Object Detection and Trcacking Based on Convolutional Neural Networks for High-Resolution Optical Remote Sensing Video

Object Detection and Trcacking Based on Convolutional Neural Networks for High-Resolution Optical Remote Sensing Video

Table of Contents




Abstract

Object detection algorithms, from high-resolution optical remote sensing images, have been booming from the last few years. However, object tracking for high-resolution optical remote sensing video is a challenging task due to the large number and small size of objects. In this paper, we propose an object detection and tracking method based on deep convolutional neural networks for wide-swath high-resolution optical remote sensing videos. The proposed method firstly segments each frame of a video into sub-samples using a sliding window of fixed size. In order to detect the objects appearing at the edge of the sliding window efficiently, we use an overlapping sliding window sampling method. Further, we design a network fusing region of interests (ROIs) of the previous and current frames to track the objects that occurred in the previous frames of the video. RoIs of the previous frame is applied directly to the feature layer of the current frame. Finally, for each frame, we merge the detection and tracking results of sub-samples by the non-maximum suppression (NMS) method. The experimental results on our dataset demonstrate the validity and generality of the proposed detection algorithm.

  • Author Keywords

    • Object detection,
    • object tracking,
    • high resolution optical remote sensing,
    • deep convolutional neural network
  • IEEE Keywords

    • Remote sensing,
    • Object detection,
    • Proposals,
    • Optical imaging,
    • Optical sensors,
    • Classification algorithms,
    • Convolutional neural networks

 

Introduction

Object detection and tracking have been one of the most important applications of remote sensing. Along with the development of sensor technology and the aerospace industry, the resolution of remote sensing data has increased dramatically. In the high resolution of remote sensing data, the contours and the textures of objects are clearly observed and informative for object detection and tracking, and object architectures are closer to characteristics of human ocular perception than SAR images. Object detection and tracking for optical remote sensing data become more significant. Object detection and tracking is an important branch in the field of computer vision, and many mature algorithms are proposed. Previous works on object detection and tracking can be divided into two categories, for natural videos and remote sensing videos. In the field of remote sensing, remote sensing data brings some new challenges in object detection and tracking using deep convolutional neural networks, e.g. complex background information, small scale, and multi-oriented property of objects. A lot of techniques [1-2] have been presented to detect and track remote sensing objects. However, most of the detection and tracking methods for remote sensing data are based on moving objects and the size of the videos is small.

In order to detect and track static and moving objects together in the wide swath high-resolution optical remote sensing videos, we proposed a detection and tracking method based on an object detection algorithm for natural images. The object detection algorithms for natural images can also be divided into two categories, two-stage algorithms, and one stage algorithms. In the first category, the detection algorithms [3-4] are based on region proposal methods. These algorithms firstly get a region of interests by a region proposal method and then get detection results of the proposals by a classifier. For example, Grishick et al proposed a fast and precise detection algorithm called Faster R-CNN [4], which includes a region proposal network (RPN) for generating region proposals and a network using these proposals to detect objects. In the second category, the detection algorithms [5-6] have only one stage without region proposal module. For example, Yolo [6] uses only one network to achieve the region proposal task and classification task. By comparison, one stage algorithms have faster detection speed, and two-stage algorithms can obtain better detection results for small objects.

Because the size of objects in high-resolution optical remote sensing videos generally is small, our detection and tracking method is based on Faster R-CNN. We firstly segment each frame of a video into a sub-sample in order to adapt videos with any size. Secondly, these sub-samples are fed into the network, where ROIs of the previous and current frames can be fused to get the detection and tracking of results of these sub-samples. Finally, we get the detection and tracking results for the frames by merging the results of subsamples of each frame. The experimental results show the proposed method can obtain better results for wide-swath high-resolution optical remote sensing videos.

conclusion

This paper proposes an efficient object detection and tracking method for a wide-swath high-resolution optical remote sensing video. The proposed method aims to detect and track objects from high-resolution optical remote sensing videos. Our method firstly segments each frame of a video into sub-samples, so we do not need to worry about the challenges about the size of the data. The method tracks the objects by fusing the ROIs of the current frame and the previous frame. The experimental results show its ability to solve the object detection problem effectively and accurately.

About KSRA

The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.

KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.

Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.

The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.

FULL Paper PDF file:

Object Detection and Tracking Based on Convolutional Neural Networks for High-Resolution Optical Remote Sensing Video

Bibliography

author

B. Hou, J. Li, X. Zhang, S. Wang, and L. Jiao

Year

2019

Title

Object Detection and Tracking Based on Convolutional Neural Networks for High-Resolution Optical Remote Sensing Video

Publish in

IGARSS 2019 – 2019 IEEE International Geoscience and Remote Sensing Symposium, Yokohama, Japan, 2019, pp. 5433-5436

Doi

10.1109/IGARSS.2019.8898173

PDF reference and original file: Click here

 

 

+ posts

Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.

Website | + posts

Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.

Website | + posts

Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.