Hybrid quantum-classical 3D object detection using multi-channel quantum convolutional neural network

The Journal of Supercomputing, Journal Year: 2025, Volume and Issue: 81(3)

Published: Feb. 1, 2025

A Novel Continual Learning and Adaptive Sensing State Response‐Based Target Recognition and Long‐Term Tracking Framework for Smart Industrial Applications

Expert Systems, Journal Year: 2025, Volume and Issue: 42(5)

Published: April 7, 2025

ABSTRACT Purpose With the rapid development of artificial intelligence technology, highly intelligent and unmanned factories have become an important trend. In complex environments smart factories, long‐term tracking inspection specified targets, such as operators special products, well comprehensive visual recognition decision‐making capabilities throughout whole production process, are critical components automated factories. However, challenges target occlusion disappearance frequently occur, complicating tracking. Currently, there is limited research specifically focused on developing robust frameworks for particularly those designed to integrate with embedded platforms overcome various challenges. Methods We first construct three new benchmark datasets in workshop environment a factory (referred SF‐Complex3 data), which include challenging conditions complete partial targets. A brain memory‐inspired approach used determine uncertainty estimation parameters, including confidence, peak‐to‐sidelobe ratio average peak‐to‐correlation energy, develop continual learning‐based adaptive model update method. Additionally, we design lightweight detection automatically detect locate targets initial frame during re‐detection. Finally, algorithm ground mobile robots aerial vehicles‐based imaging processing equipment build framework, Results conducted extensive tests UAV20L datasets. The proposed demonstrates performance improvement 6% when addressing key attributes, compared state‐of‐the‐art methods. was capable running efficiently platforms, UAVs, at real‐time speed 36.4 frames per second. Conclusions SFC‐RT framework effectively addresses loss within environments. meets requirements performance, robustness design, making it suited practical deployment.

Language: Английский

A Novel Multi-Sensor Nonlinear Tightly-Coupled Framework for Composite Robot Localization and Mapping

Sensors, Journal Year: 2024, Volume and Issue: 24(22), P. 7381 - 7381

Published: Nov. 19, 2024

Composite robots often encounter difficulties due to changes in illumination, external disturbances, reflective surface effects, and cumulative errors. These challenges significantly hinder their capabilities environmental perception the accuracy reliability of pose estimation. We propose a nonlinear optimization approach overcome these issues develop an integrated localization navigation framework, IIVL-LM (IMU, Infrared, Vision, LiDAR Fusion for Localization Mapping). This framework achieves tightly coupled integration at data level using inputs from IMU (Inertial Measurement Unit), infrared camera, RGB (Red, Green Blue) LiDAR. real-time luminance calculation model verify its conversion accuracy. Additionally, we designed fast approximation method weighted fusion features frames based on values. Finally, optimize VIO (Visual-Inertial Odometry) module R3LIVE++ (Robust, Real-time, Radiance Reconstruction with LiDAR-Inertial-Visual state Estimation) camera's capability acquire depth information. In controlled study, simulated indoor rescue scenario dataset, system demonstrated significant performance enhancements challenging conditions, particularly low-light environments. Specifically, average RMSE ATE (Root Mean Square Error absolute trajectory Error) improved by 23% 39%, reductions 0.006 0.013. At same time, conducted comparative experiments publicly available TUM-VI (Technical University Munich Visual-Inertial Dataset) without image input. It was found that no leading results were achieved, which verifies importance fusion. By maintaining active engagement least three sensors all times, boosts robustness both unknown expansive environments while ensuring high precision. enhancement is critical applications complex environments, such as operations.

Language: Английский