A Real-Time Data Analysis Method for Mobile Reconnaissance Equipment Based on Multi-Sensor Fusion

Abstract

This invention discloses a real-time data analysis method for mobile reconnaissance equipment based on multi-sensor fusion, belonging to the field of multi-sensor fusion technology. It includes a communication relay terminal generating graded early warning commands based on a dynamic threat index; and a motion monitoring terminal adjusting the equipment's motion parameters and executing obstacle avoidance strategies based on the early warning commands and the dynamic threat index. This invention enables the equipment to adopt strategies such as emergency braking, deceleration and replanning, or enhanced environmental monitoring for different threat levels by generating graded early warning commands through the communication relay terminal, thereby improving the equipment's flexibility and safety in complex environments. Furthermore, by adjusting the equipment's motion parameters and executing obstacle avoidance strategies based on the early warning commands and the dynamic threat index through the motion monitoring terminal, combined with PID controller optimization of the motion trajectory, the obstacle avoidance capability and motion stability of the equipment in complex environments are improved.

Description

Technical Field

[0001] This invention relates to the field of multi-sensor fusion technology, and in particular to a real-time data analysis method for mobile reconnaissance equipment based on multi-sensor fusion. Background Technology

[0002] With the rapid development of information technology and sensor technology, mobile reconnaissance equipment is finding increasingly diverse applications in military reconnaissance, security monitoring, and smart city management. In complex field environments, high-risk disaster sites, or covert reconnaissance missions, the equipment needs to perceive environmental changes in real time, monitor its own movement, and track target characteristics. This places stringent demands on the real-time fusion and analysis capabilities of multi-source heterogeneous data, as traditional single-sensor data analysis methods are no longer sufficient to meet the needs of accurate decision-making in complex scenarios.

[0003] Currently, while existing mobile reconnaissance equipment data analysis technology has achieved preliminary fusion of multi-sensor data, in the data preprocessing stage, there is a lack of efficient alignment and cleaning mechanisms for heterogeneous data that are not synchronized in time and space, resulting in insufficient accuracy in feature matrix construction. In the risk assessment process, static weight prediction models are mostly used, which cannot be dynamically adjusted according to the degree of environmental interference, the reliability of motion trajectory and the quality of target features, resulting in deviations between threat index calculation and actual risk scenarios.

[0004] However, current technologies face significant challenges in practical applications: environmental factor data, motion state data, and target feature data collected by multi-source sensor arrays suffer from spatiotemporal biases and noise interference, making it difficult for traditional cleaning algorithms to guarantee data integrity; environmental interference increases the standard deviation of sensor data, and existing interference suppression models cannot dynamically adapt to different attenuation factors; acceleration deviations and heading angle errors between IMU and GPS data introduce drift risks in motion trajectory confidence calculations; the signal-to-noise ratio differences and motion prediction biases of target recognition sensors are not fully quantified, resulting in the fusion index failing to accurately reflect target characteristics; fixed-weight multimodal risk prediction models lack adaptive update mechanisms, making them prone to misjudging threat indices in complex scenarios, thereby affecting the accuracy of graded early warnings and the effectiveness of equipment obstacle avoidance strategies.

[0005] Therefore, it is essential to invent a real-time data analysis method for mobile reconnaissance equipment based on multi-sensor fusion to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide a real-time data analysis method for mobile reconnaissance devices based on multi-sensor fusion, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a real-time data analysis method for mobile reconnaissance equipment based on multi-sensor fusion, comprising an environmental perception terminal, a motion monitoring terminal, a weight optimization terminal, a data fusion terminal, a decision control terminal, a communication relay terminal, and a mobile reconnaissance equipment, specifically including the following steps:

[0008] S1. The environmental perception terminal collects heterogeneous environmental data through a multi-source sensor array during the operation of the mobile reconnaissance device, forming a multi-source spatiotemporal synchronous dataset; the multi-source spatiotemporal synchronous dataset includes environmental factor data, motion state data, and target feature data.

[0009] S2. The data fusion terminal performs spatiotemporal alignment and data cleaning on multi-source spatiotemporal synchronous datasets to construct a standardized feature matrix.

[0010] S3. The decision control terminal analyzes the standardized feature matrix to obtain the environmental interference suppression coefficient, motion trajectory confidence and target feature fusion index, and inputs the three coefficients into the multimodal risk prediction model to output the dynamic threat index.

[0011] S4. The communication relay terminal generates graded early warning instructions based on the dynamic threat index;

[0012] S5. The motion monitoring terminal adjusts the device's motion parameters and executes obstacle avoidance strategies based on early warning commands and dynamic threat indices.

[0013] S6. The weight optimization terminal dynamically updates the weight coefficients in the multimodal risk prediction model based on the dynamic threat index.

[0014] Preferably, the environmental factor data includes light intensity, temperature, humidity, atmospheric pressure, wind speed, electromagnetic field strength, and dust concentration; the motion state data includes three-dimensional acceleration, angular velocity, geomagnetic azimuth, and GPS positioning coordinates; and the target feature data includes the sensor signal-to-noise ratio and measured motion vector.

[0015] Preferably, the environmental interference suppression coefficient is specifically:

[0016]

[0017] Where N is the number of environmental sensors, σ i Let μ be the standard deviation of the data collected by the i-th sensor. i Let be the mean value of the data collected by the i-th sensor, k be the environmental attenuation factor, Δt be the length of the data acquisition time window, and e be the natural constant.

[0018] Preferably, the confidence level of the motion trajectory is specifically:

[0019]

[0020] Among them, is the acceleration vector measured by the IMU, is the acceleration vector estimated by GPS, a max is the maximum allowable acceleration of the device, △θ is the absolute value of the deviation between the geomagnetic azimuth and the GPS heading angle, and π is the pi.

[0021] Preferably, the target feature fusion index is specifically:

[0022]

[0023] Among them, M is the number of target recognition sensors, SNR j is the signal-to-noise ratio of the j-th sensor, ω j is the weight of the j-th sensor, is the measured motion vector of the target, is the motion prediction vector, v max is the maximum tracking speed of the device.

[0024] Preferably, the multi-modal risk prediction model is specifically:

[0025]

[0026] Among them, Q = α × (β1 × C1 + β2 × C2 + β3 × C3 - γ), C1 is the environmental interference suppression coefficient, C2 is the motion trajectory confidence, C3 is the target feature fusion index, α is the scale factor, γ is the risk determination threshold, e is the natural constant, and β1, β2, and β3 are weight coefficients, β1 + β2 + β3 = 1 and β1, β2, and β3 ∈ [0, 1].

[0027] Preferably, the hierarchical warning instruction is:

[0028] When the dynamic threat index E ≥ E1, trigger the first-level warning instruction and start the emergency braking and defense protocol;

[0029] When E2 ≤ E < E1, trigger the second-level warning instruction and activate the device deceleration and path replanning;

[0030] When 0 ≤ E < E2, trigger the third-level warning instruction and enable the enhanced environmental monitoring mode and the basic avoidance strategy.

[0031] Preferably, the execution process of the motion monitoring terminal is:

[0032] A1. Analyze the warning instruction level and call the preset motion parameter mapping table;

[0033] A2. Calculate the obstacle avoidance vector based on the dynamic threat index:

[0034]

[0035] Where λ is the obstacle avoidance gain coefficient. The measured motion vector of the target. For motion prediction vectors;

[0036] A3. Adjust the motor torque and steering angle using a PID controller to ensure the equipment's motion trajectory meets the following requirements:

[0037]

[0038] Among them, a limit The preset value for the final upper limit of the device's acceleration, a max The maximum acceleration preset value for the equipment is given by , E is the dynamic threat index, and min is the minimum value function.

[0039] Preferably, the dynamic update of the weighting coefficients specifically involves:

[0040]

[0041] in,

[0042]

[0043] η(E)=η0×(1+0.5×E),

[0044] Where e is the natural constant, These are the final weighting coefficients after dynamic updates. The intermediate weights are scaled by the dynamic threat index. Here, m(E) is the threat response strength coefficient, η(E) is the weight adjustment rate, η0 is the baseline learning rate, and C is the weights adjusted by short-term bias. i To evaluate the coefficients in real time, The moving average of the coefficients. Let i represent the degree of fluctuation of the coefficient, and E represent the dynamic threat index, where i = 1, 2, 3.

[0045] The technical effects and advantages of this invention are as follows:

[0046] 1. This invention calculates the environmental interference suppression coefficient, motion trajectory confidence, and target feature fusion index through a decision control terminal, and outputs a dynamic threat index using a multimodal risk prediction model, thereby realizing a multi-dimensional quantitative assessment of environmental risks and providing a scientific basis for graded early warning.

[0047] 2. This invention generates graded early warning commands based on the dynamic threat index through a communication relay terminal, enabling the equipment to take strategies such as emergency braking, deceleration and replanning, or enhanced environmental monitoring in response to different threat levels, thereby improving the equipment's flexibility and safety in dealing with complex environments.

[0048] 3. This invention adjusts the device's motion parameters and executes obstacle avoidance strategies based on early warning commands and dynamic threat indices through a motion monitoring terminal, and optimizes the motion trajectory by combining a PID controller, thereby improving the device's obstacle avoidance capability and motion stability in complex environments;

[0049] 4. This invention achieves adaptive adjustment of model parameters by dynamically updating the weight coefficients in the multimodal risk prediction model based on the dynamic threat index through a weight optimization terminal, enabling the system to maintain accurate risk prediction capabilities in different scenarios. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the device connection according to the present invention.

[0051] Figure 2 This is a schematic diagram of the method steps of the present invention.

[0052] Figure 3 This is a schematic diagram illustrating the execution process of the motion monitoring terminal of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] This invention provides, for example Figure 1 The device connection diagram shown includes an environmental sensing terminal, a motion monitoring terminal, a weight optimization terminal, a data fusion terminal, a decision control terminal, a communication relay terminal, and a mobile reconnaissance device;

[0055] This invention provides, for example Figure 2 The real-time data analysis method for mobile reconnaissance equipment based on multi-sensor fusion, as shown, specifically includes the following steps:

[0056] S1. The environmental perception terminal collects heterogeneous environmental data through a multi-source sensor array during the operation of the mobile reconnaissance device, forming a multi-source spatiotemporal synchronous dataset; the multi-source spatiotemporal synchronous dataset includes environmental factor data, motion state data, and target feature data.

[0057] Furthermore, in the above technical solution, the environmental factor data includes light intensity, temperature, humidity, atmospheric pressure, wind speed, electromagnetic field strength, and dust concentration; the motion state data includes three-dimensional acceleration, angular velocity, geomagnetic azimuth, and GPS positioning coordinates; and the target feature data includes the sensor signal-to-noise ratio and measured motion vector.

[0058] It should be noted that the light intensity is detected in real time by a photosensitive sensor, such as a photoresistor or a silicon photovoltaic cell, and the light signal is converted into an electrical signal and then quantized and output.

[0059] The temperature is detected by temperature sensors, such as thermocouples, thermistors, and digital temperature chips, and the temperature value is reflected by the change in resistance or voltage.

[0060] The humidity is measured by a humidity sensor, such as a capacitive or resistive humidity sensor, to measure the water vapor content in the air, and the humidity is calculated by utilizing the effect of water molecules on the dielectric constant or resistance of the sensor.

[0061] The atmospheric pressure is measured using a pressure sensor, such as a MEMS pressure chip, to detect changes in atmospheric pressure and output a real-time pressure value in conjunction with an altitude calibration algorithm.

[0062] The wind speed is measured by a wind speed sensor, such as an impeller anemometer or an ultrasonic anemometer, which measures the airflow speed. The impeller rotation or ultrasonic time difference is converted into a wind speed value.

[0063] The electromagnetic field strength is detected by electromagnetic field sensors, such as Hall effect sensors and coil sensors, and quantified by the change in induced electromotive force or magnetic field.

[0064] The dust concentration is detected by a dust sensor, such as an optical dust sensor or a capacitive dust sensor, to measure the content of suspended particulate matter in the air, and the concentration is calculated using the principles of light scattering or charge induction.

[0065] The three-dimensional acceleration is measured by a triaxial accelerometer, such as a MEMS accelerometer, in the X, Y, and Z axes, and outputs three-dimensional vector data.

[0066] The angular velocity is measured using a three-axis gyroscope, such as a MEMS gyroscope, to measure the rotational angular velocity around three coordinate axes, which is then converted into an electrical signal through the Coriolis effect.

[0067] The geomagnetic azimuth angle is calculated by using a magnetometer, such as a magnetoresistive sensor, to detect the direction of the Earth's magnetic field, and combining the data from a gyroscope and an accelerometer to determine the azimuth angle of the device.

[0068] The GPS positioning coordinates are obtained by receiving satellite signals through the GPS module, parsing coordinate information such as longitude, latitude, and altitude, and combining them with a timestamp to achieve positioning.

[0069] The signal-to-noise ratio of the sensor is calculated by a signal processing algorithm, which is the ratio of the power of the output signal of the target sensor, such as a camera or radar, to the noise power, reflecting the signal quality.

[0070] The measured motion vector is tracked by visual sensors, such as infrared cameras, millimeter-wave radar or lidar, and the displacement, velocity and direction of the target are calculated by combining multiple frames of data to form a three-dimensional motion vector. The technical principle is to use target detection algorithms, such as YOLO, to identify the target, and to track the target's motion trajectory and solve the real-time motion vector through optical flow or Kalman filtering.

[0071] All of the above data are collected in real time by the multi-source sensor array of the environmental sensing terminal. After the various sensors convert physical quantities into electrical signals, they are converted into heterogeneous environmental data through analog-to-digital conversion and preliminary signal processing.

[0072] S2. The data fusion terminal performs spatiotemporal alignment and data cleaning on multi-source spatiotemporal synchronous datasets to construct a standardized feature matrix.

[0073] It is important to understand that the execution process of the data fusion terminal is as follows: First, spatiotemporal alignment is performed. Addressing the differences in timestamps between data collected by different sensors in the environmental perception terminal, interpolation algorithms or clock synchronization protocols are used to align heterogeneous data to a unified time reference, ensuring data temporal consistency. Simultaneously, the local coordinate systems of each sensor are transformed into a unified global coordinate system using a coordinate transformation matrix, enabling spatial mapping between motion state data and environmental factor data. Second, data cleaning is performed. Kalman filtering, median filtering, or Gaussian filtering algorithms are used to denoise the raw sensor data. Then, statistical methods or machine learning algorithms are used to identify outliers in the data and repair them through methods such as adjacent data interpolation and regression prediction. For missing data, multiple interpolation or time-series prediction models are used to fill in the gaps, ensuring the integrity of the dataset. Finally, a standardized feature matrix is ​​constructed. The cleaned heterogeneous data is converted to a unified numerical range to eliminate dimensional differences. Key features are then extracted from the cleaned data, arranged into multidimensional feature vectors according to time series, and integrated into a standardized feature matrix. Rows represent time series, and columns represent different types of features, thus providing structured input for the multimodal risk prediction model of the decision control terminal.

[0074] S3. The decision control terminal analyzes the standardized feature matrix to obtain the environmental interference suppression coefficient, motion trajectory confidence and target feature fusion index, and inputs the three coefficients into the multimodal risk prediction model to output the dynamic threat index.

[0075] Furthermore, in the above technical solution, the environmental interference suppression coefficient is specifically:

[0076]

[0077] Where N is the number of environmental sensors, σ i Let μ be the standard deviation of the data collected by the i-th sensor. i Let be the mean value of the data collected by the i-th sensor, k be the environmental attenuation factor, Δt be the length of the data acquisition time window, and e be the natural constant.

[0078] It should be noted that the value of the environmental degradation factor k is set as follows: Where ln is the logarithm to the base e, T 1 / 2 The half-life of environmental interference is specifically taken as the coefficient of variation of sensor data. The time required for the temperature to decay from peak value to 50%;

[0079] It should be noted that the formula for the environmental interference suppression coefficient C1 is designed based on multi-sensor data stability assessment, time decay effect, and interference degree quantification. In this formula, It is the square of the coefficient of variation, used to measure the relative dispersion of single-sensor data. While eliminating the influence of the mean dimension, it amplifies the impact of fluctuations through squaring, reflecting the degree of data instability caused by environmental interference. The average interference levels of all sensors are calculated to achieve multi-sensor fusion, avoiding the influence of single sensor anomalies and comprehensively reflecting the overall environmental interference level; -k×Δt As an exponential decay term, it describes the temporal correlation of environmental interference, so that as the time window Δt extends, the contribution of early data to the current interference assessment decays exponentially, and the decay rate is controlled by k, thereby weakening the influence of historical data and enhancing the response to real-time interference; the whole is structured by C1 = 1 - interference comprehensive term, so that C1 is negatively correlated with the degree of interference. When there is no environmental interference, C1≈1, and C1 decreases when the interference is strong, which intuitively quantifies the intensity of environmental interference.

[0080] Furthermore, in the above technical solution, the confidence level of the motion trajectory is specifically as follows:

[0081]

[0082] in, The acceleration vector measured by the IMU. a is the acceleration vector calculated by GPS. max Δθ is the maximum permissible acceleration of the device, Δθ is the absolute value of the deviation between the geomagnetic azimuth angle and the GPS heading angle, and π is pi.

[0083] It's important to understand that the formula for the motion trajectory confidence level C2 is designed around the consistency verification of multi-source sensors. It comprehensively evaluates the reliability of trajectory estimation by quantifying the deviation between acceleration and heading. In the online motion consistency dimension, IMUs offer strong real-time performance but are prone to drift, while GPS provides absolute reference but is susceptible to obstruction and interference. Excessive acceleration deviation between the two suggests sensor failure or environmental interference. Vector difference... The norm is converted to a scalar deviation, and divided by a. max Normalization, in order to eliminate the influence of dimensions, unifies the measurement to the dimension of the proportion of equipment motion capability. Squaring amplifies the impact of deviations; for example, if the deviation reaches 'a'... max When the denominator is 2, the confidence level of this part drops to one-half, while The structure ensures that when the acceleration is completely consistent, this part approaches 1, and the larger the deviation, the closer it is to 0, which intuitively reflects the confidence level of linear motion. In the dimension of angular motion consistency, geomagnetic heading is easily affected by magnetic interference, and GPS heading depends on continuous positioning. Excessive deviation between the two suggests that the heading measurement is unreliable. △θ is taken as an absolute value because the heading deviation has no direction and only focuses on the magnitude. The theoretical maximum deviation is π. If it exceeds this, it is equivalent to a small reverse angle. Dividing by π and normalizing to the [0,1] interval, the denominator is... When the heading is perfectly aligned, this part approaches 1; when the deviation reaches 180°, it drops to half, reflecting the physical intuition that the local confidence level of the heading dimension is halved when the heading is reversed. At the multi-dimensional fusion level, the consistency of acceleration and heading are independent of each other but synergistically affect the confidence level of the motion trajectory. The product operation amplifies the defects of the two dimensions. For example, when the deviation of each dimension causes the confidence level of the motion trajectory to be halved, the overall confidence level drops to one-quarter. Only when both are aligned at the same time does C2 approach 1. The mathematical structure strictly constrains the destruction of the confidence level of the motion trajectory by inconsistency.

[0084] Furthermore, in the above technical solution, the target feature fusion index is specifically:

[0085]

[0086] Where M is the number of target recognition sensors, and SNR j Let ω be the signal-to-noise ratio of the j-th sensor. j Let j be the weight of the j-th sensor. The measured motion vector of the target. v is the motion prediction vector. max This is the maximum tracking speed of the device.

[0087] It is important to know that the sensor weight ω j Setting the value: Among them, R j Let A be the reference signal-to-noise ratio of the j-th sensor. jThis is the priority coefficient for the j-th sensor type. The specific value is set based on industry experience. For example, if the sensor type is millimeter-wave radar, the priority coefficient is set to 1 because it has strong anti-interference capabilities; if the sensor type is lidar, the priority coefficient is set to 0.9 because it has high accuracy but is affected by weather; if the sensor type is infrared camera, the priority coefficient is set to 0.8 because it has advantages at night.

[0088] It should be noted that the first part of the formula for the target feature fusion index C3 is obtained through... The aim is to comprehensively evaluate the signal reliability of target recognition sensors, such as cameras and radar, among which the signal-to-noise ratio (SNR) is crucial. j The signal clarity of the j-th sensor is directly quantified, while the preset weight ω is used. j The weighting is dynamically allocated based on the environmental adaptability of the sensor type. For example, radar has a higher weight under strong electromagnetic interference, while optical sensors have a lower weight in dusty environments. Through weighted summation and weight normalization, this part compresses the result to the [0, 1] interval: the output approaches 1 when all sensors have extremely high signal-to-noise ratios, and approaches 0 if a sensor fails or the signal quality drops sharply. This effectively avoids system misjudgment caused by a single sensor failure and improves the anti-interference capability in multi-source heterogeneous environments; the latter half of the formula is... Focusing on resolving the risk of discrepancy between predicted and actual target motion, this is achieved by calculating the actual motion vector. With prediction vector The Euclidean distance deviation, divided by the device's maximum tracking speed v. max Normalization is performed to convert absolute deviation into a relative quantity: if the measured value is exactly the same as the prediction (deviation is 0), this item outputs 1; if the deviation exceeds v... max If the target suddenly maneuvers at high speed, the output approaches 0. This design dynamically reflects the real-time accuracy of the tracking system, avoiding missed alarms due to model lag or environmental disturbances. Ultimately, a strong two-dimensional correlation is achieved through the multiplication operation C3 = (data quality term) × (motion consistency term). Its physical meaning is that defects in either data quality or motion consistency will significantly lower the C3 value. For example, if the sensor signal-to-noise ratio is high but motion prediction is severely inaccurate, the data quality term is approximately 1, the motion consistency term is approximately 0, and C3 is approximately 0; conversely, if motion prediction is accurate but the sensor signal is interfered with, the data quality term is approximately 0, the motion consistency term is approximately 1, and C3 is approximately 0. This design forces the system to output a high C3 value only when perception is reliable and prediction matches, fundamentally reducing the misjudgment rate in complex scenarios.

[0089] Furthermore, in the above technical solution, the multimodal risk prediction model specifically refers to:

[0090]

[0091] Where Q = α × (β1 × C1 + β2 × C2 + β3 × C3 -γ), C1 is the environmental interference suppression coefficient, C2 is the moving trajectory confidence level, C3 is the target feature fusion index, α is the scale factor, γ is the risk determination threshold, e is the natural constant, and β1, β2, and β3 are weight coefficients, where β1 + β2+ β3 = 1 and β1, β2, and β3 ∈ [0, 1].

[0092] It should be noted that the initial values of the weight coefficients β1, β2, and β3 are set as follows: for scenarios with complex environments and low-speed movement, β1 = 0.4, β2 = 0.3, β3 = 0.3; for scenarios involving high maneuverability and target tracking priority, β1 = 0.2, β2 = 0.5, β3 = 0.3; for scenarios with harsh environments and no specific targets, β1 = 0.6, β2 = 0.3, β3 = 0.1;

[0093] The value of the scale factor α is set to 2.5;

[0094] The value of the risk determination threshold γ is set as: γ = γ0 + △γ env +△γ task where γ0 = 0.6, △γ env is the environmental compensation term, and its calculation formula is where N working is the number of effective working environment sensors, N total is the total number of environmental sensors, is the standard deviation of the environmental interference suppression coefficient, is the moving average of the environmental interference suppression coefficient, and △γ task is the task compensation term, which is specifically set as follows: in scenarios at night or in bad weather, △γ task = 0.15; in the reconnaissance mode, △γ task = -0.1; in the regular patrol mode, △γ task = 0.

[0095] S4. The communication relay terminal generates a hierarchical warning instruction according to the dynamic threat index;

[0096] Furthermore, in the above technical solution, the hierarchical warning instruction is:

[0097] When the dynamic threat index E ≥ E1, a first-level warning instruction is triggered to activate the emergency braking and defense protocol;

[0098] When the dynamic threat index E2 ≤ E < E1, a second-level warning instruction is triggered to activate the device deceleration and path replanning;

[0099] When the dynamic threat index 0 ≤ E < E2, a third-level warning instruction is triggered to enable the enhanced environmental monitoring mode and the basic avoidance strategy.

[0100] It should be noted that E1 = 0.7 and E2 = 0.3 can be adjusted according to the actual situation. For example, in a complex static environment, E2 can be reduced to 0.25 to extend the activation time of the environmental monitoring enhancement mode.

[0101] It should be noted that the emergency braking is achieved by performing a torque reversal operation through the motion monitoring terminal, so that the equipment acceleration meets the required standard. To achieve an emergency stop; the defense protocol involves activating a physical protection mechanism, such as starting a protective shield, while simultaneously blocking critical modules of the equipment to prevent external intrusion.

[0102] The active deceleration is to limit the acceleration to the upper limit 'a' set by the preset parameter group in Table 1. limit =0.5×a max The PID controller smoothly reduces speed; the path replanning provides a standardized environment matrix for the data fusion terminal, and the motion monitoring terminal recalculates the obstacle avoidance vector based on the standardized environment matrix. Use A* or RRT algorithms to generate new paths;

[0103] The environmental monitoring enhancement mode activates backup sensor resources and optimizes operating parameters, such as increasing the infrared camera frame rate to 60fps to enhance dynamic target capture capabilities and expanding the lidar scanning angle to 180° to cover a wider monitoring area; the basic avoidance strategy sets the acceleration upper limit 'a' according to the preset parameter group in Table 1. limit =0.2×a max Furthermore, the steering angle change rate is constrained to ≤15° / s by a PID controller to ensure that the motion trajectory meets the requirements.

[0104] S5. The motion monitoring terminal adjusts the device's motion parameters and executes obstacle avoidance strategies based on early warning commands and dynamic threat indices.

[0105] Furthermore, in the above technical solution, the execution process of the motion monitoring terminal is as follows: Figure 3 As shown:

[0106] A1. Parse the warning command level and call the preset motion parameter mapping table;

[0107] A2. Calculate obstacle avoidance vectors based on dynamic threat index:

[0108]

[0109] Where λ is the obstacle avoidance gain coefficient. The measured motion vector of the target. For motion prediction vectors;

[0110] A3. Adjust the motor torque and steering angle using a PID controller to ensure the equipment's motion trajectory meets the following requirements:

[0111]

[0112] Among them, a limit The preset value for the final upper limit of the device's acceleration, a max The maximum acceleration preset value for the equipment is given by , E is the dynamic threat index, and min is the minimum value function.

[0113] It should be noted that the preset motion parameter mapping table is shown in Table 1:

[0114] Table 1

[0115] Warning Level Dynamic Threat Index E Preset parameter group Level 1 warning <![CDATA[E≥E1]]> <![CDATA[{λ=0.5,a limit =0.8×a max }]]> Level II Warning <![CDATA[E2≤E<E1]]> <![CDATA[{λ=1.2,a limit =0.5×a max }]]> Level III Warning <![CDATA[0≤E<E2]]> <![CDATA[{λ=2,a limit =0.2×a max }]]>

[0116] Among them, a limit The device's final acceleration upper limit is preset, and the device's motion trajectory must meet the following requirements in the initial stage.

[0117] S6. The weight optimization terminal dynamically updates the weight coefficients in the multimodal risk prediction model based on the dynamic threat index.

[0118] Furthermore, in the above technical solution, the dynamic update of the weighting coefficients specifically involves:

[0119]

[0120] in,

[0121]

[0122] η(E)=η0×(1+0.5×E),

[0123] Where e is the natural constant, These are the final weighting coefficients after dynamic updates. The intermediate weights are scaled by the dynamic threat index. Here, m(E) is the threat response strength coefficient, η(E) is the weight adjustment rate, η0 is the baseline learning rate, and C is the weights adjusted by short-term bias. i To evaluate the coefficients in real time, The moving average of the coefficients. Let i represent the degree of fluctuation of the coefficient, and E represent the dynamic threat index, where i = 1, 2, 3.

[0124] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A mobile investigation equipment real-time data analysis method based on multi-sensor fusion, characterized in that, The application relates to a mobile detection device, comprising an environment perception terminal, a motion monitoring terminal, a weight optimization terminal, a data fusion terminal, a decision control terminal, a communication relay terminal and the mobile detection device, and specifically comprising the following steps: S1, the environment perception terminal collects heterogeneous environment data through a multi-source sensor array when the mobile detection device is running, and forms a multi-source time-space synchronous data set; the multi-source time-space synchronous data set comprises environment factor data, motion state data and target feature data; S2, the data fusion terminal performs time-space alignment and data cleaning on the multi-source time-space synchronous data set, and constructs a standardized feature matrix; S3, the decision control terminal analyzes the standardized feature matrix to obtain an environment interference suppression coefficient, a motion trajectory confidence degree and a target feature fusion index, and inputs the three coefficients into a multi-modal risk prediction model to output a dynamic threat index; the environment interference suppression coefficient is specifically: ， where N is the number of environmental sensors, σ i is the standard deviation of data collected by the i-th sensor, μ i is the mean of data collected by the i-th sensor, k is an environmental attenuation factor, At is the length of the data collection time window, and e is the natural constant. the motion trajectory confidence degree is specifically: ， wherein, is the acceleration vector measured by the IMU, is the acceleration vector calculated by the GPS, max is the maximum allowed acceleration of the device, and is the absolute value of the deviation between the magnetic azimuth and the GPS heading angle, and π is the circle constant. the target feature fusion index is specifically: ， where M is the number of target recognition sensors, SNR j is the signal-to-noise ratio of the jth sensor, ω j is the jth sensor weight, is the target measured motion vector, is the motion prediction vector, v max is the maximum tracking speed of the device; S4, the communication relay terminal generates a hierarchical early warning instruction according to the dynamic threat index; S5, the motion monitoring terminal adjusts device motion parameters based on the early warning instruction and the dynamic threat index, and executes an obstacle avoidance strategy; S6, the weight optimization terminal dynamically updates weight coefficients in the multi-modal risk prediction model based on the dynamic threat index. 2.The real-time data analysis method based on multi-sensor fusion for mobile investigation equipment according to claim 1, characterized in that, The environment factor data comprises illumination intensity, temperature, humidity, atmospheric pressure, wind speed, electromagnetic field intensity and dust concentration; the motion state data comprises three-dimensional acceleration, angular velocity, geomagnetic azimuth angle and GPS positioning coordinates; and the target feature data comprises a signal-to-noise ratio of a sensor and a measured motion vector. 3.The real-time data analysis method based on multi-sensor fusion for mobile investigation equipment according to claim 1, characterized in that, The multi-modal risk prediction model is specifically: ， wherein Q=alpha*(beta1*C1+beta2*C2+beta3*C3-gamma), C1 is the environment interference suppression coefficient, C2 is the motion trajectory confidence degree, C3 is the target feature fusion index, alpha is a scale factor, gamma is a risk judgment threshold value, e is a natural constant, beta1, beta2 and beta3 are weight coefficients, beta1+beta2+beta3=1 and beta1, beta2 and beta3 belong to the range of [0, 1]. 4.The real-time data analysis method based on multi-sensor fusion for mobile investigation equipment according to claim 1, characterized in that, The hierarchical early warning instruction is: when the dynamic threat index E is greater than or equal to E1, a first-level early warning instruction is triggered, and an emergency braking and defense protocol is started; when the dynamic threat index E2 is less than or equal to E and E is less than E1, a second-level early warning instruction is triggered, and device deceleration and path re-planning are activated; when the dynamic threat index 0 is less than or equal to E and E is less than E2, a third-level early warning instruction is triggered, and an environment monitoring enhancement mode and a basic avoidance strategy are enabled.

5. The real-time data analysis method of a mobile investigation device based on multi-sensor fusion according to claim 1, characterized in that, The execution process of the motion monitoring terminal is: A1, the early warning instruction level is analyzed, and a preset motion parameter mapping table is called; A2, an obstacle avoidance vector is calculated based on the dynamic threat index: ， wherein λ is an obstacle avoidance gain coefficient, is a target measured motion vector, is a motion prediction vector; A3, a PID controller is used to adjust motor torque and steering angle, so that the device motion trajectory meets: ， wherein a limit is a device final acceleration upper limit preset value, a max is a device maximum acceleration preset value, E is a dynamic threat index, and min is a minimum function. 6.The real-time data analysis method based on mobile investigation equipment with multi-sensor fusion of claim 1, wherein, The dynamic update of the weight coefficients is specifically: ， wherein , ， ， ， wherein e is a natural constant, is the dynamically updated maximum weight coefficient, is the intermediate weight scaled by the dynamic threat index, is the weight corrected by short-term deviation, m(E) is a threat response intensity coefficient, η(E) is a weight adjustment rate, η0 is a baseline learning rate, C i is a real-time evaluation coefficient, is a coefficient moving average value, is a coefficient fluctuation degree, E is a dynamic threat index, wherein i = 1, 2, 3.

Images