Abstract: Geometric measurement is one of important means controlling the manufacture of parts and the quality of assemblies. So raising the accuracy, speed and detection ratio of geometric measurement is the urgent affairs. The algorithms of geometric measurement model, image processing and line edge fitting for computer vision are introduced in this paper. Moreover, auto-adaptable image edge detection method is advanced. This paper essentially analyzes the advantage of original Max imum Hough Transform(MHT) with weakness. Base on MHT, Improved Hough Transform (IHT) is put forward to adapt for geometric measurement.IHT is make used of to proceed line edge fitting in the actual engineering. The auto-adaptable image edge detection method is applied to detect image edge in real project, and the region of image edge detected is limited within 1 pixel. IHT is also applied to fit line edge. Comparing with MHT, the result of IHT indicates its speed quick, the accuracy high, and anti-jamming-quality strong, exactly said that the accuracy of IHT fitted line is less than 0.1 pixel for polar distance and less than 0.01deg for polar angle.