| description abstract | Over the last few years, convolutional neural networks (CNNs) have been applied to detect corrosion in images. Unfortunately, the corrosion is detected in bounding boxes, without precisely segmenting the corrosion elements in irregular boundary shapes, and thus it is difficult to assess them quantitatively, such as in terms of corrosion areas and corrosion severity, which are important for engineers to evaluate the performance and condition of an inspection target. In addition, training an effective CNN model requires creating a training data set by labeling the corrosion pixels in each image, which is tedious and labor intensive. This paper presents a semantic segmentation deep learning approach together with an efficient image labelling tool for rapidly preparing large training data sets, and effectively detecting, segmenting, and evaluating corrosions in the images. The image labeling tool was developed by implementing a texture-based unsupervised image segmentation method, integrated with red-green-blue (RGB) feature-based classifier optimization. The tool enables users to construct a pixel-based corrosion classifier with small set of manually labeled images. This small set of labeled images is used for optimizing the pixel-based corrosion classifier to automatically generate corrosion segments for a large number of training images. A CNN model with semantic segmentation feature then is trained for corrosion detection and segmentation. Finally, a corrosion evaluation method is proposed for classifying each pixel of a corrosion segment into user-prescribed categories such as heavy corrosion, medium corrosion, and light corrosion. The integrated approach was tested on images collected by professional inspection engineers. The results indicated that the proposed approach is practically applicable for corrosion assessment for a wide range of industrial facilities and civil infrastructures. | |
