The three-line structured light vision sensor is widely used in industrial field measurement due to it has many advantages such as speedy and abundant information. In order to calibrate the three-line structured light vision sensor on-field with high accuracy and high efficiency, a new calibration method based on support vector machine was proposed. Firstly, a calibration target was designed. Secondly, calibration images were captured and feature points were identified. And then sub-pixel coordinates of feature points were extracted. Thirdly, a direct mapping model according to image coordinates and three dimension coordinates of feature points was built based on support vector machine. Finally, image coordinates of the calibration points were put into the model and their three dimension coordinates could be obtained. So the three-line structured light vision sensor could be calibrated directly. Experimental results demonstrated that this direct calibration method had high accuracy; its mean absolute error was 0.021 1 mm in Y direction and 0.015 0 mm in Z direction. It concludes that this method is easy, fast and suitable for field calibration.