Rozwój, kalibracja i walidacja lokalizatora bazującego na jednej kamerze smartfonu

W pracy przedstawiono rozwój, metody kalibracji i walidację lokalizatora medycznego opartego o kamerę urządzenia mobilnego, mającego w przyszłości zastąpić system lokalizacyjny Polaris. Opisano sposób detekcji markerów Rigid Body (RB) przy pomocy współczynnika nierówności perymetrycznej oraz kryterium powierzchni, przedstawiono wpływ pozycji markerów na obrazie kamery na dokładność odczytu pozycji, a także zaprezentowano zupełnie nowy rodzaj markerów RB, opracowanych przez firmę Aesculap, przystosowanych do pracy z urządzeniami mobilnymi. Przeprowadzono również testy potwierdzające zasadność wprowadzonych optymalizacji i rozwiązań.

---

The following report presents an alternative to the NDI Polaris localizer. This work focuses on optimization and implementation of new image processing algorithms for a medical mobile localizer and it also concerns the newest Rigid Body markers developed by Aesculap AG. It was necessary to use Samsung Galaxy S4 with Android 4.4 „KitKat” OS and Eclipse IDE version 4.3.1 with OpenCV 2.4.8 libraries to perform the development. Tests were conducted on a special plate Gold Bench with known dimensions. Camera calibration was carried out according to the method described in [4]. Within the scope of the current work there was usage of the isoperimetric quotient and the implemented criterion of the pixel area of detected objects, to eliminate detection of small light reflections as false Rigid Body markers (Fig. 5). The last software modification aimed at adding a circle on the camera view which was located in the area of the best camera accuracy (Fig. 6). These optimizations enable better detection of RB markers and more accurate estimation of their position. There is also presented the new design of the RB marker which has an additional collar enabling dissipation of smartphone LED diode light around the marker. It is crucial innovation which eliminates artifacts on camera view, caused by light reflections from RB. The results of the tests were satisfying. The distance 150 mm between two RB was measured using the smartphone camera from two different distances to Gold Bench of values 50 and 100 cm. The mean values from 10 measurements were 149,62 and 149,33, whereas the standard deviations values were σ±0,21 mm and σ±0,24 mm, respectively. It is planned to migrate the application from Android platform to Apple iOS in the future.