Research On Intelligent Subject Autonomic Perception Method Based On Multi-physics Domain Multimodal Information Parallel Fusion And Conditional Adversarial Learning

Typical intelligent entities such as smart cars,intelligent machine tools,intelligent robots,and intelligent electronic skins provide convenience and support for the development of modern society.In the face of increasingly complex consumer demands,manufacturing processes,and complex and complex environments,traditional intelligent entities adopt The single physical domain autonomous sensing method is not sufficient to cope with the dynamic changes of this complex,high latitude,rapid change,nonlinearity and uncertainty.In order to avoid the inaccuracy of intelligent subjects in complex environments and conditions,and even the false perception results affect the subsequent intelligent ability of independent reasoning,selfdetermination,independent learning,and independent decision-making,it is necessary to collect stable and high-quality multi-modal information.By utilizing the complementary relationship and coupling relationship between multiple physical domain information,information fusion and reconstruction are used to recover and reconstruct information interference and destruction caused by complex and variable environments,and multiple modes are used to perceive multiple measures from multiple measures.Improve the accuracy and stability of perception.As the most basic and important basic problem,the self-perceived problem of intelligent subject has high academic value and application value.The main research contents of this dissertation include:Studying the multi-physics domain information acquisition method of intelligent subject,constructing information field and multi-physics domain information model for intelligent subject multi-modal information fusion problem,expounding large-capacity information collection and parallel processing under visual modal and non-visual mode The time-frequency domain analysis and processing method is proposed to optimize the machine vision imaging based on image quality.The method achieved the current best level on the Gopro and kohler test benchmarks,and the target recognition accuracy increased by 0.15 mAP.Furthermore,the multi-modal multi-mode parallel fusion model of multi-physics domain multi-modal information is constructed for multi-physics domain multi-modal information parallel fusion problem.A multi-physics multi-modal information fusion method based on improved variational self-encoder is proposed.The multi-modal information fusion was performed by infrared image and digital image,and the image semantic recognition ratio was increased by 5.1%.The multi-model autonomic perception method based on conditional adversarial learning is studied.The influence of autonomic perception on the mechanism of intelligent intelligence is analyzed.The CPS-Agent and visual cognition mechanism are used to model its autonomic perception.The domain knowledge base system for intelligent subject autonomous perception,through the conditional adversarial learning algorithm,optimizes the convolutional neural network based on hidden Markov chain,and proposes the intelligent subject multi-model autonomic perception method.In order to verify the validity of the proposed general method of intelligent subject autonomous perception,this dissertation designs an unmanned sandbox and intelligent subject autonomous sensing simulation system,and applies the above method to the sandbox.Experiments show that the method can be used in complex environments.The stable,reliable and rapid decision making,taking into account the accuracy and real-time,has a very practical significance for the resolution of autonomic perception problems such as driverless cars,intelligent machine tools,and intelligent robotic electronic skin.