As information technologies have advanced greatly in the recent years, the security problem of information networks becomes all the more important. As a result, biometric identification techniques have been given considerable attention. Fingerprint-related techniques, due to their desirable properties, e.g., universality, perpetuity, collectability, and particularity, are most widely applied and documented. However, in practice, collected fingerprint images are not always of good shape. They often are noisy or contain only partial fingerprints. Therefore, in this article, the authors propose a partial fingerprint comparison technique, called the triangular scheme, to combine with a coefficient of orientation deflection variation and other advancements, to obtain satisfactory matching based on only partial fingerprints. By the proposed scheme, they show that within a wide range of average reliability when the test image and a database image could have only five matched minutia points, the authors have both fault resistance rate and fault acceptance rate error values down to 29%, and the total matching accuracy reaches 71%.