In Phase I, we developed a practical sensor platform suitable for mounting on the back of train. A large amount of IR images associated with rail tracks were collected on an MBTA train. Algorithms were developed to identify both right and left tracks from a moving train. To help detect track cracks, we developed new algorithms for IR image enhancement and track crack identification. Algorithms were also developed to detect the rail cracks in real time. In Phase II, we will further develop new software algorithms solving some issues observed in Phase I. A smaller sensor platform will be developed for data collection on MBTA train, which incorporates a higher resolution IR camera and smaller gyro stabilizer. Using this new sensor platform, we will collect more data of both regular rail tracks and track cracks. We will then develop software algorithms for image quality improvement under various operating conditions, thermal signature variations, and discrimination between true track cracks and background clutters. In Phase II, we will work with engineers in both Vehicle Department and Track Department of MBTA. To support our Phase II effort, MBTA has provided an endorsement letter. Feedback from MBTA will be incorporated into the final system.