Date of Graduation
Statler College of Engineering and Mineral Resources
Civil and Environmental Engineering
John P Zaniewski
Compaction is the last and possibly the most important phase in construction of asphalt concrete (AC) pavements. Compaction densifies the loose (AC) mat, producing a stable surface with low permeability. The process strongly affects the AC performance properties. Too much compaction may cause aggregate degradation and low air void content facilitating bleeding and rutting. On the other hand too little compaction may result in higher air void content facilitating oxidation and water permeability issues, rutting due to further densification by traffic and reduced fatigue life. Therefore, compaction is a critical issue in AC pavement construction.;The common practice for compacting a mat is to establish a roller pattern that determines the number of passes and coverages needed to achieve the desired density. Once the pattern is established, the roller's operator must maintain the roller pattern uniformly over the entire mat.;Despite the importance of uniform compaction to achieve the expected durability and performance of AC pavements, having the roller operator as the only mean to manage the operation can involve human errors.;With the advancement of technology in recent years, the concept of intelligent compaction (IC) was developed to assist the roller operators and improve the construction quality. Commercial IC packages for construction rollers are available from different manufacturers. They can provide precise mapping of a roller's location and provide the roller operator with feedback during the compaction process.;Although, the IC packages are able to track the roller passes with impressive results, there are also major hindrances. The high cost of acquisition and potential negative impact on productivity has inhibited implementation of IC.;This study applied computer vision technology to build a versatile and affordable system to count and map roller passes. An infrared camera is mounted on top of the roller to capture the operator view. Then, in a near real-time process, image features were extracted and tracked to estimate the incremental rotation and translation of the roller. Image featured are categorized into near and distant features based on the user defined horizon. The optical flow is estimated for near features located in the region below the horizon. The change in roller's heading is constantly estimated from the distant features located in the sky region. Using the roller's rotation angle, the incremental translation between two frames will be calculated from the optical flow. The roller's incremental rotation and translation will put together to develop a tracking map.;During system development, it was noted that in environments with thermal uniformity, the background of the IR images exhibit less featured as compared to images captured with optical cameras which are insensitive to temperature. This issue is more significant overnight, since nature elements are not able to reflect the heat energy from sun. Therefore to improve roller's heading estimation where less features are available in the sky region a unique methodology that allows heading detection based on the asphalt mat edges was developed for this research. The heading measurements based on the slope of the asphalt hot edges will be added to the pool of the headings measured from sky region. The median of all heading measurements will be used as the incremental roller's rotation for the tracking analysis.;The record of tracking data is used for QC/QA purposes and verifying the proper implementation of the roller pattern throughout a job constructed under the roller pass specifications.;The system developed during this research was successful in mapping roller location for few projects tested. However the system should be independently validated.
Niki Rashidi, Mohammad Hadi, "Application of computer vision for roller operation management" (2016). Graduate Theses, Dissertations, and Problem Reports. 7337.