The vast network of bridges in the United States raises a high requirement for its maintenance and rehabilitation. The massive cost of manual visual inspection to assess the conditions of the bridges turns out to be a burden to some extent. Advanced robots have been leveraged to automate inspection data collection. Automating the segmentations of multiclass elements, as well as surface defects on the elements, in the large volume of inspection image data would facilitate an efficient and effective assessment of the bridge condition. Training separate single-task networks for element parsing (i.e., semantic segmentation of multiclass elements) and defect segmentation fails to incorporate the close connection between these two tasks in the inspection images where both recognizable structural elements and apparent surface defects are present. This paper is motivated to develop a multitask deep neural network that fully utilizes such interdependence between bridge elements and defects to boost the performance and generalization of the model. Furthermore, the effectiveness of the proposed network designs in improving the task performance was investigated, including feature decomposition, cross-talk sharing, and multi-objective loss function. A dataset with pixel-level labels of bridge elements and corrosion was developed for training and assessment of the models. Quantitative and qualitative results from evaluating the developed multitask deep neural network demonstrate that the recommended network outperforms the independent single-task networks not only in performance (2.59% higher mIoU on bridge parsing and 1.65% on corrosion segmentation) but also in computational time and implementation capability.

翻译：美国庞大的桥梁网络对维护和修复其庞大的桥梁网络提出了很高的要求。评估桥梁条件的人工直观检查的庞大成本在某种程度上是一个负担。先进的机器人已被用于自动收集检查数据。在大量的检查图像数据中,多级元素的分块以及元素表面缺陷的自动化将有助于对桥梁状况进行高效率和有效的评估。培训单级任务网络对元素分解(即多级元素的语义分解)和缺陷分解功能进行单独分析,这在一定程度上是一个负担。高级机器人被利用来自动收集检查数据数据。在检查图像中,既存在可识别的结构元素,也存在明显的表面缺陷。本文的动机是开发一个多塔级的深层神经网络,充分利用桥梁元素和缺陷之间的这种相互依存性能,以提高模型的性能和总体化。此外,对改进任务绩效的拟议网络设计的有效性进行了调查,包括特性分解、交叉交谈共享和多目标损失功能。在检查中,没有将这两项任务分级的等级标签纳入检查区段图像中,而没有在可识别的结构结构结构结构中,也没有在深度网络的高级时间网络执行能力上建立更高等级标签。推荐的模型,并演示了标准化网络业绩模型。