Adaptive enhancement method for sensor data in high-dust environment

Abstract

The invention belongs to the technical field of signal processing, and particularly relates to an adaptive enhancement method for sensor data in a high-dust environment, which comprises the following steps: acquiring original data of an underground multi-source sensor and carrying out space-time registration to obtain aligned multi-channel data frames; obtaining the infrared reliability based on the local standard deviation of the infrared channel data and the visible light channel data and the average backscattering intensity of the visible light channel; dynamically fusing the short-wave infrared and visible light data by using infrared reliability to obtain a pre-fused signal field, and performing edge enhancement according to a spatial distance gradient mode length to obtain enhanced feature data; counting the average intensity value of the enhanced feature data in the forward view field feature region, and resolving a speed control instruction in combination with the infrared reliability; and the input voltage frequency of a chassis driving motor and the opening degree of a brake hydraulic valve are adjusted according to the instruction. According to the method, the problems of low data contrast and feature loss in a high-dust environment are effectively solved, and equipment adaptive safety control based on environmental perception is realized.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology. More specifically, this invention relates to an adaptive enhancement method for sensor data in high-dust environments. Background Technology

[0002] Downhole unmanned operation systems generally rely on the real-time acquisition and processing of multi-source sensing data to support autonomous navigation and operational decisions. However, in actual mining and transportation conditions, there are generally high concentrations of suspended particulate matter and unstructured interference in the environment, which poses a severe challenge to the acquisition and transmission of sensor data. Particles in the air produce strong scattering and attenuation effects on data signals, causing problems such as a sharp drop in signal-to-noise ratio, compression of dynamic range, and loss of effective information entropy in the data stream acquired by conventional optical sensors. This makes it difficult for the back-end processor to extract high-confidence feature vectors from the degraded data.

[0003] To compensate for the perception limitations of a single data source in a specific spectrum, short-wave infrared (SWIR) sensing technology has been introduced into data acquisition systems. Due to its longer wavelength, the SWIR band exhibits a lower signal attenuation rate when penetrating small particle media, enabling it to capture the physical characteristic signals of obscured targets. However, limited by sensor manufacturing processes, SWIR data channels generally suffer from low spatial resolution, significant thermal noise interference, and a lack of high-frequency detail information, making it difficult to use as an independent input source for high-precision navigation and control systems.

[0004] Existing multi-sensor data fusion (MSDF) technologies aim to integrate data streams from different spectra through algorithmic models. Common methods include signal-level fusion based on weighted averages and decision-level fusion based on feature vector concatenation. However, existing data processing architectures often use preset static weight matrices or fixed logical rules for channel synthesis, lacking a dynamic quantitative evaluation mechanism for real-time environmental noise levels. When downhole dust concentration fluctuates drastically, static calculation models cannot adaptively adjust the contribution ratio of each data channel, easily leading to the introduction of redundant noise or loss of key gradient features in the synthesized data. Furthermore, while complex calculation models based on deep neural networks (DNNs) can improve data processing quality, their high computational cost and inference latency make it difficult to meet the stringent requirements of downhole edge computing nodes for real-time performance and low power consumption. Moreover, existing sensing data processing flows are often disconnected from the underlying motion control logic, lacking a closed-loop control mechanism that can directly modulate actuator parameters based on data reliability. Summary of the Invention

[0005] To address the technical problems of signal quality degradation from multi-source sensor data in high-dust environments, the lack of environmental adaptability in existing static fusion calculation models, and the disconnect between sensing data and motion control logic, this invention provides an adaptive enhancement method for sensor data in high-dust environments. The method includes: acquiring raw data from downhole multi-source sensors and performing spatiotemporal registration to obtain aligned multi-channel data frames; extracting infrared and visible light channel data from the multi-channel data frames and obtaining infrared reliability based on the infrared and visible light channel data; fusing the infrared and visible light channel data using the infrared reliability to obtain a pre-fused signal field; performing edge enhancement based on the spatial distance gradient magnitude of the pre-fused signal field to obtain enhanced feature data; obtaining the average intensity value of the enhanced feature data within the forward field of view feature area; obtaining speed control commands based on the average intensity value and infrared reliability; and adjusting the input voltage frequency of the chassis drive motor and the opening degree of the brake hydraulic valve according to the speed control commands.

[0006] This invention acquires raw data from multiple downhole sensors and performs spatiotemporal registration. It then calculates infrared reliability using infrared and visible light channel data to dynamically adjust the fusion weights. Additionally, it combines the spatial distance gradient modulus of the pre-fused signal field for edge enhancement. Finally, it adjusts the chassis drive motor and brake hydraulic valve based on the average intensity value within the forward field of view feature area and the infrared reliability. This approach preserves high-frequency details of visible light while introducing the penetrating characteristics of short-wave infrared, improving imaging clarity in high-dust environments. Furthermore, it enables adaptive speed adjustment based on infrared reliability and obstacle conditions, ensuring the safe operation of the equipment in complex downhole environments.

[0007] Preferably, the spatiotemporal registration includes: using feature-point-based automatic registration technology to perform geometric correction on the data collected by the short-wave infrared optical sensor and the visible light optical sensor, and to align the timestamps of the collected data.

[0008] This invention utilizes feature-point-based automatic registration technology to perform geometric correction on data collected by short-wave infrared optical sensors and visible light optical sensors, and uses a hardware-triggered synchronization protocol for timestamp alignment. This eliminates spatial misalignment and time delay caused by differences in installation location and acquisition frequency between multiple source sensors, ensuring that the fusion algorithm processes multi-channel data frames that are spatially aligned at the same time. This reduces ghosting or blurring phenomena caused by data asynchrony in the fusion matrix, providing an accurate data foundation for subsequent feature extraction.

[0009] Preferably, the infrared reliability satisfies the expression: In the formula, for Infrared reliability at all times; The base gain scaling factor; for Local standard deviation of real-time infrared channel data; for Local standard deviation of visible light channel data at any given time; To prevent the minimum value where the denominator is zero; It is a nonlinear response exponent; This is the scattering sensitivity coefficient; for The average backscattering intensity of the visible light channel at any given time; This serves as a reference value for the signal response intensity in a clear environment. It is a natural constant; For time step; The symbol is for logarithms; It is a natural exponential function.

[0010] This invention establishes an infrared reliability expression that includes local standard deviation and the average backscattering intensity of the visible light channel. The local standard deviation reflects the change in matrix contrast, and the average backscattering intensity of the visible light channel is used to assess dust concentration. This automatically increases the infrared reliability when dust concentration increases or lighting conditions deteriorate, making the fusion result more dependent on infrared channel data. This reduces signal matrix whitening and detail loss caused by visible light scattering, ensuring consistent imaging quality under different dust concentration environments.

[0011] Preferably, the base gain scaling factor is set to 1.0.

[0012] Preferably, the enhanced feature data satisfies: In the formula, for Enhanced feature data output at each time step; for Normalized data intensity of the infrared channel at any given time; for Normalized data intensity of the visible light channel at any given time; This is a weighted bias term for the infrared spectrum; This refers to the edge enhancement amplitude coefficient. Gradient sensitivity factor; express Spatial distance gradient of the pre-fused signal field at any time The modulus length; for Infrared reliability at all times; It is the hyperbolic tangent function; This is the modulus symbol.

[0013] This invention introduces an enhanced feature data expression that includes the hyperbolic tangent function and the spatial distance gradient magnitude. By utilizing the nonlinear saturation characteristics of the hyperbolic tangent function to constrain the gradient magnitude, it enhances the contrast between the roadway wall and equipment contour edges in the pre-fused signal field while suppressing the excessive amplification of high-frequency noise. This results in the output matrix reducing the generation of false high-frequency details while maintaining contour clarity, thereby improving the ability of the subsequent vision system to recognize the geometry of obstacles.

[0014] Preferably, the weighted bias term of the infrared spectrum is set to 0.6.

[0015] Preferably, the speed control command satisfies the expression: In the formula, for Speed ​​control commands at any given moment; The maximum cruising speed allowed by the device; for The average intensity value of the enhanced feature data within the forward field of view feature region at any given moment; For obstacle braking sensitivity; Environmental risk coefficient; for Infrared reliability at all times; It is a natural exponential function.

[0016] This invention establishes a speed control command expression that associates the maximum cruising speed, average intensity value, and infrared reliability. It maps the obstacle intensity within the forward field of view feature area and the infrared reliability of the environment into the speed control logic. When an increase in obstacle intensity or an increase in infrared reliability indicates environmental deterioration, the command value is smoothly reduced by utilizing the decay characteristics of the natural exponential function. This enables early deceleration when visibility is obstructed or there is a potential collision risk, thereby reducing the risk of collision accidents involving underground mobile equipment.

[0017] Preferably, the obstacle braking sensitivity is set to 3.5.

[0018] Preferably, the forward field of view feature region is mapped to a matrix plane based on the vehicle width and braking distance.

[0019] Preferably, adjusting the input voltage frequency of the chassis drive motor and the opening degree of the brake hydraulic valve includes: when the infrared reliability is high and the average intensity value in the forward field of view feature area exceeds a preset safety threshold, adjusting the input voltage frequency of the chassis drive motor and the opening degree of the brake hydraulic valve using the vehicle controller.

[0020] The beneficial effects of this invention are as follows: This invention analyzes the average backscattering intensity and local contrast of the visible light channel to obtain infrared reliability that reflects the current infrared reliability status. Based on this, it dynamically adjusts the fusion ratio of infrared data and visible light data, solving the problem of poor adaptability of traditional fixed-weight fusion methods when dust concentration fluctuates. This enables the equipment to retain high-resolution high-frequency details in low-dust environments and to use infrared characteristics to penetrate obstructions in high-dust environments, thereby always obtaining usable environmental perception data in the ever-changing underground working environment.

[0021] This invention utilizes an edge enhancement mechanism combining spatial distance gradient magnitude and nonlinear functions. Addressing the uneven illumination and complex signal intensity distribution in underground roadways, it selectively sharpens key geometric features in the pre-fused signal field. This enhances the clarity of roadway walls, personnel, and equipment outlines while using energy constraints to limit thermal noise interference, generating enhanced feature data with sharp edges and a suitable signal-to-noise ratio. This provides more discernible visual information for target recognition and path planning in underground unmanned driving systems.

[0022] This invention achieves closed-loop linkage between the perception front end and the control back end by deeply coupling the visual features of the matrix domain with the chassis drive logic of the control domain. It directly uses the average intensity value of the forward field of view feature area in the enhanced feature data and the environmental infrared reliability to generate speed control commands, enabling the mobile device to automatically adjust its driving speed according to the changing trend of infrared reliability and the potential threat of obstacles in front. This reduces the risk of underground transportation safety accidents caused by human operation delays or visual blind spots in extremely dusty environments. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the adaptive enhancement method for sensor data in a high-dust environment according to the present invention; Figure 2 This is a schematic diagram illustrating the changes in environmental dust intensity and infrared reliability over time. Figure 3 This is a schematic diagram illustrating the normalized intensity distribution across the edge of an obstacle; Figure 4 This is a schematic diagram illustrating the adjustment of equipment travel speed in a high-dust environment. Detailed Implementation

[0024] 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, not all, of the embodiments of the present invention. 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.

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] This invention discloses an adaptive enhancement method for sensor data in high-dust environments, referring to... Figure 1 This includes steps S1 to S4: S1. Acquire raw data from downhole multi-source sensors and perform spatiotemporal registration to obtain aligned multi-channel data frames.

[0027] It should be noted that the present invention aims to acquire sensing information with different physical characteristics by arranging short-wave infrared optical sensors and visible light optical sensors on underground mobile equipment, thereby solving the problem of inaccurate sensing in the complex environment of coal mine roadways. Since the viewing angle range and data acquisition speed of different sensors are inconsistent, the present invention utilizes existing feature alignment and time alignment technologies to ensure that infrared and visible light data are completely consistent in space and time, generating a standardized data stream that can be directly processed.

[0028] Specifically, this invention acquires raw data from downhole multi-source sensors, uses feature-point-based automatic registration technology to perform geometric correction on the data collected by short-wave infrared optical sensors and visible light optical sensors, and uses a hardware-triggered synchronization protocol to perform timestamp alignment to acquire aligned multi-channel data frames.

[0029] S2. Extract the infrared channel data and visible light channel data from the multi-channel data frame, and obtain the infrared reliability based on the infrared channel data and visible light channel data.

[0030] It should be noted that dust in the tunnel is randomly scattered, causing the signal matrix of the visible light optical sensor to be blurred to varying degrees at different locations. This phenomenon can interfere with subsequent identification and judgment. This invention can evaluate the usability of the current visible light data in real time by calculating the contrast strength of the local area of ​​the matrix and the average backscattering intensity of the visible light channel, thereby determining how much weight needs to be added to the infrared data so that the material can be seen clearly in dense smoke or heavy dust.

[0031] Specifically, the average backscattering intensity of the visible light channel for all data points in the visible light channel at the current moment is calculated and used as the average backscattering intensity of the visible light channel; the infrared confidence level is obtained by using the infrared channel data and visible light channel data in the aligned multi-channel data frame, and the infrared confidence level satisfies the expression:

[0032] In the formula, express Infrared reliability at all times; Indicates the base gain scaling factor; express Local standard deviation of real-time infrared channel data; express Local standard deviation of visible light channel data at any given time; This indicates a minimum value to prevent the denominator from being zero; Indicates the nonlinear response index; This represents the scattering sensitivity coefficient; express The average backscattering intensity of the visible light channel at any given time; This represents the reference signal response intensity benchmark value under clear conditions.

[0033] In the formula, the local standard deviation of the visible light channel data on the left side of the expression is... A smaller value indicates lower signal matrix contrast and more severe dust obstruction, which increases the value of the fraction, thereby driving the infrared reliability. Increase; the average backscattering intensity of the visible light channel on the right side of the expression. The larger the value, the stronger the light scattering and the more blurred the environment. The exponentiation reduces the denominator, thus decreasing the infrared reliability. This increases the value, thus providing a weight reference for subsequent processing.

[0034] Furthermore, in this invention, the reference signal response intensity benchmark value The base gain scaling factor is calculated using the average of historical data collected in a cleanroom environment. Setting it to 1.0 will prevent the minimum value where the denominator is zero. Set as .

[0035] Furthermore, the present invention sets a nonlinear response index. The range is 1.5 to 3.0, and the scattering sensitivity coefficient is set. The range is from 0.5 to 2.0; if the nonlinear response index Selecting an excessively large value will affect the infrared reliability. Numerical oscillations occur during calculations, if the scattering sensitivity coefficient... If the selected value is too small, it cannot accurately reflect subtle changes in the tunnel environment; this invention selects... It is 2.2 and A value of 1.0 ensures that the infrared reliability remains stable even when dust concentration changes abruptly. It can quickly follow without significant fluctuations, improving the consistency of data enhancement.

[0036] For example, Figure 2This is a schematic diagram showing the change of environmental dust intensity and infrared reliability over time. The diagram shows that as the average backscattering intensity of the visible light channel increases, the infrared reliability calculated by the system increases synchronously. This indicates that the algorithm can perceive and identify the degree of environmental degradation in real time and automatically enhance the importance of the infrared channel in subsequent fusion.

[0037] S3. The infrared channel data and the visible light channel data are fused through infrared reliability to obtain a pre-fused signal field; edge enhancement is performed based on the spatial distance gradient modulus of the pre-fused signal field to obtain enhanced feature data.

[0038] It should be noted that if infrared and visible light data are simply superimposed, the thermal noise inherent in the infrared matrix will be amplified, and the originally clear lines of the visible light may be lost. This invention, by adding edge energy constraints, can more clearly delineate the outlines of personnel, tunnel walls, and machines while penetrating dust.

[0039] Specifically, this invention combines infrared reliability to perform fusion calculations on multi-source channel data to obtain enhanced feature data.

[0040]

[0041] In the formula, express Enhanced feature data output at each time step; express Normalized data intensity of the infrared channel at any given time; express Normalized data intensity of the visible light channel at any given time; Represents the weighted bias term of the infrared spectrum; Indicates the edge enhancement amplitude coefficient; This represents the gradient sensitivity factor; express Spatial distance gradient of the pre-fused signal field at any time The length of the module.

[0042] In the formula, the first part of the expression utilizes infrared confidence level. Adjust the proportion of the infrared component; when the infrared reliability... As the value increases, the weight of the infrared component becomes higher, thus extracting infrared thermal features in dusty environments; the latter part of the expression uses the hyperbolic tangent function to adjust the spatial distance gradient magnitude. Processing is performed when the spatial distance gradient magnitude is... When the value increases, it indicates the presence of a clear contour, and the calculated enhancement feature data... The value of will also increase, resulting in higher contrast at the edges of the signal matrix.

[0043] Furthermore, the present invention uses the weighted bias term of the infrared spectrum. Setting it to 0.6 will adjust the gradient sensitivity factor. Set to 1.0 meter.

[0044] Furthermore, the present invention sets an edge enhancement amplitude coefficient. The range is 0.1 to 0.5; if the edge enhancement amplitude coefficient If the value is too small, the sharpening effect on the edges of the signal matrix will be insignificant; if the value is too large, false ghosting of lines will occur. This invention selects... The value is 0.3, which can improve the clarity of the outline while avoiding the introduction of too much interference noise, making the shape of the object in the well easier to identify.

[0045] For example, Figure 3 This is a schematic diagram of the normalized intensity distribution across the edge of an obstacle. The diagram shows that the enhanced feature data has higher contrast than the original visible light channel data and short-wave infrared channel data at locations where the data position changes drastically. Its curve transitions more sharply and the waveform is smoother at the edge, indicating that the method effectively compensates for the contour blur caused by dust scattering through edge energy constraints.

[0046] S4. Obtain the average intensity value of the enhanced feature data within the forward field of view feature area, and obtain the speed control command based on the average intensity value and infrared reliability; adjust the input voltage frequency of the chassis drive motor and the opening degree of the brake hydraulic valve according to the speed control command.

[0047] It should be noted that the computing resources of underground mobile equipment are limited. This invention performs data statistics by monitoring only the core area in front of the vehicle, aiming to achieve the safest obstacle avoidance logic with the least amount of computation. By combining the current dust interference situation, this invention can assess the safe driving speed of the equipment in real time and ensure that it can automatically slow down and stop when visibility is poor.

[0048] Specifically, the present invention obtains the average intensity value of the enhanced feature data within the forward field of view feature region, and combines it with the infrared reliability to calculate the speed control command.

[0049] The speed control command satisfies the expression:

[0050] In the formula, for Speed ​​control commands at any given moment; The maximum cruising speed allowed by the device; for The average intensity value of the enhanced feature data within the forward field of view feature region at any given moment; For obstacle braking sensitivity; Environmental risk coefficient; for Infrared reliability at all times; It is a natural exponential function.

[0051] In the formula, the average strength value A larger value indicates a higher probability of an obstacle ahead, leading to a larger absolute value of the negative exponent and thus affecting the speed control command. Reduce; at the same time, infrared reliability A higher value indicates a higher concentration of dust in the environment, which increases braking force and forces the equipment to decelerate earlier in poor conditions, thus protecting the safety of downhole operations.

[0052] Furthermore, in this invention, the forward field-of-view feature region is pre-calibrated and calculated based on a pinhole optical sensor model and "perspective projection transformation technology"; environmental risk coefficient. Set to 0.5. This sets the obstacle braking sensitivity. The range is 2.0 to 5.0; if the obstacle braking sensitivity is... If the selected value is too large, the equipment will frequently brake suddenly, leading to instability; if the selected value is too small, there is a risk of not being able to brake in time. This invention selects... With a speed of 3.5, it balances driving smoothness and safety, and enables intelligent speed adjustment in dusty environments.

[0053] Furthermore, the present invention utilizes infrared reliability. and speed control commands An adaptive enhancement method for sensor data in high-dust environments was achieved by directly adjusting the frequency of the chassis drive motor and the opening of the brake valve through the vehicle controller.

[0054] For example, Figure 4 This is a schematic diagram of the equipment's speed adjustment in a high-dust environment. The diagram shows that during operation, due to the combined effects of increased ambient dust concentration and the intensity of forward-detected targets, the speed control command is actively reduced relative to the equipment's maximum allowable cruising speed. After the risk is eliminated, the speed command smoothly recovers, demonstrating the system's adaptive obstacle avoidance control capability under complex working conditions.

Claims

1. A method for adaptive augmentation of sensor data in high dust environments, characterized in that, The method comprises the following steps: acquiring original data of downhole multi-source sensors and performing spatiotemporal registration to obtain an aligned multi-channel data frame; extracting infrared channel data and visible light channel data from the multi-channel data frame, and acquiring infrared reliability based on the infrared channel data and the visible light channel data; fusing the infrared channel data and the visible light channel data based on the infrared reliability to obtain a pre-fusion signal field; performing edge enhancement based on a spatial distance gradient module length of the pre-fusion signal field to obtain enhanced feature data; acquiring an average intensity value of the enhanced feature data in a forward field of view feature area, and acquiring a speed control instruction based on the average intensity value and the infrared reliability; and adjusting an input voltage frequency of a chassis drive motor and an opening degree of a brake hydraulic valve based on the speed control instruction.

2. The method for adaptive augmentation of sensor data in high dust environments of claim 1, wherein, The spatiotemporal registration comprises: performing geometric correction on data collected by a short-wave infrared optical sensor and a visible light optical sensor by using an automatic registration technology based on feature points, and performing timestamp alignment on the collected data.

3. The method for adaptive augmentation of sensor data in high dust environments of claim 1, wherein, The infrared reliability satisfies the expression: ； wherein is the infrared reliability at time instant is the base gain scaling factor; is the local standard deviation of the infrared channel data at time instant is the local standard deviation of the visible light channel data at time instant is a small value to prevent division by zero; is the non-linear response exponent; is the scattering sensitivity coefficient; is the average backscatter intensity of the visible light channel at time instant is the reference signal response intensity baseline value in clear conditions; is the natural constant; is the time step; is the logarithmic sign; is the natural exponential function.

4. The method for adaptive augmentation of sensor data in high dust environments of claim 3, wherein, The basic gain scaling factor is set to 1.

0.

5. The method for adaptive augmentation of sensor data in high dust environments of claim 1, wherein, The enhanced feature data satisfies: ； wherein is the enhanced feature data output at the time instant; is the normalized data intensity of the infrared channel at the time instant; is the normalized data intensity of the visible light channel at the time instant; is a weighted bias term for the infrared spectrum; is an edge enhancement amplitude coefficient; is a gradient sensitivity factor; denotes the spatial distance gradient of the pre-fusion signal field at the time instant is the modulus of the spatial distance gradient of the pre-fusion signal field at the time instant is the infrared reliability at the time instant; is a hyperbolic tangent function; is the modulus sign.

6. The method for adaptive augmentation of sensor data in high dust environments of claim 5, wherein, The weighted bias term of the infrared spectrum is set to 0.

6.

7. The method for adaptive augmentation of sensor data in high dust environments of claim 1, wherein, The speed control instruction satisfies the expression: ； wherein is the speed control command at the instant; is the maximum cruise speed allowed by the device; is is the average intensity value of the enhanced feature data within the forward field of view feature zone at the instant; is the obstacle braking sensitivity; is the environmental risk coefficient; is is the infrared reliability at the instant; is a natural exponential function.

8. The method for adaptive augmentation of sensor data in high dust environments of claim 7, wherein, The obstacle braking sensitivity is set to 3.

5.

9. The method for adaptive augmentation of sensor data in high dust environments of claim 7, wherein, The forward field of view feature area is mapped to a matrix plane according to a vehicle width and a braking distance.

10. The method for adaptive augmentation of sensor data in high dust environments of claim 1, wherein, The adjustment of the input voltage frequency of the chassis drive motor and the opening degree of the brake hydraulic valve comprises: when it is detected that the infrared reliability is large and the average intensity value in the forward field of view feature area exceeds a preset safety threshold, adjusting the input voltage frequency of the chassis drive motor and the opening degree of the brake hydraulic valve by using a vehicle-mounted controller.

Images