During and after disasters, highways provide vital routes for emergency services, relief efforts, and evacuation activities. Thus, a timely and reliable assessment of disaster impacts on highways is critical for decision-makers to quickly and effectively perform relief and recovery efforts. Recently, social media has increasingly been used in disaster management for obtaining a rapid, public-centric assessment of disaster impacts due to its near real-time, social and informational characteristics. Although promising, the employment of social media for assessing disaster impacts on highways is still limited due to the inability of extracting accurate highway-related data from social media. To overcome this limitation, a systematic approach is proposed to identify highway-related data from social media for assessing disaster impacts on highways, and a case study of Hurricane Harvey in Houston, TX is employed for the demonstration. The approach is constructed through three steps: (1) building data sources for social media and highways of interest in Houston, respectively; (2) adapting the social media data to each highway through a developed mapping algorithm; (3) assessing disaster impacts through analyzing social media activities in terms of their intensity, geographic, and topic distributions. Results show that the proposed approach is capable of capturing the temporal patterns of disaster impacts on highways. Official news and reports are employed to validate the assessed impacts.