Fast Digital Image Correlation Research Articles

A New Approach to External and Internal Fingerprint Registration With Multisensor Difference Minimization

External fingerprints captured by conventional scanners are easily affected by pollution or abrasion. Optical coherence tomography can capture stable internal fingerprints located inside fingertips but with comparable unstable qualities. External and internal fingerprints have advantages and disadvantages that can be used to complement each other. Registration is the premise of the fusion of these fingerprints. Multisensor differences between external and internal fingerprints, including image style differences and deformations, make the registration challenging. This paper presents a new multisensor fingerprint registration method to align external and internal fingerprints finely for the first time. The proposed method includes two stages. First, a cycle generative adversarial network with a structural constraint is designed to unify the image style of multisensor fingerprints. It can improve the similarity of fingerprints while preserve the valley–ridge structures. Then, a two-step registration with fast digital image correlation and outlier rejection is proposed to realize global to local registration. Comparisons have been performed to demonstrate the advantage of the proposed method in registration accuracy and matching performance. Fingerprint fusion results also verify that the registration and fusion of external and internal fingerprints are conducive to supplement fingerprint information.

Fast digital image correlation using parallel temporal sequence correlation method

Digital image correlation (DIC) is a noncontact technique that is widely used for deformation measurement, but improving the calculation efficiency to achieve real-time DIC calculation has always been a big concern. A parallel temporal sequence DIC method is proposed, which chooses seed points to determine the integer-pixel displacement and applies the moving least-squares fitting technique to acquire the subpixel displacement. This method avoids traditional complex iterations and takes full advantage of the GPU parallel computing. Results of a simulation experiment and an actual experiment demonstrate the accuracy and efficiency of the proposed algorithm. The calculation speed in the simulation experiment of the proposed method achieved 463,320 POI/s, whereas the speed in the actual experiment was 432,866 POI/s, when the speed of the ICGN method was 2700 POI/s and 2074 POI/s under the same accuracy, respectively. Also, the subpixel displacement calculation made up less than 1% of the entire calculation. The computational efficiency could be further enhanced if a faster integer-pixel displacement calculation method is discovered or a parallel algorithm is used.