Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automatic Defect Detection in Sewer Network Using Deep Learning Based Object Detector

Bach Ha, Birgit Schalter, Laura White, Joachim Koehler

TL;DR

The paper tackles automatic defect detection in sewer networks to reduce labor-intensive CCTV inspections. It introduces a deep learning detector based on EfficientDet-D0 trained on a new dataset totaling $14.7$ km of sewer pipes, with expert-defined bounding boxes for 9 defect types plus a connection. The authors advocate processing unrolled 2D projections of pipe interiors and employ a bounding-box detector to enable localization and size measurement, while addressing Annotation challenges via a two-round refinement and a practical running-meter evaluation metric. Results show credible detection of key defects and practical deployment considerations, though spatial defects remain challenging due to lack of depth information; the work also provides concrete guidance on annotation, evaluation, and post-processing to improve real-world utility.

Abstract

Maintaining sewer systems in large cities is important, but also time and effort consuming, because visual inspections are currently done manually. To reduce the amount of aforementioned manual work, defects within sewer pipes should be located and classified automatically. In the past, multiple works have attempted solving this problem using classical image processing, machine learning, or a combination of those. However, each provided solution only focus on detecting a limited set of defect/structure types, such as fissure, root, and/or connection. Furthermore, due to the use of hand-crafted features and small training datasets, generalization is also problematic. In order to overcome these deficits, a sizable dataset with 14.7 km of various sewer pipes were annotated by sewer maintenance experts in the scope of this work. On top of that, an object detector (EfficientDet-D0) was trained for automatic defect detection. From the result of several expermients, peculiar natures of defects in the context of object detection, which greatly effect annotation and training process, are found and discussed. At the end, the final detector was able to detect 83% of defects in the test set; out of the missing 17%, only 0.77% are very severe defects. This work provides an example of applying deep learning-based object detection into an important but quiet engineering field. It also gives some practical pointers on how to annotate peculiar "object", such as defects.

Automatic Defect Detection in Sewer Network Using Deep Learning Based Object Detector

TL;DR

The paper tackles automatic defect detection in sewer networks to reduce labor-intensive CCTV inspections. It introduces a deep learning detector based on EfficientDet-D0 trained on a new dataset totaling km of sewer pipes, with expert-defined bounding boxes for 9 defect types plus a connection. The authors advocate processing unrolled 2D projections of pipe interiors and employ a bounding-box detector to enable localization and size measurement, while addressing Annotation challenges via a two-round refinement and a practical running-meter evaluation metric. Results show credible detection of key defects and practical deployment considerations, though spatial defects remain challenging due to lack of depth information; the work also provides concrete guidance on annotation, evaluation, and post-processing to improve real-world utility.

Abstract

Maintaining sewer systems in large cities is important, but also time and effort consuming, because visual inspections are currently done manually. To reduce the amount of aforementioned manual work, defects within sewer pipes should be located and classified automatically. In the past, multiple works have attempted solving this problem using classical image processing, machine learning, or a combination of those. However, each provided solution only focus on detecting a limited set of defect/structure types, such as fissure, root, and/or connection. Furthermore, due to the use of hand-crafted features and small training datasets, generalization is also problematic. In order to overcome these deficits, a sizable dataset with 14.7 km of various sewer pipes were annotated by sewer maintenance experts in the scope of this work. On top of that, an object detector (EfficientDet-D0) was trained for automatic defect detection. From the result of several expermients, peculiar natures of defects in the context of object detection, which greatly effect annotation and training process, are found and discussed. At the end, the final detector was able to detect 83% of defects in the test set; out of the missing 17%, only 0.77% are very severe defects. This work provides an example of applying deep learning-based object detection into an important but quiet engineering field. It also gives some practical pointers on how to annotate peculiar "object", such as defects.
Paper Structure (11 sections, 7 figures, 2 tables)

This paper contains 11 sections, 7 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Sewer Maintenance
  4. DL based Vision Object Detection
  5. METHOD
  6. Dataset Acquisition and Annotation Problems
  7. Neural Network Training
  8. Problem with mAP and Network evaluation
  9. RESULT
  10. FUTURE WORK
  11. CONCLUSIONS

Figures (7)

  • Figure 1: New inspection workflow with automatic object detector integration.
  • Figure 2: Visualization of data within the pipeline.
  • Figure 3: Distribution of defect types.
  • Figure 4: Examples for how fissures could be labeled, either coarsely (top) or finely (bottom).
  • Figure 5: Differences in experts' annotations (top), and the networks' detections (bottom) of formless defects, in this case surface damages, which are marked with pink boxes
  • ...and 2 more figures