Machine vision is widely used in industry for non-contact dimensional measurement. However, existing point correspondence reconstruction methods suffer from low speed and low efficiency as considerable amount of time is required to process large amount of point data. In this paper, a fast measurement method based on feature line reconstruction of stereo vision is proposed. Under proposal, a few pairs of image lines are extracted from manufacturing part images acquired from vision system. Then, three-dimensional space contour lines are reconstructed based upon the developed technique. Subsequently, target measurement can be calculated directly based on the matched image lines and camera matrices. Dwelling on feature lines of target object instead of feature points, the proposed method is capable of fast and efficient data processing for real-time measurement and around three times faster than existing point-based method while maintained similar level of accuracy. The proposed method is illustrated through experiments on measuring dimensions of a plaster model within laboratory and hot forgings in the workshop. The proposed method can be applied and integrated in existing vision system for hot manufacturing part measurement and is of practical importance for industrial real-time measurement.