3D Benchmark and Performance Evaluation for Pavement Crack Detection

作者: Ran Ji , Cheng Chen , Kelvin C P Wang

DOI:

关键词:

摘要: Computerized crack detection is the core functionality of any fully automatic pavement distress survey. Many algorithms have been developed over past few years. But absence a systematic performance evaluation method makes it very difficult to distinguish performances various algorithms. In order address this problem, benchmark based on 3D surface data introduced in paper (the benchmark). The dataset (image library) contains rich sources types data. An improved also along with benchmark. proposed uses distance transform as tolerance setting and resulting Precision-Recall curve assist decision making. This provides an intuitive way selecting better under specific circumstances user’s requirements. Four existing are benchmarked against using method. Performance comparison sensitivity analysis test results conducted. A discussion

参考文章(0)