Abstract: This paper designs a new set of algorithms to locate centers of artificial marks pictured from long distance in an outdoor environment, and applies it to complete the building deformation monitoring tasks. This paper aims at pixel-level edge positioning and data quality control. It firstly proposes the maximum gradient projection notation to obtain the pixel-level edge of a mark. An improved edge thinning method is then presented and the curvature is introduced for further screening. A novel pixel-level edge positioning process is formed based on the above three steps to achieve a complete and accurate thinning edge and also able to avoid interferences from outdoor condition. A data quality control mechanism is then designed based on the statistical theory after the sub-pixel center localization. Only those localization results which meet the statistical requirement are kept to participate in the final calculation. The negative effects caused by the shaking of the imaging equipment can thus be controlled. Experiments are conducted in various environment with different background and photographic ranges. The results prove the image measurement method presented in this paper shows its advantage in stable and accurate localization capability, fast processing speed, low requirement for imaging equipment, and strong adaption to outdoor environment. Using a mainstream SLR camera with a standard lens, its positioning accuracy is less than 3.5 mm within 200-meter photographic range and therefore can be used in real building deformation monitoring scenarios.

Key words: building deformation monitoring, maximum gradient projection, edge thinning, curvature, data quality control