This paper focuses on the development of a system to detect if a nearby car poses a collision danger to a bicycle, and to sound a loud horn to alert the car driver if a collision danger is detected. A sensing and estimation system suitable for use on a bicycle is therefore developed in order to track the trajectories of vehicles in a traffic intersection. An inexpensive solid-state low-density Lidar mounted at the front of an instrumented bicycle is used. The low angular resolution of the sensor creates many challenges. These challenges are addressed in this research by clustering based approaches for assigning measurement points to individual vehicles, by introducing a correction term with its own dynamic model for position measurement refinement, and by incorporating multi-target tracking using global nearest neighbor data association and interacting multiple model extended Kalman filtering. The tracking performance of the developed system is evaluated by both simulation and experimental results. Scenarios that involve straight driving (in all four directions) and left turning (opposing) vehicles at a traffic intersection are considered. Experimental results show that the developed system can successfully track cars accurately in these scenarios in spite of the low measurement resolution of the sensor.