This paper presents an autonomous vision-based net-recovery system for small fixed-wing unmanned aerial vehicles (UAVs). A fixed-wing UAV platform is constructed using various avionic sensors, and integrated with a flight control system and a vision system. The ground operation system consists of a vision station and ground control station that provide operation commands and monitor the UAV status. The vision algorithm to detect the recovery net and provide the bearing angle to the guidance algorithm is explained, along with the discussions on the techniques employed to improve the reliability of visual detection. The system identification process and controller are described, which enables to track given waypoints and to approach the detected net under the pursuit guidance law. Experimental results show the autonomous capabilities including take-off, waypoint following, and vision-based net recovery. The proposed technique can be an effective solution to recover fixed-wing UAVs without resorting to a complicated structure such as an instrumented landing system or expensive sensors such as a differential GPS.