The warping deformation of flexible printed circuit (FPC) would result in flexible deformation of the image acquired, which reduces the detection performance of the system. To solve this problem, this paper propose a solution based on establishing an image flexible deformation model and analyzing deformation mechanism. Firstly, the circuit skeleton is extracted after image preprocessing and is delivered to dynamic template scale space (DTSS) model for acquiring clean skeleton image without branches noise. Then, whole endpoints of previous skeleton are extracted and abnormal structure regions are located by morphological operation and local template matching. Endpoints within abnormal structure regions are removed.. Thus, the endpoints of circuit break are acquired, and circuit break defects are detected through the congruent relationship between circuit break endpoints and locations. Experimental results show that the approach proposed in this paper is robust to flexible deformation. The CDR reaches to 99.30% and FDR reaches to 2.25%, which shows the validity of this method. Compared with other methods, CDR increases 5.03% at least, and FDR decreases 8.22%, which shows the advantages of this method in practical application.