Fast and robust image sequence mosaicking of nursery plug tray images  

作　　者：Suiyan Tan Xu Ma Long Qi Zehua Li 

机构地区：[1]College of Engineering,South China Agricultural University,Guangzhou 501642,China [2]College of Electronic Engineering,South China Agricultural University,Guangzhou 501642,China

出　　处：《International Journal of Agricultural and Biological Engineering》2018年第3期197-204,共8页国际农业与生物工程学报（英文）

基　　金：The authors gratefully acknowledge the financial support from the National Key Research and Development Program of China(Grant No.2017YFD0700802);the National Natural Science Foundation of China(Grant No.51505156);the Earmarked Fund for Modern Agro-industry Technology Research System(Grant No.CARS-01-43).

摘　　要：A machine vision based system to estimate the nursery plug tray sowing quantity is necessary in the rice seedling nursery.Because the super hybrid rice is of small diameter,in order to obtain better estimation results of sowing quantity,only part of the tray scene is captured by machine vision system.Therefore,nursery tray image sequence mosaic algorithm is required so as to obtain the whole nursery tray sowing performance.In this paper,a fast nursery plug tray image mosaic technique based on Phase Correlation and Speeded up Robust Features(the PC-SURF)algorithm was introduced.To reduce huge computational complexity in feature extraction and avoid wrong match points from non-overlapping regions between two adjacent images,firstly the proposed method used Phase Correlation to approximately locate the overlapping regions between two sequential images so as to narrow down the processing area for further feature extraction.Then,the SURF algorithm was implemented in the overlapping region of the images to perform image registration and image blending.Image registration and blending were then performed using the RANSAC algorithm and transition smoothing method.Finally,sequential images were warped into a single frame to produce a panoramic tray image with high resolution and better quality.The test results showed that the PC-SURF algorithm greatly outperforms the SURF algorithm in point matching accuracy,time consumption,and image mosaic accuracy.The average feature point matching accuracy of the PC-SURF algorithm was improved by approximately 7.14%.The implementation time was almost three times faster.The Root Mean Squared Error(RMSE)of image blending in the R,G,B channels(RMSEr,RMSEg,and RMSEb)decreased by approximately 8.4%,6.9%,6.9%respectively.The RMSE of image registration RMSEreg decreased by 33.8%.In addition,the speed of the nursery tray image mosaic technology could meet the requirements for real-time implementation in super hybrid rice automated sowing machines.

关 键 词：super hybrid rice nursery plug tray image mosaic phase correlation SURF algorithm 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]