Quality Scoring - Pavement Markings
- Mark Christian
Currently U.S. DOT has a retroreflectivity standard for longitudinal pavement markings. Many road authorities rely on a subjective assessment for quality and maintenance planning of non-longitudinal markings in urban areas. To help road agencies prioritize maintenance of pavement markings, our approach utilizes a visual observation method with machine learning to produce a condition score on pavement markings. The intent of the quality scoring is to enable a scalable and standardized approach to identify and prioritize pavement markings for review by subject matter experts for roadway improvement and/or maintenance.
The training dataset consists of image crops (BBox) of various pavement markings including crosswalks, stop lines, arrows and bicycle markings. Each image crop has been subjectively assessed for quality using a 1 through 4 scoring system as described below.
Quality Score | Criteria |
|---|---|
1 = Poor | Pronounced signs of defects that can significantly affect the function of the pavement marking. [~>=40% worn, missing or faded to where the crosswalk is difficult to discern]. |
2 = Fair | Pronounced signs of defects that can affect the function of the pavement marking. [~20% - 40% of crosswalk marking is worn, missing and/or faded]. |
3 = Acceptable | Minimal visible signs of defects (wear and/or fading). |
4 = Good | No visible signs of defects. [0% of marking missing] |
Quality Score | Crosswalk | Stop Line | Turn Arrow | Bicycle | Words |
|---|---|---|---|---|---|
1 |       |  
|  
|
|   
|
2
|        |   
| 
|
|
|
3 |    
|   
|  
| 
|     
|
4 |    
|  
|  
|  
|    
|