Asset Details
Do you wish to reserve the book?
High-density foreground object detection in optical remote sensing images via semantic fusion and box alignment
Hey, we have placed the reservation for you!
By the way, why not check out events that you can attend while you pick your title.
Oops! Something went wrong.
Looks like we were not able to place the reservation. Kindly try again later.
Are you sure you want to remove the book from the shelf?
High-density foreground object detection in optical remote sensing images via semantic fusion and box alignment
Do you wish to request the book?
High-density foreground object detection in optical remote sensing images via semantic fusion and box alignment
Please be aware that the book you have requested cannot be checked out. If you would like to checkout this book, you can reserve another copy
We have requested the book for you!
Your request is successful and it will be processed during the Library working hours. Please check the status of your request in My Requests.
Oops! Something went wrong.
Looks like we were not able to place your request. Kindly try again later.
Journal Article
High-density foreground object detection in optical remote sensing images via semantic fusion and box alignment
2024
Overview
Accuracy and effectiveness towards multiscale and dense remote sensing multivariate 2D information with object detection of bi-directional learning method remains challenging. Most methods require the design of complex network structures or bounding box loss functions, thus neglecting computational cost and training noise. To facilitate practical applications, a novel optical remote sensing of the bi-directional learning object detection (ORS-BLOD) is proposed in this paper. In the method, the positive direction mechanism contains two feature re-identification convolutional modules, which can effectively distinguish complex internal texture features and improve the accuracy of small objects. The method further designs a novel auxiliary-point balancing IoU (ABIoU) loss in the reverse direction mechanism. The novel loss not only can avoid the local optimum solutions of Euclidean distance term in single-pair points regression but also can avoid IoU loss non-converging for local aspect ratio, which can realize the stability of the loss values and the direct measure of the side length. During the training phase, ABIoU loss does not produce additional parameters and improves the accuracy of box position and the integrity of aspect ratio. mAP50 of our method can, respectively, reach 73.3%, 87.03% and 56.84% on DIOR, DIOR6 and VOC2007 object detection data sets, and the high-precision and portability of our method are revealed by extensive experiment results and analysis.
Publisher
Springer Berlin Heidelberg,Springer Nature B.V
Subject
