To study the variation law of driver's sight distance at tunnel entrance and exit, an actual vehicle test was conducted on an expressway in Chongqing (a total length of 160 km, 36 tunnels, including 4 short tunnels, 8 middle tunnels, 14 long tunnels, and 10 extra-long tunnels) in non-peak hours. Eye-tracking data, illuminance, and velocity were collected by eye-tracker, illuminance meter, and non-contact velocimetry with multi-function respectively. Based on the image illuminance extraction method, the illuminance of tunnel entrance and exit were obtained through Matlab. According to the tunnel parameters in reality, the scale model test was designed, and the driving simulation test was carried out to obtain the driver's sight distance at the tunnel entrance and exit. The support vector regression sight distance calculation model was constructed to analyze the relationship among the driver's driving sight distance, velocity, and illuminance to inspect the driver's sight distance at the entrance and exit of the tunnel and put forward some improvement measures.