮度图像增强方法

Non-uniform and low illumination image enhancement for cabinet surface defect detection

光照条件是大尺寸机柜表面缺陷检测的重要影响因素。当光照分布不均匀或光照强度不足时,采集得到的机柜表面图像质量低,造成缺陷检测误差。为此,提出一种融合卡通纹理分解和最优双曲正切曲线的图像增强方法。首先,采用导向滤波将机柜表面图像分解为卡通图和纹理图,利用高斯尺度空间理论建立光照模型,实现不均匀光照去除;其次,研究图像的双曲正切曲线性质,通过图像加权拉伸实现低亮度图像增强;最后,采用对比度、亮度和灰度方差乘积对图像增强效果进行评价,同时对增强前和增强后的图像进行缺陷检测,进行对比分析验证。实验结果表明,该方法能实现光照不均且低亮度的机柜表面图像增强,机柜表面缺陷检测的准确率显著提升,召回率提高了29%,F值提高了21%。

Illumination plays an important role in the surface defect detection of large cabinet.The quality of cabinet surface image captured in uneven or low illumination condition is poor,which may lead to defect detection error.To solve this problem,an image enhancement method is proposed by combining cartoon texture decomposition and optimal hyperbolic tangent curve algorithm.Firstly,cartoon and texture maps are separated from cabinet images using an orientation filter.The image illumination model is also formulated based on the Gaussian scale space theory,and the uneven illumination is removed.Secondly,the hyperbolic tangent curve is used to enhance the low-illumination image by the weighted stretching.Finally,the performance of the proposed image enhancement method is evaluated using the contrast,brightness and gray-scale variance product parameters.The method performance is also evaluated based on the comparison results of defect detection on the original captured image and the enhanced images.Experimental results show that the proposed method is suitable to enhance the cabinet image captured under the uneven and low illumination condition.The accuracy of defect detection on enhanced images is significantly improved.To be specific,the recall ratio is increased by 29%and the F-measure value is increased by 21%.