А.V. Prozorov - Post-graduate Student, P.G. Demidov Yaroslavl State University. E-mail: alexprozoroff@gmail.com А.L. Priorov - Dr.Sc. (Eng.), Associate Professor, P.G. Demidov Yaroslavl State University. E-mail: andcat@yandex.ru

The paper presents an algorithm of a depth map preprocessing for 3D industrial television systems. This system is used as a component of simultaneous localization and mapping technique. The use of an additional depth maps pretreatment can significantly increase the accuracy of the camera positioning using a visual odometry method. At the initial stage random defects on the depth map are localized using a sliding window of a given radius. After median filtering the depth map is free from accidental defects caused by the matching errors in the RGB-D system. Next step is an adaptive thresholding of color images. The threshold value is calculated by the average intensity of the image within the defect area. Then, the edges of objects are refined using the Canny edge detector. It allows defining areas of quasi-stationary, which are determined by the actual boundaries of objects in the scene. Next, the EBM interpolation is carried out within obtained areas. The proposed algorithm increases the accuracy of camera movement calculation of 9-10% in a simultaneous localization and mapping system.

 

1. Leonard J.J., Durrant-whyte H.F. Simultaneous map building and localization for an autonomous mobile robot // Intelligent Robots and Systems\' 91.\'Intelligence for Mechanical Systems, Proceedings IROS\'91. IEEE/RSJ International Workshop. P. 1442-1447.
2. Prozorov A., Tyukin A., Lebedev I., Priorov A. Self-localization of mobile robot in unknown environment // Open Innovations Association (FRUCT). 17-th Conference. 20-24 April 2015. P. 173-178.
3. Prozorov A., Volokhov V., Priorov A. Monocular visual odometry and 3D reconstruction // Proceedings of the 15th Conference of Open Innovations Association, FRUCT 2014. Saint-Petersburg. RussianFederation. P. 112-118.
4. Lebedev I.M., Tjukin A.L., Priorov A.L., Prozorov A.V. Adaptivnaja sistema prikladnogo televidenija opredelenija prepjatstvijj na podstilajushhejj poverkhnosti dlja mobilnojj robotizirovannojj platformy // EHlektromagnitnye volny i ehlektronnye sistemy. 2015. S. 64-69.
5. Dutta T. Evaluation of the Kinect sensor for 3-D kinematic measurement in the workplace // Applied Ergonomics. 2012. V. 43. № 4.P. 645-649.
6. Smisek J., Jancosek M., Pajdla T. 3D with Kinect // IEEE Workshop on Consumer Depth Cameras for Computer Vision. 2011.
7. PoljakovA.JU.Cifrovoeobemnoetelevidenie 3DTV // Tekhnikakinoitelevidenija. 2009. № 1.
8. Kui W., Ping A., Zhao-yang Z., Hao C., Hejian L. Fast Inpainting Algorithm for Kinect Depth Map // Journal of Shanghai University. DOI: 10.3969. 2012.
9. Fisunov A.V., Levina O.M., Franc V.V., Voronin V.V., Kozhin R.A. Metod obrabotki izobrazhenijj kart glubin, poluchennykh s pomoshhju aktivnykh 3D skanerov // Materialy XI Mezhdunar. nauch.-tekhnich. foruma «Innovacii, ehkologija i resursosberegajushhie tekhnologii». Rostov n/D: DGTU. 2014. S. 1341-1346.
10. SHapiro L., Stokman D. Kompjuternoe zrenie // M.: BINOM. 2006.
11. Criminisi A., Perez P., Toyama K. Region filling and object removal by exemplar-based inpainting // IEEE Trans. Image Process. 2004. 13 (9). P. 1200-1212.
12. Zachesov A., Erofeev M., Vatolin D.Ispolzovanie kart glubiny pri vosstanovlenii fona v videoposledovatelnostjakh // Novye informacionnye tekhnologii v avtomatizirovannykh sistemakh. 2015. № 18.
13. Canny J. A Computational Approach to Edge Detection // IEEE Transactions on Pattern Analysis and Machine Intelligence. Nov. 1986. V. PAMI-8. № 6. P. 679-698.
14. Kuemmerle R., Steder B., Dornhege C., Ruhnke M., Grisetti G., Stachniss C., Kleiner A. On Measuring the Accuracy of SLAM Algorithms // Journal of Autonomous Robots. 2009. 27(4). P. 387-407.
15. Sturm J., Engelhard N., Endres F., Burgard W., Cremers D. A benchmark for the evaluation of RGB-D SLAM systems // Intelligent Robots and Systems (IROS). 2012 IEEE/RSJ International Conference. Vilamoura. 2012. P. 573-580.