تطبیق تصاویر سنجش از دور با استفاده از آشکارساز SURF بهبود یافته و توصیف‌گر BRISK نامتغیر با جهت در محیط شبیه‌ساز توابع تبدیل مستوی
محورهای موضوعی : پردازش تصویر و ویدئو
فاطمه خلیلی
1
,
فربد رزازی
2
,
ابوالفضل حسینی
3
1 - دانشکده مهندسی مکانیک، برق و کامپیوتر- واحد علوم و تحقیقات، دانشگاه آزاد اسلامی، تهران، ایران
2 - دانشکده مهندسی مکانیک، برق و کامپیوتر- واحد علوم و تحقیقات، دانشگاه آزاد اسلامی، تهران، ایران
3 - گروه مهندسی برق- مرکز تحقیقات و توسعه فناوری‌های پیشرفته صنعت برق و الکترونیک، واحد یادگار امام خمینی (ره) شهر ری، دانشگاه آزاد اسلامی، تهران، ایران
کلید واژه: الگوریتم SURF, انطباق تصاویر, تصاویر سنجش ازدور, توصیف‌گرBRISK , فیلتر مورفولوژی.,
چکیده مقاله :
از جمله مشکلات موجود در انطباق تصاویر سنجش از دور این است که تصاویر توسط سنسورهای متنوع و در زمان‌های مختلف و با زاویه‌های انحراف متنوع گرفته شده‌اند. برای حل این مشکل الگوریتم‌هایی برای بهبود انطباق پیشنهاد شده‌اند. یکی از متداول‌ترین روش‌ها، استفاده از الگوریتم SURF (ویژگیهای مقاوم سریع) است که نسبت به تغییر مقیاس، چرخش، تغییر روشنایی و نویز تا حدودی مقاوم و به زاویه انحراف تصاویر تا حدود 45 درجه پاسخ‌گو است. اما روی هم‌افتادگی و نزدیکی نقاط کلیدی استخراج شده در این الگوریتم زیاد است و عملاً توزیع پذیری مکانی مناسبی از نقاط کلیدی را ارائه نمیدهد. این پژوهش به دنبال روش‌هایی است که نسبت به پارامترهای تابع تبدیل آفین مقاوم باشد. در مقاله حاضر از محیط شبیه‌ساز IMAS (انطباق تصویر با شبیه‌ساز توابع تبدیل مستوی) که توزیع‌پذیری مناسبی از نقاط کلیدی را پیشنهاد و به اختلاف زاویه بیشتری نسبت به SURF پاسخ‌گوست، استفاده شده است. برای یافتن مرزها و لبه‌هایی با وضوح بیشتر در تصویر از فیلتر مورفولوژی استفاده شده و برای آشکارسازی نقاط کلیدی، ایده جرم تصویر به کاربرده شده است که جهت اصلی نقاط ویژگی را مشخص و چرخش‌های تغییر ناپذیر را توصیف می‌کند. در بخش توصیف‌گر از توصیف‌گر RBRISK (نقاط کلیدی دودویی مقیاس‌پذیر مقاوم و تغییرناپذیر در برابر دوران) که نسبت به دوران پایدار است، استفاده شده است. نتایج عملی آزمایشها نشان‌دهنده آن است که روش پیشنهادی در تصاویر ماهواره میزان انطباق را تا حدود 10 درصد بهبود بخشیده و از سرعت اجرای مناسبی در کاربردهای آنلاین برخوردار است.
Remote sensing images are often captured by a variety of sensors at different times and with various deviation angles. This makes the matching procedure of image pairs be a challenge. To solve this problem, some algorithms have been proposed to improve this matching. One of the most popular methods is SURF (Speedup robust features) algorithm, which is somewhat resistant to scale changes, rotation of images, brightness variation, and noise. In addition, the algorithm is suitable for the image deviation angles up to 45 degrees. However, the overlap and proximity of the extracted key points in this algorithm are high and it does not provide a suitable spatial distribution for the key points. This study is looking for a method that is resistant to the changes of affine transformation parameters. We use an IMAS (Image matching by affine simulation) simulator environment, which offers a suitable distribution of key points and can be considered as a solution to more angle differences than SURF. A morphology filter is used to find the boundaries and the edges with more clarity in the images. To reveal the key points, the images centers of mass are employed, which address the main direction of feature points and describe the invariable rotation. In addition, RBRISK (Rotation invariant binary robust invariant scalable key point) descriptor is employed in the algorithm which is temporally stable. The results of the experiments show that the proposed method improved the matching rate in satellite images by about 10% with suitable computational complexity.
[1] G. Khademi, H. Ghassemian, "Second-order total generalized variation regularization for pansharpening", IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1-5, Dec 2020 (doi: 10.1109/LGRS.2020.3043435).
[2] A. Saboori, H. Ghassemian, F. Razzazi, "Active multiple kernel fredholm learning for hyperspectral images classification", IEEE Geoscience and Remote Sensing Letters, vol. 18, no. 2, pp. 356-360, Feb. 2021 (doi: 10.1109/LGRS.2020.2969970).
[3] S.A. Hosseini, H. Ghassemian, "Rational function approximation for feature reduction in hyperspectral data", Remote Sensing Letters, vol. 7, no. 2, pp. 101-110, Nov 2015 (doi: 10.1109/ICSIPA.2015.7412241).
[4] A. Sedaghat, H. Ebadi," Remote sensing image based on adaptive binning SIFT descriptor", IEEE Trans. on Geoscience and Remote Sensing, vol. 53, no. 10, pp. 5283-5293, Oct 2015 (doi: 10.1109/TGRS.2015.2420659).
[5] J. Zhang, W. Ma, Y. Wu, L. Jiao, "Multimodel remote sensing image registration based on image transfer and local feature", IEEE Geoscience and Remote Sensing Letters, vol. 16, no. 8, pp. 1210-1214, Aug. 2019 (doi: 10.1109/LGRS.2019.2896341).
[6] A. Ishihara, H. Aga, Y. Ishihara, H. Ichikawa, H. Kaji, K. Kawasaki, D. Kobayashi, T. Kobayashi, K. Nishida, T. Hamasaki, H. Mori, Y. Morikubo," Integrating both parallax and latency compensation into video see-througth head-mounted display", IEEE Trans. on Visualization and Computer Graphics. vol. 29, no. 5, pp. 2826-2836, May 2023 (doi: 10.1109/TVCG.2023.3247460).
[7] Z. Fang, X. Yu, J. Pan, N. Fan, H. Wang, J. Qi," A fast image mosaicking method based on iteratively minimizing cloud coverage areas", IEEE Geoscience and Remote Sensing Letters. vol. 18, no. 8, pp. 1371-1375, Aug. 2021 (doi: 10.1109/LGRS.2020.2998920).
[8] R. Grompone, G. Randull, "A sub-pixel edge detector: On implementation of the canny devernay algorithm", Image Processing on Line, vol. 7, pp. 347-372, Sept. 2017 (doi: 10.5201/ipol.2017.216).
[9] P. Moreno, A. Bernadino, J. Santos-Victor," Improving the SIFT descriptor with smooth derivative filters", Pattern Recognition Letters, vol. 30, pp. 18-26, Jan. 2009 (doi: 10.1016/j.patrec.2008.08.012).
[10] S. Jiang, U. Jzang, B. Wang, X. Zhu, M. Xiang, X. FU, X. Sun, "Registration of SAR and optical images by weighted SIFT based on phase congruency", Proceeding of the IEEE/IGARSS, pp. 8885-8888, Valencia, Spain, July 2018 (doi: 10.1109/IGARSS.2018.8519181).
[11] H. Bay, T. Tuytelaars, L. Van Gool," SURF: Speeded up robust features", Proceeding of the ECCV, vol. 3951, pp. 404-417, Berlin, 2006 (doi: 10.1007/11744023_32).
[12] A. Sedaghat, M. Mokhtarzade, H. Ebadi, "Uniform robust scale-invariant feature matching for optical remote sensing images", IEEE Trans. on Geoscience and Remote Sensing, vol. 49, pp. 4519-4527, Nov. 2011 (doi: 10.1109/TGRS.2011.2144607).
[13] S. Wany, H. You, K. Fu," BFSIFT: A novel method to find feature matches for SAR image registration", IEEE Geoscience and Remote Sensing Letters, vol. 9, pp. 649-653, July 2012 (doi: 10.1109/LGRS.201¬1.2177437).
[14] J.M. Morel, G. Yu, "ASIFT: An algorithm for fully affine invariant comparison", SIAM Journal of Ima-ging Sciences, vol. 2, no. 2, pp. 11-38, Feb. 2011 (doi: 10.5201/ipol.2011.my-asift).
[15] X. Liu, Z. Tian, Q. Lu, L. Yang, Ch. Chai," A new affine invariant descriptor framework in shearlet domain for image SAR multiscale registration", International Journal of Electronics and Commu¬nic-ations, vol. 67, no. 9, pp. 743-753, Sept. 2013 (doi: 10.1016/j.aeue.2013.03.002).
[16] S.A. Khan, Z. Saleem, "A comparative analysis of SIFT, SURF, KAZE, AKAZE, ORB, and BRISK", Proceeding of the IEEE/ICOMET, pp. 1-10, Sukkur, Pakistan, March 2018 (doi: 10.1109/ICOMET.2018.8346440).
[17] G. Jiang, L. Liu, W. Zhu, S. Yin, S. Wei, "A 181, GOPS AKAZE accelerator employing discrete-time cel¬lu¬lar neural networks for real-time feature extraction", Sensors, vol. 15, pp. 22509-22529, Sept. 2015 (doi: 10.3390/s150922509).
[18] P. Soleimani, D.W. Capson, K.F. Li, "Real-time FPGA-based implementation of the AKAZE algorithm with nonlinear scale space generation using image partitioning", Journal of Real-Time Image Processing, vol. 18, pp. 2123-2134, Dec. 2021 (doi: 10.1007/s11554-021-01089-9).
[19] W. Zhang, C. Sun," Corner detection using multi-directional structure tensor with multiple scales", International Journal of Computer Vision, vol. 128, pp. 438-459, Feb. 2020 (doi: 10.1007/s11263-019-012¬57-2).
[20] T. Gao, J. Jing, Ch. Liu, W. Zhang, Y Gao, Ch. Sun, "Fast corner detection using approximate form of second-order Gaussian directional derivative", IEEE Access, vol. 8, pp. 194092-194104, Oct. 2020 (doi: 10.1109/ACCESS.2020.3032751).
[21] B. Pan, R. Jiao, J. Wang, Y. Han, H. Hang, "SAR image registration based on KECA. SAR-SIFT operator”, Proceeding of the IEEE/CEI, pp. 114-119, Nanjiang, China, Nov 2022 (doi: 10.1109/CEI57409.2022.9950203).
[22] Z. Hossien-Nejad, M. Nasri, M. Baharlouie, "Image mosaicing based on adaptive sample consensus method and a data dependent blending algorithm", Signal Processing and Renewable Energy, vol. 6, pp. 1-12, Sept. 2022 (dor: 20.1001.1.25887327.2022.6.3.1.1).
[23] S. Leutenegger, M. Chli, R.Y. Siegwart, "BRISK: Binary robust invariant scalable key points", Proceeding of the IEEE/ICOCV, Barcelona, Spain, Nov. 2011 (doi: 10.1109/ICCV.2011.6126542).
[24] N. Mentzer, J. Mahr, G. Payá-Vayá, H. Blume," Online stereo camera calibration for automotive vision based on HW-accelerated A-KAZE-feature extraction", Journal of Systems Architecture, vol. 97, pp. 335–348, Aug. 2019 (doi: 10.1016/j.sysarc.2018.11.003).
[25] P.F. Alcantarilla, A. Bartoli, A.J. Davison, "KAZE features", Computer Vision-EECV, vol. 7577, pp. 214-227, Berlin, Sept. 2012 (doi: 10.1007/978-3-642-33783-3_16).
[26] D. Quan, H. Wei, Sh. Wang, Y. Li, J. Chanussot, Y. Guo, B. Hou, L. Jiao, "Efficient and Robust: A cross-modal registration deep wavelet learning method for remote sensing images", IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 16, pp. 4739-4754, May 2023 (doi: 10.1109/JSTARS.2023.3276409).
[27] Y. Hong, C. Leng, X. Zhang, Z. Pei, I. Cheng, A. Basu," Remote sensing image registration based on histogram of oriented local binary pattern descriptor", Remote Sensing, vol. 13, no. 12, Article Number: 2328, June 2021 (doi: 10.3390/rs13122328).
[28] W. Zhang, Y. Zhao, "SAR and optical image registration based on uniform feature points extraction and consistency gradient calculation", Applied Sciences, vol. 13, no. 3, Article Number: 1238, Jan. 2023 (doi: 10.3390/app13031238).
[29] D. Mishkin, J. Matas, M. Perdoch," MODS: FAST and robust method for two-view matching", Computer Vision and Image Understanding, vol. 141, no. 12, pp. 81-93, Dec. 2015 (doi: 10.1016/j.cviu.2015.08.005).
[30] M. Rodriguez, J. Delon, J. Michel, "FAST affine invariant image matching", Image Processing on Line, vol. 8, pp. 251-281, Dec 2018. (doi: 10.5201/ipol.2018.225).
[31] S. Jain, S. Kumar, R. Shettigar," Comparative study on SIFT and SURF face feature descriptors", Proceeding of the IEEE/ICICC, Coimbatore, India, March 2017 (doi: 10.1109/ICICCT.2017.7975187).
[32] M. Gharibi, S. Mirzakuchaki, "Improving the performance of SURF algorithm descriptors for image matching", Electronic Industries Quarterly, vol. 7, no. 1, pp. 75-88, April 2015 (in Persian).
[33] E. Rublee, V. Rabaud, K. Konolige, G. Bradski," ORB: An efficient alternative to SIFT or SURF", Proceeding of the IEEE/ICCV, pp. 2564-2571, Barcelona, Spain, Nov. 2011 (doi: 10.1109/ICCV.2011.6126544).
[34] E. Azimi, A.R Behrad, M.B. Gaznavi-Ghoushchi, J. Shanbehzadeh, "A fully pipelined and parallel hardware architecture for real-time BRISK salient point extraction", Journal of Real-Time Image Processing, vol. 16, pp. 1859-1879, Oct. 2019 (doi: 10.1007/s11554-017-0693-4).
[35] F. Khalili, H. Ghassemian, "Classification of remote sensing images using transformation methods and spatial features", Proceeding of the MVIPC, Isfahan, Iran, Nov. 2016 (in Persian).
[36] Y. Zhao, R. Hong, J. Jiang, "Visual summarization of image collections by fast RANSAC", Nearocomputing, vol. 172, pp. 48-52, Jan. 2016 (doi: 10.1016/j.neucom.2014.09.095).
[37] K. Li, Y. Zhang, Z. Zhang, G. Lai, "A course-to- fine registration strategy for multi-sensor images with large resolution difference", Remote Sensing, vol. 11, no.4, Article Number: 470, Feb. 2019 (doi: 10.3390/rs11040470).
