Camera calibration accuracy determines the precision of vision-based measurement. To address the issues of limited inclination angle detection and low calibration accuracy, this paper proposes a binocular camera calibration method for target images with large inclination angles. By clustering the geometric feature data of target marked points, the paper designs a marked point extraction algorithm without prior threshold parameters to enhance the capability of detecting target images with large inclination angles. Meanwhile, the paper uses local deformation matching of marked points to replace direct detection according to the matching correlation between the ideal target plane images without inclination angles and target images with inclination angles. In addition, in order to improve the detection accuracy of the real circle center, the projection deviation is estimated by calculating the optimal local deformation parameter. Simulation and experimental results demonstrate that the proposed calibration method is more sensitive in detecting inclination angles than the traditional method. The calibration accuracy for the simulation images is improved by up to 82%, and that for experimental calibration images is enhanced by up to 60%.