Multi-sensor fusion life detection positioning system and positioning method

Abstract

The invention relates to the field of emergency rescue, disaster search and rescue and life detection and positioning, and discloses a multi-sensor fusion life detection and positioning system comprising a sensor array module which is composed of a plurality of different types of sensors and is used for collecting environmental data in a target area; the invention also discloses a multi-sensor fusion life detection positioning method, which comprises the following steps: S1, deploying a multi-type sensor array in a target area, and collecting environment data of the target area; s2, performing noise filtering and signal enhancement processing on the original data acquired by the sensor, and extracting key features of the life body; through adaptive multi-modal fusion, deep learning feature recognition, nonlinear TDOA optimization and reinforcement learning path planning, the problems of failure of a single sensor, low signal-to-noise ratio misjudgment, complex environment positioning errors, static path planning lag and the like are solved, the life detection precision and rescue response efficiency are improved, and the method is suitable for an extreme disaster environment.

Description

Technical Field

[0001] The present invention relates to the fields of emergency rescue, disaster search and rescue, and life detection and positioning, and specifically to a multi-sensor fusion life detection and positioning system and a positioning method. Background Art

[0002] In the field of emergency rescue and disaster search and rescue, how to quickly and accurately detect and locate life is one of the key technologies for rescue operations. After sudden disasters such as earthquakes, mining disasters, and fires, trapped people are often in extremely complex environments. They may be buried under rubble, exposed to high temperatures or thick smoke, or even have limited mobility due to lack of oxygen or injuries. In this case, traditional manual search methods are inefficient and difficult to detect life in a timely manner. There is an urgent need to use advanced detection technology to improve rescue efficiency.

[0003] Existing life detection technologies usually rely on a single type of sensor, such as infrared thermal imaging sensors, radar detectors, acoustic sensors, etc. These sensors can detect life signals under certain environmental conditions, but in complex rescue environments, they are often affected by various factors, resulting in reduced detection accuracy.

[0004] Infrared detection technology relies on the temperature difference between living things and the environment, and determines the target location by detecting thermal radiation. However, in a fire or high-temperature environment, background heat sources may cause infrared detectors to misjudge, making it difficult to distinguish real life signals. In addition, infrared detection is easily affected by obstacles, and it is difficult to obtain effective life signals when trapped people are completely buried.

[0005] Acoustic detection technology uses microphones or ultrasonic sensors to collect the breathing, heartbeat or knocking sounds of living things. This technology can effectively detect weak life signals in a quiet environment, but in noisy rescue sites (such as rescue machinery operation, wind noise, building collapse, etc.), environmental noise will seriously interfere with the acoustic detection signal. In addition, trapped people may be unable to actively send out sound signals due to coma or physical exhaustion, making the acoustic detection method ineffective in some cases.

[0006] Radar detection technology uses electromagnetic waves to penetrate obstacles to identify the presence of life and analyzes it based on the signal reflection characteristics. This method is suitable for walls or ruins of a certain thickness, but in complex multipath environments, the reflection and attenuation of radar signals may lead to false targets or positioning errors. In addition, metal materials may block radar signals, affecting detection results.

[0007] Based on the detection of life forms, existing life form positioning methods mainly rely on technologies such as time difference ranging (TDOA), received signal strength (RSSI) and triangulation. These methods can provide relatively accurate positioning information in open environments, but in complex scenes such as building ruins and mines, signal propagation is affected by obstacles, multipath reflection, etc., resulting in a significant increase in ranging errors. In addition, the signal propagation speed may be affected by changes in the environmental medium, such as high humidity, smoke, and changes in air pressure, which further reduces the stability and reliability of existing positioning technologies.

[0008] Rescue path planning is an important part of the life detection and positioning system. Existing path planning methods are mostly based on static maps for calculation, such as the Dijkstra algorithm, the A* algorithm, etc. These methods can calculate the shortest path in theory. However, in a real disaster environment, the rescue team's route may be affected by environmental changes at any time, such as new collapses, fire spread, toxic gas diffusion, etc., making the original path unusable. Since traditional path planning methods lack dynamic adjustment capabilities, once an emergency occurs, manual re-planning of the path is often required, resulting in delays in rescue time and affecting rescue efficiency.

[0009] In general, the existing life detection and positioning technologies still have obvious limitations in actual rescue applications. Single sensor technology is easily restricted by specific environmental factors, which affects the stability of detection; the accuracy of existing positioning methods decreases in complex environments, making it difficult to meet the needs of high-precision rescue; path planning lacks dynamic adjustment capabilities and is difficult to adapt to the rapid changes in the rescue site environment. These problems restrict the practicality of life detection and positioning systems in disaster relief, and there is an urgent need to provide more adaptable and efficient technical solutions to address the special challenges of complex environments, so as to improve the overall capabilities of detection, positioning, and rescue planning. Summary of the invention

[0010] In view of the shortcomings of the existing technology, the present invention provides a multi-sensor fusion life detection and positioning system and a positioning method, which solve the problems of existing life detection and positioning systems in complex environments due to the limitations of a single sensor, a fixed feature fusion method, positioning accuracy affected by occlusion, and a lack of real-time optimization of path planning.

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-sensor fusion life detection and positioning system, comprising: A sensor array module, which is composed of multiple sensors of different types and is used to collect environmental data in the target area; A data preprocessing module, which is used to filter out noise and enhance signals of the data collected by the sensor, and extract key features of vital signs therefrom; A feature fusion analysis module, which is used to weightedly fuse the features from each sensor, use a deep learning model to evaluate whether the signal source is a living organism, the deep learning model includes but is not limited to a convolutional neural network and a recurrent neural network, and automatically adjust the weight of the fusion feature; A positioning algorithm module, which is used to calculate the preliminary position of the life form based on the time difference ranging data of the signal source and the triangulation method; A result verification module is used to evaluate the credibility of the positioning results; The emergency command center module is used to receive the final positioning results and update the rescue path planning in real time, guiding the rescue team to quickly locate and rescue.

[0012] Preferably, the sensor includes but is not limited to an infrared sensor, an acoustic sensor, a gas sensor and a vibration sensor.

[0013] Preferably, the data preprocessing module includes: A noise filtering unit, which is used to remove noise from environmental factors, including but not limited to wind speed and electromagnetic interference; A signal enhancement unit, which is used to increase the strength of weak vital sign signals; The feature extraction unit is used to extract key features such as body temperature, breathing rate, gas concentration and vibration frequency of living organisms from the signal.

[0014] Preferably, the feature fusion analysis module includes: A feature fusion unit, which is used to perform weighted fusion of features obtained from different sensors, so that the system can automatically adjust the fusion weight according to the reliability and information quality of different sensors; The signal recognition unit is used to compare the fused features with the life data database using a deep learning algorithm to determine whether the signal source is a living organism.

[0015] Preferably, the positioning algorithm module includes: TDOA calculation unit, which is used to estimate the position of the target through the time difference between sensors; A triangulation unit is used to combine the known positions of multiple sensors and the received signal propagation time differences to optimize positioning accuracy through triangulation.

[0016] Preferably, the TDOA calculation unit uses a time difference ranging formula to estimate the position of the target, and the formula is: in, For the signal to reach the sensor The time difference, is the actual position of the target life form, For the The known locations of the sensors, is the signal propagation speed, For sensor The receiving time.

[0017] Preferably, the result verification module includes: A credibility evaluation unit, which is used to evaluate the credibility of the positioning result based on the signal-to-noise ratio and positioning error factors of the signal; The credibility judgment unit is used to judge whether the credibility is higher than a threshold.

[0018] Preferably, the credibility evaluation unit uses a credibility calculation formula to evaluate the credibility of the positioning result, and the formula is: in, is the signal-to-noise ratio, is the actual position of the target life form, is the estimated position calculated by the algorithm, is the adjustment coefficient, which controls the impact of position error on credibility.

[0019] Preferably, the emergency command center module includes: A path planning unit, which is used to calculate and optimize the rescue path in real time based on the positioning results; A real-time feedback unit is used to adjust the path planning in real time according to the on-site rescue progress data.

[0020] The present invention also provides a multi-sensor fusion life detection and positioning method, comprising the following steps: S1. Deploy a multi-type sensor array in the target area to collect environmental data of the target area; S2, performing noise filtering and signal enhancement processing on the raw data collected by the sensor to extract the key features of the living organism; S3, weighted fusion of feature information extracted by each sensor, and use of deep learning model to analyze whether the signal source is a living organism; S4, calculating the preliminary position of the target life form based on the time difference ranging and triangulation method of the signal source; S5. Conduct credibility assessment on the initially calculated positioning results to determine the validity of the results; S6. If the credibility is higher than the preset threshold, the positioning result is confirmed to be valid and sent to the emergency command center to guide the rescue team to perform the task; if the credibility is low, return to continue monitoring.

[0021] The present invention provides a multi-sensor fusion life detection positioning system and positioning method. It has the following beneficial effects: 1. The present invention proposes an adaptive weight adjustment mechanism for multimodal sensor data, which, combined with the differentiable feature weight optimization method of the deep learning model, can dynamically adjust the weights of different sensor features in the fusion calculation, so that the system can adapt to signal changes in different disaster environments (such as high temperature, strong noise, and low visibility). Compared with the existing fixed weight fusion or simple feature splicing methods, the differentiable feature weight self-optimization strategy of the present invention can effectively solve the problems of single sensor signal failure and high misrecognition rate in specific environments, thereby improving the robustness and generalization ability of life recognition, and is suitable for large-scale unstructured disaster sites.

[0022] 2. The present invention proposes a feature clustering neural network that combines time series and spatial distribution characteristics. It optimizes the clustering distribution of target features through a spatiotemporal constraint loss function, effectively enhancing the ability to recognize life signals in a low signal-to-noise ratio environment. Compared with the existing single-frame feature extraction and simple time series matching methods, the cross-time frame feature joint learning mechanism of the present invention solves the problem of incorrect recognition caused by short-term signal fluctuations and noise pollution, ensuring that life and environmental noise can still be reliably distinguished under extremely low signal-to-noise ratios (such as buried pressure, mining disasters, and high background noise environments).

[0023] 3. The present invention innovatively proposes a TDOA positioning correction algorithm that integrates nonlinear multi-objective optimization. Combined with the dynamic correction parameters of the signal propagation model, the system can optimize the positioning calculation in complex occlusion environments (such as collapsed buildings and underground spaces). Compared with the existing static TDOA or single-path propagation models, the nonlinear multi-objective adaptive optimization algorithm of the present invention solves the positioning deviation caused by signal reflection, attenuation, and environmental noise, so that high-precision life positioning can be achieved even under non-line-of-sight (NLOS) signal propagation conditions, providing more reliable location information for disaster relief.

[0024] 4. The present invention proposes a dynamic optimization strategy for disaster rescue paths based on reinforcement learning (RL). By constructing an environmental feedback reward function, the system can adaptively adjust the rescue path in different disaster sites (such as earthquake ruins, fire spreading areas, and mine collapses). Compared with the existing fixed path planning or preset weight path optimization methods, the reinforcement learning path dynamic optimization mechanism of the present invention solves the problem of rescue failure or delay caused by sudden changes in the environment (such as increased obstacles and path blocking), ensuring that rescue operations can always be executed on the optimal path in the case of dynamic changes at the disaster site, thereby improving the response speed and success rate of life rescue. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a system framework diagram of the present invention; Figure 2 is a flow chart of the system of the present invention; Figure 3 It is a framework diagram of the data preprocessing module of the present invention; Figure 4 This is a framework diagram of the feature fusion analysis module of the present invention; Figure 5 It is a positioning algorithm module framework diagram of the present invention; Figure 6 It is a result verification module framework diagram of the present invention; Figure 7 It is a module framework diagram of the emergency command center of the present invention; Figure 8 The figure is a flow chart of the positioning method of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings of the present invention specification to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] Please see attached Figure 1 -Attached Figure 2 The embodiment of the present invention provides a multi-sensor fusion life detection and positioning system, including: A sensor array module, which is composed of multiple sensors of different types and is used to collect environmental data in the target area; Sensors include, but are not limited to, infrared sensors, acoustic sensors, gas sensors, and vibration sensors.

[0028] Specifically, the sensor array module of the present invention is composed of multiple different types of sensors, the purpose of which is to comprehensively collect environmental data in the target area. Each sensor has different working principles and characteristics. Through the complementarity of multiple sensors, more accurate and comprehensive life detection and positioning results can be provided. In this embodiment, the sensor array module includes but is not limited to infrared sensors, acoustic sensors, gas sensors and vibration sensors. These sensors can work effectively under different environmental conditions and complement each other to improve the detection accuracy and anti-interference ability of the system.

[0029] In this embodiment, the structure design and working mode of the sensor array module are as follows: Infrared sensor: Infrared sensor is used to detect the temperature change of living things. Infrared sensor can effectively capture the temperature difference, especially in the ruins, fire and other scenes, the temperature of living things is obviously different from the temperature of the surrounding environment. The signal strength of infrared sensor is affected by distance and environmental interference, so compensation and signal enhancement processing are required in complex environments.

[0030] Acoustic sensor: Acoustic sensors are used to monitor physiological signals such as breathing and heartbeat of living organisms. Acoustic sensors use ultrasonic or sound wave technology to detect the presence of living organisms through the reflection and transmission time difference of sound waves between the sensor and the target. The intensity and frequency changes of acoustic signals can provide information about the location and activity status of the target. Acoustic signals are often interfered by environmental noise (such as wind noise, mechanical noise, etc.), so it is necessary to improve the recognition rate of signals through noise filtering and signal enhancement technology.

[0031] Gas sensor: Gas sensors are used to detect gas components emitted by living organisms, such as carbon dioxide and ammonia. Living organisms release specific gas components when breathing. Gas sensors can capture these gas molecules in the air and identify the activities of living organisms. Gas sensors determine the presence of living organisms by measuring changes in gas concentration. Gas sensors are easily affected by environmental factors such as air flow and humidity during detection. Therefore, data needs to be filtered and compensated in the system.

[0032] Vibration sensor: Vibration sensor is used to capture the weak vibrations of the ground or target surface caused by the activities of living things. In environments such as ruins and mines, the activities of living things will generate tiny vibration signals. Vibration sensors can accurately capture these signals in the sensor array and further analyze them. The signals of vibration sensors are usually affected by the surrounding environment (such as wind, vibration of mechanical equipment, etc.), so signal filtering technology must be used to reduce false detections.

[0033] In this embodiment, the sensor array module works together between sensors. After collecting environmental data, the sensor array module processes different types of sensor signals through the data preprocessing module to remove noise, enhance signals, and extract key features of living organisms.

[0034] The data from each sensor is analyzed through weighted fusion, and a deep learning model (such as convolutional neural network CNN or recurrent neural network RNN) is used to evaluate whether the signal source is a living organism. The weights of the fusion features are automatically adjusted to provide high-precision life detection judgment.

[0035] The advantage of the sensor array module is that the data from multiple types of sensors complement each other and can more accurately identify the presence of life in complex environments, thereby improving the accuracy and reliability of the system.

[0036] As an option, the sensor array module in the present invention can flexibly select the configuration of sensors according to the needs of different environmental scenarios. For example, in scenarios such as mines or ruins, a combination of vibration sensors and acoustic sensors can be deployed first, while in high-temperature environments such as fires, the configuration of infrared sensors will be more important. In addition, gas sensors can provide supplementary information based on changes in gas concentration on site to further optimize the positioning capability of the system.

[0037] Specifically, the design of the sensor array module takes into account the characteristics of multi-sensor fusion. The signal characteristics of each sensor are weighted during the data fusion stage to ensure that the contribution of different sensors is dynamically adjusted according to their data quality. This solution can effectively cope with the challenges of environmental interference and complex background noise while improving the accuracy of life detection.

[0038] In one possible implementation, the working mode and parameter configuration of each sensor can be adjusted in real time through the sensor array module to ensure that the system can maintain high detection accuracy and response speed under different environmental conditions.

[0039] Please see attached Figure 3 , a data preprocessing module, which is used to filter out noise and enhance signals of the data collected by the sensor, and extract key features of vital signs from them; The data preprocessing module includes: A noise filtering unit, which is used to remove noise from environmental factors, including but not limited to wind speed and electromagnetic interference; A signal enhancement unit, which is used to increase the strength of weak vital sign signals; The feature extraction unit is used to extract key features such as body temperature, breathing rate, gas concentration and vibration frequency of living organisms from the signal.

[0040] Specifically, the data preprocessing module is a key component of the present invention, which is responsible for the preliminary processing of the environmental data collected by the sensor. Its main tasks include removing environmental noise, enhancing weak vital signs signals, and extracting key features related to life from the signals. The data preprocessing module processes and optimizes multiple sensor data to ensure that downstream modules (such as feature fusion analysis module and positioning algorithm module) can obtain clearer and more reliable signal information, thereby improving the precision and accuracy of life detection.

[0041] In this embodiment, the specific composition and working mode of the data preprocessing module are as follows: The main task of the noise filtering unit is to remove the noise caused by environmental factors. Various interference factors in the environment, such as wind speed changes, electromagnetic interference, weather fluctuations, etc., will affect the sensor's signal, thereby affecting the detection accuracy of vital signs. To this end, the noise filtering unit uses a variety of filtering techniques to remove these irrelevant signals.

[0042] Typically, the noise filtering unit uses a Kalman filtering technique that smoothes the sensor data based on minimizing the estimation error.

[0043] As an option, the noise filtering unit may also use wavelet transform, which is particularly suitable for the case where there are multi-frequency components in the signal and can effectively remove high-frequency noise.

[0044] Specifically, the Kalman filter uses a recursive calculation method to continuously correct the estimate of the signal. The formula is as follows: in, For the The predicted value at time, is the Kalman gain, For the The observed value at time, is the observation matrix, is the updated estimate.

[0045] The main function of the signal enhancement unit is to increase the strength of weak vital signs signals to ensure that signals with low signal-to-noise ratios can be fully identified. In complex environments, the signals generated by living organisms are usually weak, so signal enhancement technology is needed to increase their strength to ensure that downstream modules can accurately analyze them.

[0046] In some embodiments, the signal enhancement unit enhances the signal through adaptive filtering technology. Adaptive filtering can dynamically adjust the filtering parameters according to the changes of the real-time collected data to optimize the signal enhancement effect.

[0047] As an option, the signal enhancement unit can use the local maximum detection method to extract sudden changes in the life signal (such as heartbeat, breathing, etc.) and further enhance the signal strength of these important features.

[0048] Specifically, the signal enhancement process can be described by the following formula: in, is the enhanced signal, is the original signal, is the background noise signal, is the enhancement factor, which determines the strength of signal enhancement.

[0049] The feature extraction unit is used to extract key features of vital signs from the signal after noise filtering and enhancement processing, including body temperature, breathing rate, gas concentration and vibration frequency, etc. These features are the basis for the subsequent feature fusion analysis module to identify and locate life forms.

[0050] Specifically, body temperature signals usually come from infrared sensors. When extracting body temperature characteristics, the temperature change rate analysis method is used. By calculating the temperature change rate, the heat source changes of living organisms can be accurately identified.

[0051] The respiratory and heart rate characteristics are usually collected by acoustic sensors, and the signals are analyzed by spectrum analysis to extract the heart and respiratory signals in specific frequency bands. The respiratory and heart rates of living organisms are extracted by fast Fourier transform (FFT) processing of the signal spectrum.

[0052] Gas concentration characteristics are obtained through gas sensors. The extracted characteristics are usually the rate of change of gas concentration, gas concentration It can be expressed by the following formula: in, is the gas concentration, is the gas volume, is the detection area of ​​the sensor, By calculating the time rate of change of concentration, the breathing pattern of the living organism can be extracted.

[0053] The vibration frequency is obtained through the vibration sensor, and feature extraction uses frequency domain analysis to identify the main frequency components of the vibration signal. It can be described by the following formula: in, is the vibration signal, is the initial amplitude of the signal, is the angular frequency of vibration, is the phase shift, is the attenuation factor, is a positive attenuation coefficient, is a sine function.

[0054] As an option, during the feature extraction process, the features of multiple sensors can be combined and weighted. Through weighted averaging or principal component analysis (PCA), the signal redundancy between different sensors can be further reduced and the accuracy of feature extraction can be improved.

[0055] In one possible implementation, during the data preprocessing phase, the signals of all sensors can be collected synchronously and processed in real time by an integrated preprocessing system. The system will adaptively adjust the filtering and enhancement parameters according to the quality of the data collected by the sensor, thereby ensuring that the data of each sensor can be processed under optimal conditions.

[0056] Please see attached Figure 4 , a feature fusion analysis module, which is used to weightedly fuse the features from each sensor, use a deep learning model to evaluate whether the signal source is a living organism, the deep learning model includes but is not limited to a convolutional neural network and a recurrent neural network, and automatically adjust the weight of the fusion feature; The feature fusion analysis module includes: A feature fusion unit, which is used to perform weighted fusion of features obtained from different sensors, so that the system can automatically adjust the fusion weight according to the reliability and information quality of different sensors; The signal recognition unit is used to compare the fused features with the life data database using a deep learning algorithm to determine whether the signal source is a living organism.

[0057] Specifically, the feature fusion analysis module is one of the core components of the present invention. Its main function is to weight the feature information from different types of sensors and analyze the fused features through a deep learning model to determine whether the signal source is a living organism. The key to this module is how to automatically adjust the fusion weights of each sensor feature to ensure that the system can adaptively adjust the fusion strategy according to the reliability and information quality of different sensors, thereby improving the accuracy and robustness of life detection.

[0058] In this embodiment, the structure and working principle of the feature fusion analysis module are as follows: The task of the feature fusion unit is to perform weighted fusion of the features obtained from each sensor. Different types of sensors (such as infrared, acoustic, gas, vibration, etc.) have their own advantages and limitations, and the feature fusion unit comprehensively considers the data quality of each sensor in a weighted manner, so that the system can adaptively adjust the fusion weight according to the actual situation.

[0059] In general, feature fusion uses the weighted average method to calculate the weighted sum of each sensor feature. This weighted sum can reflect the importance of different sensor information, and the weight can be adjusted through the system's real-time evaluation feedback mechanism.

[0060] As an option, the feature fusion unit can use principal component analysis (PCA) technology to reduce the dimension of the high-dimensional feature space, extract the most important features, reduce the impact of redundant information, and further improve the fusion accuracy.

[0061] Specifically, during the fusion process, the following weighted calculation formula can be used: in, is the fused feature vector, For the The features extracted by the sensors are For the The weight of each sensor, is the number of sensors. Weight It can be dynamically adjusted based on the signal quality (such as signal-to-noise ratio) of each sensor to ensure that sensors with higher quality contribute more to the fusion result.

[0062] The task of the signal recognition unit is to use deep learning algorithms to analyze the fused features and evaluate whether the signal source is a living organism. The deep learning model can be trained based on a large database of vital signs, enabling the system to learn the potential patterns of life signals and thus more accurately determine whether the signal source is a living organism.

[0063] In one possible implementation, the signal recognition unit uses a convolutional neural network (CNN) or a recurrent neural network (RNN) to analyze the fused features. CNN can efficiently process image data and is suitable for analyzing spatial features, while RNN is suitable for processing time series data and can capture the law of signal changes over time.

[0064] Specifically, in a convolutional neural network, after several layers of convolution and pooling, the feature map will be gradually reduced and abstracted, and finally the high-dimensional features will be mapped to classification results through the fully connected layer. The output of the CNN is used to determine whether the signal source is a living organism through the softmax activation function.

[0065] As an option, the signal recognition unit can also use a long short-term memory (LSTM) network to process time series features. The LSTM network can process long-distance dependent information in time series and is very effective in capturing changes in signals such as breathing and heartbeat of living organisms.

[0066] In this implementation, the feature fusion analysis module continuously optimizes the weights and feature selection so that the fused features can reflect the characteristics of life signals to the greatest extent, reduce the interference of environmental noise and non-living signals, and improve the judgment accuracy of the system.

[0067] In some embodiments, the feature fusion analysis module can dynamically adjust the feature weighting method according to the real-time environment and the working conditions of the sensor through incremental learning and adaptive adjustment mechanisms. This flexible feature fusion method can be optimized according to specific needs in different application scenarios to improve the adaptability of the system in various complex environments.

[0068] In this embodiment, the output result of this module not only provides the judgment information of whether the life form exists, but also provides the contribution of each sensor to the final judgment result. This feature enables the system to further improve the accuracy by analyzing the contribution of each sensor and optimizing the system parameter configuration when a misjudgment occurs.

[0069] As an extension, in order to further enhance the robustness of the system, the feature fusion analysis module can also combine traditional rule matching methods (such as threshold detection) with deep learning methods for hybrid mode decision making. That is, based on the judgment results of deep learning, threshold judgment and rule verification are added to further reduce false alarms and missed detections.

[0070] Please see attached Figure 5 , a positioning algorithm module, which is used to calculate the preliminary position of the life form based on the time difference ranging data of the signal source and the triangulation method; The positioning algorithm module includes: TDOA calculation unit, which is used to estimate the position of the target through the time difference between sensors; A triangulation unit is used to combine the known positions of multiple sensors and the received signal propagation time differences to optimize positioning accuracy through triangulation.

[0071] The time difference ranging formula is used in the TDOA calculation unit to estimate the position of the target. The formula is: in, For the signal to reach the sensor The time difference, is the actual position of the target life form, For the The known locations of the sensors, is the signal propagation speed, For sensor The receiving time.

[0072] Specifically, the positioning algorithm module is a key component of the present invention, which is responsible for estimating the initial position of the life form based on the time difference ranging (TDOA) data between sensors and triangulation. This module provides high-precision positioning results by combining multiple positioning technologies, which can effectively identify the position of life forms in complex environments and support subsequent rescue operations. In particular, the combination of TDOA calculation and triangulation enables the system to effectively overcome problems such as signal reflection and occlusion, and further improve positioning accuracy and robustness.

[0073] In this embodiment, the specific composition and working mode of the positioning algorithm module are as follows: The main task of the TDOA calculation unit is to estimate the location of the target by the time difference of the signals received by multiple sensors. TDOA technology calculates the location of the target by measuring the propagation time difference of the signal from the organism to different sensors and combining it with the known position of the sensor.

[0074] Generally, the TDOA calculation unit calculates the preliminary position of the target life form based on the time difference ranging formula, through the known signal propagation speed and the time difference between each sensor receiving the signal.

[0075] Specifically, the TDOA calculation formula is as follows: in, For the signal to reach the sensor The time difference, is the actual position of the target organism, to be solved, For the The known locations of the sensors, The speed of signal propagation, which may vary depending on the environment (such as air or other media), is usually a known constant. For sensor The reception time, usually recorded by the sensor when it receives the signal.

[0076] The key to TDOA technology is to calculate the target's position through time difference. In general, the signal propagation speed is It is known in different media (such as air, ruins, etc.), so the time difference between different sensors can be combined to infer the position of the target.

[0077] The role of the triangulation unit is to combine the known positions of multiple sensors and the time difference in propagation of the received signals to further optimize the positioning accuracy through triangulation. Triangulation uses the known positions of the sensors and the time difference in propagation of the signals between each sensor and the target to calculate the precise coordinates of the target's location.

[0078] As an option, if there are more than three sensors in the system, the triangulation unit can optimize the positioning result by combining multiple TDOA calculations and using geometric relationships.

[0079] Specifically, through the following geometric calculations, the constraint equation of the target position can be established by utilizing the signal propagation time difference between multiple sensors: in, Target to sensor The distance Target to sensor The time difference, is the signal propagation speed, For sensor Using these equations, the system can calculate the precise location of the target.

[0080] As an option, triangulation can also be combined with the least squares method to optimize the positioning results. The least squares method can further improve the positioning accuracy by minimizing the sum of squared errors.

[0081] In practical applications, the propagation speed of the signal It may be affected by environmental conditions, such as temperature, humidity or object obstruction. Therefore, during the positioning process, the TDOA calculation unit needs to adjust and compensate the signal propagation speed in real time.

[0082] Specifically, when the ambient temperature is low or the air density is high, the signal propagation speed is To compensate for this effect, the following adjustment formula can be used: in, is the adjusted signal propagation speed, is the signal propagation speed under standard conditions, is the temperature coefficient, It is the difference between the ambient temperature and the standard temperature.

[0083] In one possible implementation, the TDOA calculation unit and the triangulation unit are used in combination to ensure high-precision positioning in complex environments. In the case of a multi-sensor configuration, the triangulation unit can combine the calculation results of multiple signal propagation time differences to provide a more accurate target location.

[0084] Please see attached Figure 6 ,Result verification module, which is used to evaluate the credibility of the positioning results; The result verification module includes: A credibility evaluation unit, which is used to evaluate the credibility of the positioning result based on the signal-to-noise ratio and positioning error factors of the signal; The credibility judgment unit is used to judge whether the credibility is higher than a threshold.

[0085] The credibility evaluation unit uses a credibility calculation formula to evaluate the credibility of the positioning result, and the formula is: in, is the signal-to-noise ratio, is the actual position of the target life form, is the estimated position calculated by the algorithm, is the adjustment coefficient, which controls the impact of position error on credibility.

[0086] Specifically, the result verification module is one of the important components of the present invention, and its main function is to evaluate the credibility of the positioning result. The module ensures the accuracy and reliability of the positioning system in complex environments by strictly judging the credibility of the positioning result. By evaluating the signal-to-noise ratio (SNR) and positioning error of the signal, the result verification module can effectively determine whether the positioning result is true and reliable, and decide whether to accept the positioning result and whether further monitoring is needed based on the evaluation result.

[0087] In this embodiment, the specific composition and working principle of the result verification module are as follows: The task of the credibility assessment unit is to evaluate the credibility of the positioning results based on factors such as the signal-to-noise ratio (SNR) of the signal and the positioning error. The signal-to-noise ratio (SNR) is an important indicator for evaluating signal quality. A high SNR indicates a clear signal, small error, and high credibility. On the contrary, a low SNR indicates poor signal quality and may have large errors, so the positioning results need to be carefully verified.

[0088] Generally, the credibility evaluation unit uses the following formula to calculate the credibility of the positioning result: in, is the calculated credibility. The larger the value, the higher the credibility. It is the signal-to-noise ratio, a direct measure of signal quality. The higher the SNR value, the better the signal quality. is the actual position of the target life form, is the estimated position calculated by the positioning algorithm, It is an adjustment coefficient used to control the impact of position error on credibility. When the value is large, the impact of positioning error on credibility is greater.

[0089] Specifically, the credibility evaluation unit calculates the relationship between the signal-to-noise ratio and the estimated position error to obtain a comprehensive credibility index. When the signal quality is good and the error of the estimated position is small, the credibility will be high, indicating that the positioning result is more accurate. On the contrary, if the signal quality is poor or the position error is large, the credibility will be low, indicating that the positioning result may have a large deviation and needs further confirmation or repositioning.

[0090] The credibility judgment unit determines whether the positioning result is valid according to the credibility value calculated by the credibility evaluation unit. The unit determines whether the positioning result is reliable enough according to the preset threshold. If the credibility is higher than the set threshold, the positioning result is considered valid; if it is lower than the threshold, it returns to the step of continuing monitoring and re-positioning.

[0091] Generally, the credibility judgment unit sets a credibility threshold based on historical data or experimental results. ,like , then confirm that the positioning result is valid; if , then return to the first step and continue to detect and monitor the target.

[0092] Specifically, if the credibility calculated after credibility evaluation Exceeding the preset threshold , the system considers that the positioning result can be used for subsequent processing (such as sending it to the emergency command center for path planning). On the contrary, if the credibility does not meet the requirements, the system will perform further testing and optimize the positioning result.

[0093] As an option, in real-time applications, the confidence threshold Dynamic adjustments can be made through system learning or real-time feedback mechanisms. The thresholds can be flexibly adjusted according to different environments, the status and operation of different sensors, thereby improving the adaptability of the system.

[0094] In a possible implementation, the credibility judgment unit can also combine other auxiliary information (such as environmental changes, sensor status, etc.) to optimize the judgment process to make the judgment more accurate and efficient. For example, when the sensor is interfered, the credibility judgment unit can appropriately relax the threshold to adapt to the needs of environmental changes.

[0095] Please see attached Figure 7 , the emergency command center module is used to receive the final positioning results and update the rescue path planning in real time, guiding the rescue team to quickly locate and rescue.

[0096] The emergency command center modules include: A path planning unit, which is used to calculate and optimize the rescue path in real time based on the positioning results; A real-time feedback unit is used to adjust the path planning in real time according to the on-site rescue progress data.

[0097] Specifically, the emergency command center module is one of the key modules in the present invention. Its main function is to receive the final positioning results provided by the positioning system and update the rescue path planning in real time to ensure that the rescue team can quickly reach the location of the life body and carry out effective rescue. This module provides dynamic rescue path planning through the cooperation of the path planning unit and the real-time feedback unit, and adjusts the rescue route according to the real-time rescue progress data on site to improve rescue efficiency and personnel safety.

[0098] In this embodiment, the specific composition and working mode of the emergency command center module are as follows: The main task of the path planning unit is to calculate and optimize the path planning of the rescue team in real time based on the final positioning results provided by the positioning system, ensuring that the rescue team can reach the location of the target life form in the shortest time. The path planning unit generates one or more optimal paths for the rescue team by combining multiple algorithms (such as graph theory algorithm, Dijkstra algorithm, A* algorithm, etc.).

[0099] Generally, the path planning unit calculates the path based on the target positioning information and the existing environment map (such as the ruins structure map, the obstacle distribution at the fire scene, etc.).

[0100] As an option, the path planning unit uses the A* algorithm to perform path search. The A* algorithm combines the advantages of heuristic search and shortest path search, and can find the shortest path from the starting point to the target in a complex environment.

[0101] Specifically, in the rescue mission, the A* algorithm calculates the cost function of the path ,in is the actual cost from the starting point to the current point, is the estimated cost from the current point to the target point, is the total cost, and the path with the minimum cost is selected for search and optimization. Through this method, the path planning unit can avoid obstacles and dangerous areas while ensuring the shortest path.

[0102] The function of the real-time feedback unit is to adjust the path planning in real time according to the on-site rescue progress data to ensure that the rescue team can cope with the changing on-site conditions. As the rescue team continues to advance, new obstacles, dangers or environmental changes may appear. The real-time feedback unit dynamically optimizes the path planning by obtaining on-site rescue progress and real-time environmental data.

[0103] Generally, the real-time feedback unit is connected to on-site sensors and monitoring systems, and dynamically adjusts the path planning by obtaining real-time data such as the team's current location, task progress, and environmental changes.

[0104] As an option, the real-time feedback unit can estimate and predict the field data in real time through particle filtering or Kalman filtering technology to cope with rapid changes in the environment.

[0105] Specifically, suppose that the path planning unit has generated a preliminary rescue path, but during the rescue process, obstacles on site (such as collapsed walls or fire sources) make the original path infeasible. In this case, the real-time feedback unit recalculates the path that the rescue team can safely pass by adjusting the path planning goal. For example, the formula for updating the path state using particle filtering is: in, is the optimal path estimate at the current moment, It is the latest on-site feedback data. is the observation matrix, is the Kalman gain, and the particle filter update method can dynamically adjust the path planning according to the changes on site.

[0106] In this embodiment, the feedback mechanism and dynamic adjustment capability of the real-time feedback unit ensure the flexibility of the rescue operation. As the on-site environment changes, the real-time feedback unit can update the path information in a timely manner to ensure that the rescue team can bypass new obstacles or dangerous areas, thereby ensuring the efficiency and safety of the rescue operation.

[0107] As an extension, the collaborative work of the path planning unit and the real-time feedback unit can be achieved through multi-objective optimization, which not only considers the shortest path, but also multiple factors such as rescue efficiency, team member safety, and resource optimization, further improving the intelligence level of the system. In an emergency, the system can also allocate tasks among different rescue teams, optimize resource scheduling, and make the overall rescue operation more efficient.

[0108] Please see attached Figure 8 The multi-sensor fusion life detection and positioning method described below and the multi-sensor fusion life detection and positioning system described above can correspond to each other.

[0109] A multi-sensor fusion life detection and positioning method comprises the following steps: S1. Deploy a multi-type sensor array in the target area to collect environmental data of the target area; S2, performing noise filtering and signal enhancement processing on the raw data collected by the sensor to extract the key features of the living organism; S3, weighted fusion of feature information extracted by each sensor, and use of deep learning model to analyze whether the signal source is a living organism; S4, calculating the preliminary position of the target life form based on the time difference ranging and triangulation method of the signal source; S5. Conduct credibility assessment on the initially calculated positioning results to determine the validity of the results; S6. If the credibility is higher than the preset threshold, the positioning result is confirmed to be valid and sent to the emergency command center to guide the rescue team to perform the task; if the credibility is low, return to continue monitoring.

[0110] The method of this embodiment can be used to execute the above system embodiment, and its principles and technical effects are similar, which will not be repeated here.

[0111] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-sensor fusion life detection and positioning system, characterized in that: include: A sensor array module, which is composed of multiple sensors of different types and is used to collect environmental data in the target area; A data preprocessing module, which is used to filter out noise and enhance signals of the data collected by the sensor, and extract key features of vital signs therefrom; A feature fusion analysis module, which is used to weight the features from each sensor, evaluate whether the signal source is a living organism using a deep learning model, the deep learning model includes but is not limited to a convolutional neural network and a recurrent neural network, and automatically adjust the weight of the fusion feature; A positioning algorithm module, which is used to calculate the preliminary position of the life form based on the time difference ranging data of the signal source and the triangulation method; A result verification module is used to evaluate the credibility of the positioning results; The emergency command center module is used to receive the final positioning results and update the rescue path planning in real time, guiding the rescue team to quickly locate and rescue.

2. A multi-sensor fusion life detection and positioning system according to claim 1, characterized in that: The sensors include, but are not limited to, infrared sensors, acoustic sensors, gas sensors, and vibration sensors.

3. The multi-sensor fusion life detection and positioning system according to claim 1, characterized in that: The data preprocessing module comprises: A noise filtering unit, which is used to remove noise from environmental factors, including but not limited to wind speed and electromagnetic interference; A signal enhancement unit, which is used to increase the strength of weak vital sign signals; The feature extraction unit is used to extract key features such as body temperature, breathing rate, gas concentration and vibration frequency of living organisms from the signal.

4. The multi-sensor fusion life detection and positioning system according to claim 1, characterized in that: The feature fusion analysis module includes: A feature fusion unit, which is used to perform weighted fusion of features obtained from different sensors, so that the system can automatically adjust the fusion weight according to the reliability and information quality of different sensors; The signal recognition unit is used to compare the fused features with the life data database using a deep learning algorithm to determine whether the signal source is a living organism.

5. The multi-sensor fusion life detection and positioning system according to claim 1, characterized in that: The positioning algorithm module includes: TDOA calculation unit, which is used to estimate the position of the target through the time difference between sensors; A triangulation unit is used to combine the known positions of multiple sensors and the received signal propagation time differences to optimize positioning accuracy through triangulation.

6. The multi-sensor fusion life detection and positioning system according to claim 5, characterized in that: The TDOA calculation unit uses a time difference ranging formula to estimate the position of the target, and the formula is: in, For the signal to reach the sensor The time difference, is the actual position of the target life form, For the The known locations of the sensors, is the signal propagation speed, For sensor The receiving time.

7. The multi-sensor fusion life detection and positioning system according to claim 1, characterized in that: The result verification module comprises: A credibility evaluation unit, which is used to evaluate the credibility of the positioning result based on the signal-to-noise ratio and positioning error factors of the signal; The credibility judgment unit is used to judge whether the credibility is higher than a threshold.

8. The multi-sensor fusion life detection and positioning system according to claim 7, characterized in that: The credibility evaluation unit uses a credibility calculation formula to evaluate the credibility of the positioning result, and the formula is: in, is the signal-to-noise ratio, is the actual position of the target life form, is the estimated position calculated by the algorithm, is the adjustment coefficient, which controls the impact of position error on credibility.

9. The multi-sensor fusion life detection and positioning system according to claim 1, characterized in that: The emergency command center module includes: A path planning unit, which is used to calculate and optimize the rescue path in real time based on the positioning results; A real-time feedback unit is used to adjust the path planning in real time according to the on-site rescue progress data.

10. A multi-sensor fusion life detection and positioning method, using a multi-sensor fusion life detection and positioning system according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1. Deploy a multi-type sensor array in the target area to collect environmental data of the target area; S2, performing noise filtering and signal enhancement processing on the raw data collected by the sensor to extract the key features of the living organism; S3, weighted fusion of feature information extracted by each sensor, and use of deep learning model to analyze whether the signal source is a living organism; S4, calculating the preliminary position of the target life form based on the time difference ranging and triangulation method of the signal source; S5. Conduct credibility assessment on the initially calculated positioning results to determine the validity of the results; S6. If the credibility is higher than the preset threshold, the positioning result is confirmed to be valid and sent to the emergency command center to guide the rescue team to perform the task; if the credibility is low, return to continue monitoring.

Images