In response to the traditional detection methods of not detecting the target, not seeing it, blurring the outline and details of the image, etc., using a combination of infrared and polarization detection, the infrared image information and the polarization image information are decoded to solve the problem of not being detected or seen in various environments. For the target local feature extraction process of large amounts of data, slow extraction speed, and other problems, an improved SIFT algorithm for local feature extraction of polarized images with deep learning is proposed. Experimental results show that the modified algorithm combines the advantages of polarimetric imaging and deep learning to achieve fast feature extraction of targets in simple or complex backgrounds. The algorithm improves the speed of local feature extraction of polarized images by 3.6% and the accuracy of extraction by 2.8%. The algorithm provides the theoretical basis for target classification, identification, and tracking techniques.
In this paper, we propose an Att-Siam infrared cluster target detection and tracking algorithm to address the difficult problem of background distinction in the process of long-range detection and tracking of small targets in air clusters. The Att-Siam algorithm first introduces the attention mechanism in the high-dimensional Hilbert spatial data and the original spatial data, and then introduces the attention mechanism in the convolutional channels, the stacked channel attention mechanism, and the spatial attention mechanism. By improving the learning performance of the network structure, the proposed algorithm improves the tracking success rate and accuracy by 32.3% and 20.9%, respectively, when compared with the traditional SiamFC algorithm tested on the spatially weak target IR dataset. The experimental results show that the proposed algorithm can adapt to complex and diverse infrared aerial scenarios and achieve effective and stable real time tracking of small infrared aerial targets.
Aiming at the problems such as low efficiency of initial structure optimization design of traditional refractive optical system and overreliance on experience in structure selection. In this paper, an initial structure automatic optimization design method of refractive optical system based on deep learning is proposed. The structural characteristic data of the reference lens in the optical lens library are learned through supervised training. Unsupervised training model based on ray tracing is constructed to improve the generalization ability of deep neural network model. Through the network model generated by training, the optical system structure parameters including real glass are output, and the automatic optimal design of the initial structure of the refractive optical system is realized. The design results show that the initial structure spot radius of optical system in full field and full spectrum optimized by network model are close to the reference lens. The initial structure of the optical system can be designed according to different focal length requirements. The success rate of one million initial structures designed in this paper is greater than 96.403%, which indicates that the network model has good generalization ability. The method proposed in this paper contributes to automatically generate the initial structure of the refractive optical system rapidly and provides a new solution for the optimization of complex optical system.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.