Open AccessVehicle detection base on Road Segmentation for surveillance satellites video
Author(s) -
Yun Meng,
Xiaoyong Wang,
Shuai Yuan
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2010/1/012169
Subject(s) - segmentation , computer science , mixture model , pyramid (geometry) , artificial intelligence , computer vision , recall rate , precision and recall , frame (networking) , mathematics , telecommunications , geometry
With the development of data acquisition ability, the LEO satellites can work with a surveillance camera on board by focusing a particular area for dozens of seconds, so it drives us to develop some applications based on this data. In this paper, a data preparation method named DERS (Density Estimation based Road Segmentation) is proposed to divide the road from background to reduce the worse impact from relative-motion and then improve the precession and recall for vehicles for surveillance satellite video sequence. DERS has three main steps: frame difference, morphology, pyramid iteration, segmentation. Our experiments with the satellite videos from SkySat-1, and compared between Gaussian Mixture Model (GMM) and GMM with Ders shows that the precision rate and recall has improved of the traditional method with Ders, the precision of GMM with Ders has been improved from 64.42% to 88.71, the recall of GMM with Ders is from 76.21% to 74.16%, and the F-Score has been improved from 69.82% to 80.79%.