In this paper a real-time embedded 3D vision measurement system (a vision sensor) is realized, based on the line structured light, mainly for applications like large object reconstruction or robot obstacle avoidance. The system mainly contains a sensor and an image acquisition and processing board developed by our research team. An FPGA chip and a DSP chip are embedded in the card as the major calculation units, which make real-time computation possible. The sensor consists of a high-speed camera, a laser projector and an optical filter. The whole recognition algorithm is divided into three parts: the camera calibration, the light plane calibration and the final 3D information extraction. The first two parts are achieved with Matlab before the first use of the sensor. The last part is the process of extracting 3D information for objects. In order to achieve real-time extraction, FPGA and DSP are used to realize certain algorithms separately. Experiments show the time of processing one image using the system proposed in this paper is 9.2ms, and the reconstruction error is less than 2mm when the distance between the sensor and the object is about one meter, which is enough for large object reconstruction and robot obstacle avoidance.