Date of Award


Degree Name

Doctor of Philosophy


Civil and Construction Engineering

First Advisor

Dr. Jun-Seok Oh

Second Advisor

Dr. Valerian Kwigizile

Third Advisor

Dr. Alvis Fong


Artificial intelligence, walkability, geographic information system, image processing, automated walk score, active transportation


Walking is considered as one of the major modes of active transportation, which contributes to the livability of cities. It is highly important to ensure walk friendly sidewalks to promote human physical activities along roads. Over the last two decades, different walk scores were estimated in respect to walkability measures by applying different methods and approaches. However, in the era of big data and machine learning revolution, there is still a gap to measure the composite walkability score in an automated way by applying and quantifying the activityfriendliness of walkable streets. In this study, a street-level automated walkability score was estimated by detailing the methodology of automatic data collection procedure through applying computer vision and artificial intelligence.

The first part of the study explores the trend of walkability measures over the past two decades by considering a comprehensive literature review. The outcome shows that there are needs for tools automatically collecting walkability data by taking advantage of recent advancement in machine learning and image processing technologies. The second part of the study identifies the existing major variables related to walkability and walk-score measures. Two objective approaches, word frequency and correspondence analysis, and one subjective approach, an analytical hierarchical process was applied to identify the potential walkability variables. The third part of the study investigates the new attributes related to walkability measures by assessing the relationship between human walking activities and surrounding visual environmental attributes along the foot-walk. Statistical analysis results showed positive correlation between human walking activities and visual environmental attributes, such as surrounding building enclosure, streetlight/pole, traffic sign and billboard, street greenery, and the enclosure of the sky view factor. The fourth part of the study evaluates the pedestrian walking experience along segments mixed with pedestrians, bicyclists and e-scooter users. Higher rating and positive walking experiences were observed along the sidewalk enabled with buffer, in addition to other walkability attributes, such as high greenery, low building density, and low bike/e-scooter density.

The final part of the study combines the existing and new walkability attributes, and employs the street-level automated walk score for the city of Kalamazoo and Arlington. Semantic segmentation technique based on convolutional neural network (CNN) algorithm, along with spatial analysis was performed to automatically the walkability attributes from the Google street view images and Geographic Information System (GIS) shapefiles. Both Kalamazoo and Arlington city showed positive relationship between the computed walk score and the walking activities along the sidewalks. It was evident from the result; the computed walk score shows higher correlation to smart city indices (i.e., Health Index, Property Value, Bike and Transit score, etc.) in comparison to existing Walk Score® for both study areas

This study presents a novel methodology to measure and develop an automated street-level walkability score; which could be readily replicable and significantly reduce the labor cost, effort, and time in comparison to other traditional walkability measures.

Access Setting

Dissertation-Open Access