Remote sensing target detection method based on content awareness
A technology of content perception and target detection, which is applied in neural learning methods, instruments, biological neural network models, etc., can solve problems such as ignorance and different content information, and achieve the effect of improving accuracy and realizing high-quality detection
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0032] Such as figure 1 As shown, the remote sensing target detection network used in the embodiment of the present invention includes a feature extraction network, an RPN network, a classification module, a regression module and a content perception module; as a possible implementation mode, the remote sensing target detection network used in the embodiment of the present invention ResNet101 is used as the backbone network, and the FPN architecture is used to construct the feature pyramid as the feature extraction network, and then three horizontal anchor boxes are preset on the P3, P4, P5, P6, and P7 layers of the feature extraction network, and the feature map obtained by the feature extraction network High-quality proposal regions are given by the RPN network. The remote sensing target detection method provided by the embodiment of the present invention is based on the content awareness module, such as figure 2 As shown, it specifically includes the following steps:
[...
Embodiment 2
[0051] High aspect ratio has always been a big problem in rotating target detection. Some high aspect ratio objects will be difficult to match suitable positive samples in the RPN stage due to the fixed size of the anchor frame, which makes the high aspect ratio object Detection performance has been poor. In order to solve this problem, the embodiment of the present invention designs a loss function that can make the remote sensing target detection network pay more attention to the aspect ratio in the regression process, denoted as L wh . The basic idea of this embodiment is to make the remote sensing object detection network pay more attention to the change of the aspect ratio during the regression process by calculating the difference between the aspect ratio of the predicted frame and the real frame.
[0052] On the basis of the foregoing embodiments, the embodiments of the present invention specifically further include the following steps:
[0053] S201: The content pe...
PUM
Abstract
Description
Claims
Application Information
- R&D Engineer
- R&D Manager
- IP Professional
- Industry Leading Data Capabilities
- Powerful AI technology
- Patent DNA Extraction
Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.
© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com