With the increasing application of the vision measurement technique in the civil engineering structure health monitoring,more attention has been paid to the long-term all-weather performance of vision measurement.To explore the main error source of vision measurement technique,a new error source analysis method based on Blind Source Separation (BSS) is proposed:First, in order to construct the multi-channel signals as the input signals of the blind source separation model, Ensemble Empirical Mode Decomposition (EEMD) was used to expand the observation signal channels; then, Fast Independent Component Analysis (FastICA) algorithm was used to separate the input signals, to obtain the FastICA components; next,the correlation between each component and environmental factors such as temperature, light irradiation, etc., was analyzed to explore the error source corresponding to the principal component; finally,by using the inverse transformation of the mixed matrix obtained by the separation algorithm, the proportion of the specified error source components was calculated and the main error source of the camera measurement was determined. The error data of long-term vision measurement were analyzed by blind source separation algorithm. The results show that this algorithm has good separation effect and can effectively separate and extract the displacement error components caused by each error source. In long-term vision measurement, temperature is the primary error source.