CN-121995373-A - Mobile life detection method, system and device based on migration learning mechanism and UWB radar

CN121995373ACN 121995373 ACN121995373 ACN 121995373ACN-121995373-A

Abstract

The invention discloses a mobile life detection method, a system and a device based on a migration learning mechanism and a UWB radar, which relate to the technical field of radar signal processing and emergency rescue, and have the technical key points that the mobile life detection method and the system construct a simulation-migration-real-time reasoning technical system, generate a large amount of diversified training data with labels through high-fidelity physical simulation, and pre-train a deep neural network model by utilizing the data; after the mobile platform enters a real disaster site, the invention adopts a transfer learning technology, and the pre-training model can be quickly fine-tuned by only collecting data of a very small amount of non-living areas on site or simply calibrating the data, so that the model is suitable for the current special environment. In this way, the invention can realize high-precision mobile life detection and provide powerful support for emergency rescue work.

Inventors

ZHANG FENGYUN
YE JINYUAN
WANG FUDI
WANG MENGKUI
LUO LI
HE CHEN

Assignees

西南大学

Dates

Publication Date: 20260508
Application Date: 20251216

Claims (7)

1. The mobile life detection method based on the migration learning mechanism and the UWB radar is characterized by comprising the following steps of: S1, loading a UWB radar on a mobile platform, acquiring radar echo data by the mobile platform, and performing motion compensation on the radar echo data to construct a radar data cube; S2, performing data processing on the radar data cube, inhibiting static clutter, and generating an interested region containing potential moving targets through constant false alarm rate detection or energy detection; s3, separating a micro-motion signal from the region of interest, and extracting a vital sign signal from the micro-motion signal by using a time-frequency analysis tool; S4, constructing a simulation data set through electromagnetic simulation software and a human physiological model, pre-training the simulation data set, simulating the acquisition process of S1-S3, acquiring real data of a rescue site, performing self-adaptive fine adjustment on the pre-training model according to the real data, Inputting the vital sign signals extracted in the step S3 into a fine-tuned pre-training model, and calculating the vital sign probability, the respiratory rate and the heart rate; S5, combining the calculated distance measurement and angle measurement capacity with the UWB radar, and positioning the detected life sign; Fusing the positioning and SLAM map information, and marking the position and state of the living body on an environment map of the mobile platform.
2. The mobile life detection method based on the transfer learning mechanism and the UWB radar of claim 1, wherein the motion compensation is performed on radar echoes according to S1 mobile platform inertial measurement unit data or radar point cloud data.
3. The method for mobile life detection based on the transfer learning mechanism and the UWB radar of claim 1, wherein the S3 time-frequency analysis tool comprises discrete wavelet transform and VMD.
4. The mobile life detection method based on the transfer learning mechanism and the UWB radar of claim 1, wherein in the step S4, a simulation data set is obtained by inputting a human physiological model and radar echo data into electromagnetic simulation software, and then the simulation data set is pre-trained through a deep convolution-time sequence hybrid network to obtain a pre-trained model, wherein the pre-trained model can enable the pre-trained model to be suitable for clutter distribution and propagation characteristics of rescue sites.
5. The mobile life detection system based on the transfer learning mechanism and the UWB radar as recited in claim 1, wherein the system comprises a data acquisition module, a preprocessing module, a feature extraction module, an identification and learning module, a multi-target positioning module and an output end; The data acquisition module acquires radar echo data in the moving process by adopting a UWB radar; The preprocessing module adopts an IMU inertial sensor to implement motion compensation on radar echo data to eliminate phase errors, adopts a background cancellation and moving target indication algorithm to inhibit clutter caused by static objects or environments; The feature extraction module is used for generating an interested region containing potential moving targets through constant false alarm rate detection or energy detection, separating micro-motion signals from the interested region and extracting vital sign signals from the micro-motion signals by using a time-frequency analysis tool; the recognition and learning module constructs a simulation data set through electromagnetic simulation software and a human physiological model, and pretrains the simulation data set to obtain a pretraining model; The multi-target positioning module inputs the extracted vital sign signals to the finely tuned pre-training model, calculates the vital sign probability, the respiration rate and the heart rate, and accurately positions the multiple targets through multi-sensor data combined processing; And the output end outputs the results of the life sign probability, the respiratory rate and the heart rate calculated by the pre-training module.
6. The mobile life detection system based on the transfer learning mechanism and the UWB radar of claim 5, wherein the mobile life detection system is externally connected with a human-computer interaction terminal, the output end is the human-computer interaction terminal, and the human-computer interaction terminal displays detection results, vital sign data and a target position map.
7. The chip for implementing the mobile life detection method/system based on the migration learning mechanism and the UWB radar is characterized in that the chip is a UWB radar core plate, and the UWB radar core plate is connected with an MCU, a radio frequency front end, an antenna array, an IMU module, a high-performance embedded AI computing card and various communication interfaces.

Description

Mobile life detection method, system and device based on migration learning mechanism and UWB radar Technical Field The invention relates to the technical field of radar signal processing and emergency rescue, in particular to a mobile life detection method, a system and a device based on a migration learning mechanism and UWB radar. Background In disaster emergency rescue such as earthquake, collapse and the like, rapid and accurate detection of life signs of buried survivors is important. Traditional life detection technologies, such as optical cameras, sound wave/vibration detection and the like, have limitations, the optical cameras are influenced by visual conditions, light rays and smoke, and sound wave detection requires survivors to sound or actively strike. The ultra-wideband radar technology becomes a penetration type life detection research hot spot due to strong penetration force, high distance resolution and sensitivity to micro motions, can penetrate through a nonmetallic medium, and detects the micro motions of the thoracic cavity through a micro Doppler effect. However, when the UWB radar is used for detecting disaster sites on a mobile platform, the technical bottlenecks are faced, namely, the environment is high in complex interference, strong clutter of radar echoes in the disaster sites and platform motion interference can submerge vital sign signals, the data are few, the model generalization capability is poor, disaster scenes are many, a large number of real data training AI models are difficult to acquire, the generalization capability of the model trained by a single scene is reduced in the real scenes, false alarm and omission are caused, the mobile platform distorts signals, the platform motion causes echo signal frequency shift and phase distortion, a traditional static monitoring algorithm is not applicable, the real-time requirement is high, rescue tasks are urgent, and the system needs seconds to judge vital signs and rapidly screen multiple targets. In addition, the prior art is mostly aimed at vital sign monitoring under static and controllable environments, and lacks special optimization for mobile detection scenes and extreme environmental adaptability. The three-dimensional detection method of the underground information of the patent CN112526511A is mainly aimed at vital sign monitoring under static and controllable environments, and lacks special optimization for mobile detection scenes and extreme environment adaptability. To this end, the present invention aims to provide a mobile life detection method, system and apparatus based on a transfer learning mechanism and UWB radar to solve the above-mentioned problems. Disclosure of Invention The invention aims to solve the problems, and provides a mobile life detection method, a mobile life detection system and a mobile life detection device based on a migration learning mechanism and a UWB radar, wherein a simulation-migration-real-time reasoning technology system is built, a large amount of diversified training data with labels are generated by means of high-fidelity physical simulation, and a deep neural network model is trained in advance; after the mobile platform enters a real disaster site, a transfer learning technology is used for collecting data of a very small amount of non-living areas on site or simple calibration data, and a pre-training model is quickly fine-tuned to adapt to the current special environment, so that high-precision mobile life detection is finally achieved. In order to achieve the above purpose, the technical scheme of the invention is as follows: the invention provides a mobile life detection method based on a migration learning mechanism and UWB radar, which comprises the following steps: S1, UWB radar data acquisition and motion compensation of a mobile platform: UWB radars are carried on mobile platforms (e.g., robots) and continuously acquire echo data while moving and motion compensate them to build a radar data cube (slow time x fast time x channel); The mobile platform (such as a robot) contains inertial measurement unit data or radar point cloud data, can implement motion compensation on radar echo, and eliminates phase errors generated by the movement of the platform; s2, clutter suppression and region-of-interest generation: The background cancellation algorithm, the moving target indication algorithm and the like are applied to the data after the motion compensation, and static clutter is restrained; Generating a region of interest containing potential moving objects via constant false alarm rate detection or energy detection; S3, vital sign signal separation and feature extraction: separating a micro-motion signal from a region of interest (ROI), and extracting time domain, frequency domain and nonlinear characteristics (namely vital sign signals) related to respiration and heartbeat by using discrete wavelet transform, VMD and time-frequency analysis tools; s4, life identificat