A Method and System for Detecting Scattered Foreign Objects on the Ground Based on Passive Radio Frequency Envelope Fingerprinting

Abstract

This invention relates to the field of foreign object detection technology, and in particular to a method and system for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting. First, envelope detection is performed on non-cooperative radio frequency signals to construct a discrete envelope voltage sequence. Second, time-series transformation is performed based on the discrete envelope voltage sequence and an environmental statistical fingerprint database to obtain an envelope time-series transformation sequence. Then, a foreign object presence confidence level is generated based on the envelope time-series transformation sequence. Finally, the ground monitoring result is determined based on the foreign object presence confidence level and alarm conditions. The solution disclosed in this invention effectively solves the background drift and feature mismatch problems of passive radio frequency sensing in real-time edge deployment while achieving low-cost, lightweight edge deployment.

Description

Technical Field

[0001] This invention relates to the field of foreign object monitoring technology, and in particular to a method and system for monitoring scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting. Background Technology

[0002] In indoor or semi-enclosed environments, scattered objects such as transparent water bottles, cables, and low obstacles on the ground can easily cause service robots to get stuck or people to trip, posing common safety risks. Current monitoring methods largely rely on cameras, LiDAR, or contact sensors. However, cameras are limited by lighting and privacy compliance requirements; LiDAR has the risk of missing detections of transparent or ground-hugging flexible objects, and is also costly and consumes a lot of power; contact-based solutions suffer from latency.

[0003] Passive radio frequency sensing inherently possesses advantages in protecting privacy and resisting optical interference. However, existing technologies struggle to meet the demands of lightweight indoor edge deployments in terms of hardware complexity and target adaptability. For instance, invention patents CN105158727B and CN108875584A propose sensing schemes based on wireless channel state information (CSI) or wireless fingerprinting, but these require acquiring physical layer CSI data and performing multi-dimensional matrix processing, primarily targeting large-area human positioning or behavior recognition. Invention patent CN108398676B proposes an external radiation source radar detection method, requiring the establishment of reference and echo dual channels, high-speed IQ sampling, and complex coherent digital processing. The aforementioned schemes suffer from complex hardware links and high computational overhead, making it difficult to form low-cost, low-power micro-monitoring nodes.

[0004] Furthermore, in the fields of robotic inspection and machine vision, invention patent CN110703760A relies on security inspection robots and PTZ cameras to identify newly added suspicious objects, while invention patent CN112347876B depends on TOF depth cameras to identify obstacles. These solutions are highly dependent on mobile platforms and optical imaging links, resulting in high system integration and equipment costs, and are not suitable for large-scale, distributed fixed grid deployment in privacy-sensitive areas such as bedrooms and elderly care facilities.

[0005] In summary, existing solutions have their own advantages in human body localization, radar detection, or robot obstacle avoidance scenarios. However, in long-term, non-intrusive ground monitoring, they are prone to problems such as hardware complexity, large node size, high deployment costs, or privacy restrictions. At the same time, existing passive radio frequency solutions are prone to background drift and feature mismatch when deployed in real time at the edge. Summary of the Invention

[0006] The purpose of this invention is to provide a ground-based scattered foreign object monitoring method and system based on passive radio frequency envelope fingerprinting, so as to achieve low-cost, miniaturized, lightweight and low-power distributed indoor deployment, and overcome the problems of background drift and feature mismatch that easily occur when deploying at the edge in real time.

[0007] To achieve the above objectives, the present invention provides a method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting, the method comprising:

[0008] Acquire non-cooperative radio frequency signals containing spatial multipath disturbances at the locations of each edge node in the ground area to be monitored;

[0009] Envelope detection is performed on the non-cooperative radio frequency signal to construct a discrete envelope voltage sequence;

[0010] Based on the discrete envelope voltage sequence and the environmental statistical fingerprint database, a time series transformation is performed to obtain the envelope time series transformation sequence;

[0011] The confidence level for the presence of foreign matter is generated based on the envelope time-series transformation sequence.

[0012] The ground monitoring results are determined based on the confidence level of the presence of the foreign object and the alarm conditions.

[0013] Optionally, the step of constructing an envelope detection sequence from the non-cooperative radio frequency signal to obtain a discrete envelope voltage sequence specifically includes:

[0014] Frequency selection and filtering out interference signals in the non-cooperative radio frequency signals yields interference-free radio frequency signals near the target receiving frequency.

[0015] The gain of the interference-free radio frequency signal is adjusted to the effective input range to obtain the radio frequency power signal;

[0016] The radio frequency power signal is converted into a continuous envelope voltage signal;

[0017] The continuous envelope voltage signal is acquired using a sampling rate lower than the radio frequency carrier frequency to obtain a discrete envelope voltage sequence.

[0018] Optionally, the step of performing time-series transformation based on the discrete envelope voltage sequence and the environmental statistical fingerprint database to obtain an envelope time-series transformation sequence specifically includes:

[0019] The discrete envelope voltage sequence is processed by a filter to obtain a filtered envelope voltage sequence;

[0020] The filtered envelope voltage sequence is subjected to hardware transient truncation to obtain the truncated voltage sequence;

[0021] The cutoff voltage sequence is dynamically standardized and calibrated based on the environmental statistical fingerprint database to obtain the envelope timing transformation sequence.

[0022] Optionally, the step of dynamically standardizing and calibrating the truncated voltage sequence based on an environmental statistical fingerprint database to obtain an envelope time-series transformation sequence specifically includes:

[0023] Read the background mean and background scale that match the current node, current receiving frequency, candidate ground grid or path segment from the environmental statistical fingerprint database;

[0024] Based on the background mean and the background scale, the cutoff voltage sequence is time-calibrated to obtain an envelope time-calibration sequence.

[0025] The envelope timing calibration sequence is subjected to amplitude clipping or piecewise linear compression to obtain the envelope timing transformation sequence.

[0026] Optionally, generating the foreign object presence confidence based on the envelope time-series transformation sequence specifically includes:

[0027] The envelope time-series transformation sequence is segmented according to a preset window length and sliding step size to obtain time-series segmentation features;

[0028] The temporal segmentation features are input into the lightweight foreign object detection module to obtain the confidence level of foreign object presence.

[0029] Optionally, determining the ground monitoring results based on the confidence level of the foreign object's presence and the alarm conditions specifically includes:

[0030] The confidence of foreign object presence is weighted and fused for continuous time windows and for different edge nodes and / or different receiving frequencies to obtain a fused confidence.

[0031] Determine whether the fusion confidence score is less than a first threshold. If the fusion confidence score is less than the first threshold, output that there are no scattered foreign objects on the ground, and allow the background fingerprints of the corresponding candidate ground grids, path segments, edge nodes, or receiving frequencies to be updated according to the gating update rules. If the fusion confidence score is greater than or equal to the first threshold and less than or equal to the second threshold, enter the observation state and continue to accumulate the confidence score of subsequent time windows. If the fusion confidence score is greater than the second threshold, determine whether the alarm condition is met. If the alarm condition is met, output that there are scattered foreign objects on the ground and the candidate ground area, and issue an alarm. If the alarm condition is not met, enter the suspected observation state and pause the updating of the background fingerprints of the corresponding candidate ground grids, path segments, edge nodes, or receiving frequencies according to the gating update rules. The second threshold is greater than the first threshold.

[0032] This invention also provides a ground-based foreign object monitoring system based on passive radio frequency envelope fingerprinting, the system comprising: at least one edge node and an edge processor; each edge node comprising: a receiving antenna and an envelope detection construction module; the receiving antenna being connected to the edge processor via the envelope detection construction module;

[0033] The receiving antenna is used to receive non-cooperative radio frequency signals containing spatial multipath disturbances in the monitoring state;

[0034] The envelope detection construction module is used to construct the envelope of non-cooperative radio frequency signals to obtain discrete envelope voltage sequences.

[0035] The edge processor is used to perform time-series transformation based on the discrete envelope voltage sequence and the environmental statistical fingerprint database to obtain an envelope time-series transformation sequence; generate a foreign object presence confidence level based on the envelope time-series transformation sequence; and determine the ground monitoring results based on the foreign object presence confidence level and alarm conditions.

[0036] Optionally, the envelope detection construction module includes: a frequency selective filter, an RF amplifier, an envelope detector, and an analog-to-digital converter, and the receiving antenna is connected in sequence through the frequency selective filter, the RF amplifier, the envelope detector, the analog-to-digital converter, and the edge processor;

[0037] The frequency-selective filter is used to select the frequency and filter out interference signals in the non-cooperative radio frequency signal to obtain an interference-free radio frequency signal near the target receiving frequency point; the radio frequency amplifier is used to adjust the gain of the interference-free radio frequency signal to the effective input range to obtain a radio frequency power signal; the envelope detector is used to convert the radio frequency power signal into a continuous envelope voltage signal; and the analog-to-digital conversion unit is used to acquire the continuous envelope voltage signal at a sampling rate lower than the radio frequency carrier frequency to obtain a discrete envelope voltage sequence.

[0038] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting.

[0039] The present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting.

[0040] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0041] This invention discloses a method and system for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting. First, envelope detection is performed on non-cooperative radio frequency signals to construct a discrete envelope voltage sequence. Second, time-series transformation is performed based on the discrete envelope voltage sequence and an environmental statistical fingerprint database to obtain an envelope time-series transformation sequence. Then, a foreign object presence confidence level is generated based on the envelope time-series transformation sequence. Finally, the ground monitoring result is determined based on the foreign object presence confidence level and alarm conditions. The scheme disclosed in this invention effectively solves the background drift and feature mismatch problems of passive radio frequency sensing in real-time edge deployment while achieving low-cost, lightweight edge deployment. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a ground-scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting, according to an embodiment of the present invention.

[0044] Figure 2 This is a structural diagram of the lightweight foreign object detection module according to an embodiment of the present invention;

[0045] Figure 3 This is a structural diagram of a ground-scattered foreign object monitoring system based on passive radio frequency envelope fingerprinting, according to an embodiment of the present invention.

[0046] Among them, 101 is the receiving antenna, 102 is the frequency selective filter, 103 is the radio frequency amplifier, 104 is the envelope detector, 105 is the analog-to-digital conversion unit, and 106 is the edge processor. Detailed Implementation

[0047] 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.

[0048] The purpose of this invention is to provide a ground-based scattered foreign object monitoring method and system based on passive radio frequency envelope fingerprinting, so as to achieve low-cost, miniaturized, lightweight and low-power distributed indoor deployment, and overcome the problems of background drift and feature mismatch that easily occur when deploying at the edge in real time.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] like Figure 1 As shown, this invention discloses a method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting. The method includes:

[0051] Step S1: Acquire non-cooperative radio frequency signals containing spatial multipath disturbances at the locations of each edge node in the ground area to be monitored.

[0052] Step S2: Construct envelope detection for non-cooperative radio frequency signals to obtain discrete envelope voltage sequences.

[0053] Step S3: Perform time series transformation based on discrete envelope voltage sequence and environmental statistical fingerprint database to obtain envelope time series transformation sequence.

[0054] Step S4: Generate the confidence level of foreign object presence based on the envelope time-series transformation sequence.

[0055] Step S5: Determine the ground monitoring results based on the confidence level of the presence of foreign objects and the alarm conditions.

[0056] The following is a detailed discussion of each step:

[0057] Step S1: Acquire the non-cooperative radio frequency signals containing spatial multipath disturbances at the locations of each edge node in the ground area to be monitored. Non-cooperative radio frequency signals refer to common downlink radio frequency signals that the system does not control in terms of transmission content and timing, and are passively received only by receiving antenna 101.

[0058] The specific formula for the non-cooperative radio frequency signal of this invention is as follows:

[0059] ;

[0060] in: For discrete sampling times; This is the index for the passive radio frequency edge nodes; For the receiving frequency point index; For the candidate ground grid number index; Indicates time , No. The edge node at the ... Received radio frequency signals at each frequency point; Indicates time , No. Equivalent transmitted signal of a non-cooperative radio frequency source; Represents the convolution operation between the signal and the channel impulse response; Indicates the first The edge node at the ... Strong direct path response on each frequency point; This represents the slowly time-varying background multipath response influenced by the environment; Representation of time The relevant physical state or dynamic parameters; Indicates in candidate ground grid Additional scattering response caused by scattered foreign objects; Indicates the first The edge node at the ... At each frequency point The environmental noise term is considered. This model shows that foreign objects do not need to actively emit signals; their presence alters the multipath superposition relationship at the receiver, and the system compares the received signals... Differences in envelope characteristics can be used to apply disturbances to foreign objects. The perception.

[0061] Step S2: Construct the envelope detection for the non-cooperative radio frequency signal to obtain a discrete envelope voltage sequence, specifically including:

[0062] Step S21: Select the frequency and filter out interference signals in the non-cooperative radio frequency signal to obtain an interference-free radio frequency signal near the target receiving frequency.

[0063] Step S22: Adjust the gain of the interference-free radio frequency signal to the effective input range to obtain the radio frequency power signal.

[0064] Step S23: Convert the radio frequency power signal into a continuous envelope voltage signal. The continuous envelope voltage signal is actually a continuous low-frequency envelope voltage signal. Specifically, its effective characteristic energy reflecting foreign object disturbances is mainly concentrated in the frequency range of 0.05Hz to 20Hz.

[0065] Step S24: Acquire the continuous envelope voltage signal using a sampling rate lower than the RF carrier frequency to obtain a discrete envelope voltage sequence. This step converts the continuous domain envelope voltage into a discrete envelope voltage sequence through analog-to-digital sampling. The discrete envelope voltage samples in the discrete envelope voltage sequence and the corresponding RF power samples satisfy the following mapping relationship: The specific formula is:

[0066] ;

[0067] The radio frequency power sample is calculated from the sampled values ​​of the received radio frequency signal, and the corresponding relationship is as follows:

[0068] ;

[0069] in: For discrete-time indexing, This is the index for the passive radio frequency edge node. For the receiving frequency point number index, For the first The edge node at the ... The first sampled output at the receiving frequency point in this step The discrete envelope voltage samples constitute the discrete envelope voltage sequence finally obtained in this step; For the first The edge node at the ... The first receiving frequency point Discrete radio frequency power samples corresponding to each sampling time; For the first The envelope detection slope corresponding to each receiving frequency point The detector intercept at the corresponding frequency point. To preset the reference power, For the first The edge node at the ... The first receiving frequency point The envelope front-end error of each sampling point To continuously receive radio frequency signals at time Discrete sampled values ​​at the location, This is the sampling period used in this step.

[0070] The above discrete-time index Corresponding sampling point number; passive RF edge node number index and the index of receiving frequency points These correspond to the edge node performing this signal acquisition and the current receiving frequency, respectively; The envelope detection slope corresponding to each receiving frequency point and the detector intercept at the corresponding frequency point The inherent transformation parameter of the envelope detector circuit represents the logarithmic-linear transformation relationship between the input RF power and the output envelope voltage. The edge node at the ... The first receiving frequency point Envelope front-end error at each sampling point This includes hardware circuit noise and analog-to-digital conversion quantization error.

[0071] Step S3: Perform time series transformation based on the discrete envelope voltage sequence and the environmental statistical fingerprint database to obtain the envelope time series transformation sequence, specifically including:

[0072] Step S31: Process the discrete envelope voltage sequence using a filter to obtain the filtered envelope voltage sequence. The filter in this invention uses a unidirectional causal low-pass or band-pass filter; when using a unidirectional causal filter, an infinite impulse response filter is implemented using a second-order section, therefore the unidirectional causal filter's first... The specific formula for the second-order recurrence relation is as follows:

[0073] ;

[0074] in, For discrete-time indexing; Indicates the first The second-order section in the... Input at each sampling point; Indicates the first The second-order section in the... Output of each sampling point; , , , and For the first The filter coefficients of each second-order section. When implementing causal low-pass or band-pass filtering in a cascaded manner: for the first second-order section, its input... This is the envelope voltage sequence output by the analog-to-digital converter unit 105; for subsequent second-order sections, the input of the current second-order section is the output of the previous second-order section, i.e. The output of the last second-order section is the final filtered envelope voltage sequence. The second-order section is used to retain low-frequency envelope disturbances caused by ground debris and suppress out-of-band noise.

[0075] Step S32: Perform hardware transient truncation on the filtered envelope voltage sequence to obtain the truncated voltage sequence. The specific formula is as follows:

[0076] ;

[0077] in, Indicates the first The edge node, the first The filtered envelope voltage sequence of each receiving frequency point within the original sampling segment; This indicates the total number of sampling points in the original sampling segment, i.e., the length of the original sampling segment; Indicates the number of transient sampling points that were discarded; Represents the first in the cutoff voltage sequence One sampling point, .

[0078] The original sampling segment mentioned above refers to the continuous sampling segment after one power-on acquisition, one sleep wake-up, and one frequency point switching, or the continuous data segment obtained by dividing the filtered envelope voltage sequence according to a preset time length.

[0079] Step S33: Perform dynamic normalization calibration on the truncated voltage sequence based on the environmental statistical fingerprint database to obtain the envelope time series transformation sequence, specifically including:

[0080] Step S331: Read the background mean and background scale that match the current node, current receiving frequency, candidate ground grid or path segment from the environmental statistical fingerprint database.

[0081] Step S332: Perform time-series calibration on the cutoff voltage sequence based on the background mean and background scale to obtain the envelope time-series calibration sequence. The specific formula is as follows:

[0082]

[0083] in, Represents the first in the cutoff voltage sequence The cutoff voltage corresponding to each sampling point This represents the background mean that matches the current node, frequency point, and candidate grid. This represents the background scale value of the corresponding match. This is a very small constant used to avoid division-to-zero anomalies when the background scale is too small. The first in the envelope timing calibration sequence The envelope timing calibration sample corresponding to each sampling point.

[0084] Step S333: Perform amplitude clipping or piecewise linear compression on the envelope timing calibration sequence to obtain the envelope timing transformation sequence. When amplitude clipping is used, the clipping operation satisfies the following relationship:

[0085]

[0086] in, The first in the envelope time-series transformation sequence The envelope time-series transformation samples corresponding to each sampling point To preset the cropping threshold, This is an amplitude clipping function used to limit the amplitude of the calibrated sequence to a certain value. Within the range, The first in the envelope timing calibration sequence The envelope timing calibration sample corresponding to each sampling point.

[0087] Step S4: Generate the confidence level of foreign body presence based on the envelope time series transformation sequence, specifically including:

[0088] Step S41: Divide the envelope timing transformation sequence according to the preset window length and sliding step size to obtain timing segmentation features; when there is one edge node and one receiving frequency point, each window forms a one-dimensional timing segmentation feature; when there are multiple edge nodes or multiple receiving frequency points, use the node dimension or frequency point dimension as the channel dimension to obtain multi-channel timing segmentation features.

[0089] Step S42: Input the temporal segmentation features into the lightweight foreign object discrimination module to obtain the confidence level of foreign object presence.

[0090] Step S5: Determine the ground monitoring results based on the confidence level of the foreign object's presence and the alarm conditions, specifically including:

[0091] Step S51: Weighted fusion is performed on the foreign object presence confidence corresponding to continuous time windows, different edge nodes and / or different receiving frequencies.

[0092] Step S52: Determine if the fusion confidence is less than the first threshold; if the fusion confidence is less than the first threshold, output that there are no scattered foreign objects on the ground, and allow the background fingerprints of the corresponding candidate ground grids, path segments, edge nodes, or receiving frequencies to be updated according to the gating update rules; if the fusion confidence is greater than or equal to the first threshold and less than or equal to the second threshold, enter the observation state and continue to accumulate the confidence of subsequent time windows; if the fusion confidence is greater than the second threshold, determine if the alarm conditions are met; if the alarm conditions are met, output the scattered foreign objects on the ground and the candidate ground area, and issue an alarm; if the alarm conditions are not met, enter the suspected observation state, and pause the updating of the background fingerprints of the corresponding candidate ground grids, path segments, edge nodes, or receiving frequencies according to the gating update rules; the second threshold is greater than the first threshold.

[0093] The alarm conditions must simultaneously meet the following two conditions: the first is the time continuity condition, i.e., continuous. At least one time window The fusion confidence of each window reaches the first threshold. The second condition is the consistency condition, which requires that the same candidate ground region satisfy at least one of the following: edge node consistency condition, received frequency consistency condition, and path segment consistency condition. The candidate ground region is determined by at least two of the following: node installation location, antenna orientation, received signal variation amplitude, frequency quality index, and confidence weight.

[0094] like Figure 3 As shown, this invention also discloses a ground-based foreign object monitoring system based on passive radio frequency envelope fingerprinting. The system includes at least one edge node and at least one edge processor 106. Each edge node includes a receiving antenna 101 and an envelope detection construction module. The receiving antenna 101 is connected to the edge processor 106 through the envelope detection construction module. The envelope detection construction module includes a frequency-selective filter 102, a radio frequency amplifier 103, an envelope detector 104, and an analog-to-digital converter 105. The receiving antenna 101 is sequentially connected to the edge processor 106 through the frequency-selective filter 102, the radio frequency amplifier 103, the envelope detector 104, the analog-to-digital converter 105, and the edge processor 106. One or more edge processors 106 can be configured, and multiple edge processors 106 can provide redundant backups for each other or collaboratively complete data processing across multiple nodes.

[0095] The ground area to be monitored disclosed in this invention can be an indoor environment or a semi-enclosed scene. At least one passive radio frequency edge node (hereinafter referred to as edge node) is deployed in the ground area to be monitored. The edge node can be installed in either a fixed installation or a mobile installation. In the fixed installation method, the edge node is installed in at least one of the following locations: a corner of an indoor room, the edge of an indoor corridor, the bottom of an indoor shelf, and a low position on a wall. The continuous ground area to be monitored is discretized into multiple indexable candidate ground grids, which are used to maintain the environmental envelope statistical characteristics of the corresponding ground location and output the corresponding grid location when an abnormal disturbance is detected. In the mobile installation method, the edge node is installed on at least one of the following carriers: a sweeping robot, an indoor inspection robot, and a mobile cleaning device. The travel path in the ground area to be monitored is divided into multiple path segments.

[0096] Each edge node includes a receiving antenna 101; the receiving antenna 101 may be an omnidirectional antenna or a directional antenna facing the candidate area on the ground. The passive radio frequency edge node will be referred to as an edge node in the following text.

[0097] Edge nodes do not generate active probe waves; they only receive existing public downlink radio frequency (RF) signals in the environment. These public downlink RF signals include FM broadcasting, digital audio broadcasting, terrestrial digital television, cellular communication downlink, or other long-term stable public downlink RF signals. Taking FM broadcasting as an example, the system can select one or more frequency points within the range of 87MHz to 108MHz and calculate the envelope mean, variance, and low-frequency noise energy of each frequency point within a reference time. Frequency points with stable intensity and unsaturated front-end output are selected for monitoring.

[0098] Specifically, when selecting the receiving frequency, the system can prioritize using the 87MHz to 108MHz FM broadcast frequency band; if the FM broadcast frequency is insufficient or its stability does not meet the requirements, digital audio broadcasting, terrestrial digital television, cellular communication downlink, or other long-term stable public downlink radio frequency signals can also be selected.

[0099] The quality of the receiving frequency can be determined by the envelope variance within the reference time, low-frequency noise energy, relative envelope fluctuation between nodes, and front-end saturation state; preferably, receiving frequencies with smaller envelope fluctuations and unsaturated output of envelope detector 104 are selected for monitoring.

[0100] This invention divides the ground area to be monitored into multiple candidate ground grids. These candidate ground grids serve as spatial indexes for storing background fingerprints, calculating anomaly confidence, and outputting alarm locations. In a fixed deployment, each candidate ground grid does not require a separate edge node; in practice, one edge node can cover multiple candidate ground grids, or multiple edge nodes can jointly cover the same candidate ground grid, establishing a connection through edge node numbers, receiving frequency numbers, and candidate ground grid numbers. In a mobile platform deployment, the travel path within the ground area to be monitored is divided into multiple path segments based on distance, time, or robot mileage. Wheel speed, mileage, and attitude information are used to map the acquired anomaly time windows to the corresponding path segments or candidate ground grids, thereby obtaining the approximate location of foreign objects.

[0101] When the ground area to be monitored is free of scattered foreign objects, the edge processor 106 receives the background radio frequency envelope sequence of each edge node at at least one receiving frequency point, and establishes an environmental statistical fingerprint database based on the background radio frequency envelope sequence, according to node number, receiving frequency point number, time period, candidate ground grid number, or movement path segment number. At least two of the following constitute background fingerprint items: background mean, background scale quantity, standard deviation, median, absolute median difference, preset low-frequency band energy, and inter-node envelope ratio. These background fingerprint items are stored in the environmental statistical fingerprint database in the form of a structured data table. The environmental statistical fingerprint database can be stored in the local memory of the edge processor 106, an external non-volatile memory, or a local gateway connected to the edge processor 106. The environmental statistical fingerprint database includes at least the corresponding background mean and background scale quantity for subsequent dynamic normalization calibration and background drift suppression.

[0102] The formula for calculating the background fingerprint is:

[0103] ;

[0104] in, Indicates the first The edge node, the first The receiving frequency point, the first Background fingerprint items corresponding to each candidate ground grid or path segment; This represents the mean of the background envelope voltage sequence; This represents the background scale, which can be the standard deviation or a robust scale derived from the absolute median. Indicates the background median; Indicates the absolute median difference; This indicates the preset low-frequency band energy. Indicates the first The edge node and the first The edge node at the ... The receiving frequency and the first Envelope ratio statistics on each candidate ground grid or path segment.

[0105] Let the update threshold be The freezing threshold is ,and Less than To prevent the background model from absorbing real foreign objects, the system employs a gating update procedure to update the environmental statistical fingerprint database. Specifically, this includes: if the confidence level of foreign objects in multiple consecutive monitoring windows is lower than the update threshold... Then, the background radio frequency envelope sequence is updated using an exponential moving average method; if the confidence level is higher than the freeze threshold... If so, then pause the update of the corresponding candidate ground grid, receiving frequency point, or moving path segment. If the confidence level of the foreign object presence is within a certain range... and During this period, the system remains in observation mode, with no updates or only background fingerprint updates with a low weight. The number of continuous monitoring windows can be set to 3 to 30, the update threshold can be set to 0.20 to 0.50, and the freeze threshold can be set to 0.60 to 0.95.

[0106] like Figure 3As shown, the edge node of the present invention includes: a receiving antenna 101; in addition, the edge node of the present invention also includes: an envelope detection construction module; the envelope detection construction module includes: a frequency selective filter 102, an RF amplifier 103, an envelope detector 104, and an analog-to-digital conversion unit 105. The receiving antenna 101 is connected sequentially through a frequency-selective filter 102, an RF amplifier 103, an envelope detector 104, an analog-to-digital converter 105, and an edge processor 106. The receiving antenna 101 is used to receive non-cooperative RF signals containing spatial multipath disturbances during monitoring. The frequency-selective filter 102 is used to select the frequency and filter out interference signals in the non-cooperative RF signals to obtain interference-free RF signals near the target receiving frequency. The RF amplifier 103 is used to adjust the gain of the interference-free RF signal to the effective input range to obtain an RF power signal. The envelope detector 104 is used to convert the RF power signal into a continuous envelope voltage signal. The analog-to-digital converter 105 is used to acquire the continuous envelope voltage signal at a sampling rate lower than the RF carrier frequency to obtain a discrete envelope voltage sequence. The specific sampling rate can be adjusted according to the target object size, node height, moving speed, and the processing capability of the edge processor 106. The edge processor 106 is connected to the analog-to-digital conversion unit 105 and the local memory, respectively, and is used to receive discrete envelope voltage sequences and perform environmental statistical fingerprint database establishment, one-way causal filtering, hardware transient truncation, dynamic standardization, foreign object presence confidence calculation, gating background update, and fusion alarm decision; the local memory is used to store node number, receiving frequency point number, candidate ground grid number or path segment number, as well as corresponding background statistics and alarm thresholds; the local memory is used to store the above data.

[0107] The sampling rate of the analog-to-digital conversion unit 105 can be set within the range of 200Hz to 5000Hz. For low-speed mobile platforms or close-range fixed monitoring, a lower sampling rate can be selected to reduce power consumption; for scenarios such as warehouse passages, high-level deployments, or high-speed movement, a higher sampling rate can be selected to retain short-term envelope changes. The above range is used to cover common low-frequency envelope disturbances while keeping the data throughput and power consumption of edge nodes at a low level.

[0108] The continuous envelope voltage signal output by the envelope detector 104 has a logarithmic linear relationship with the input RF power signal. This is used to compress the absolute amplitude corresponding to strong direct waves while retaining the relative envelope changes corresponding to weak multipath disturbances. The envelope detector 104 can employ devices with logarithmic detection, square-law detection, or envelope following capabilities to convert the received power changes of the common downlink RF signal into a continuous envelope voltage signal in the analog domain. The analog-to-digital converter 105 samples the continuous envelope voltage signal at low speed and outputs a discrete envelope voltage sequence. The edge processor 106 receives the discrete envelope voltage sequence and performs filtering, hardware transient truncation, dynamic normalization, foreign object detection, and fusion decision-making. Compared to broadband RF direct sampling, this link strips the megahertz-level carrier in the analog domain, allowing the analog-to-digital converter 105 to process only the envelope band signal, thereby reducing the sampling rate, data throughput, and power consumption.

[0109] The edge processor 106 first employs a unidirectional causal filter to process the discrete envelope voltage sequence. The filter can be implemented in a second-order form, with the passband set according to the frequency band of envelope disturbances caused by ground debris at a fixed node or mobile platform, for example, covering low-frequency variations in the range of 0.05Hz to 20Hz. The same filter coefficients are used in both the training and deployment phases.

[0110] The filter of this invention can be a unidirectional causal low-pass or band-pass filter, used to retain low-frequency envelope disturbances caused by foreign objects passing over or being statically obstructed by the mobile platform, while filtering out out-of-band high-frequency noise. The filter order, passband, and transient cutoff duration can be determined using foreign object-free calibration data and maintained consistent throughout the calibration and deployment phases. For different sampling rates, the number of transient cutoff points can be proportionally converted according to the transient duration.

[0111] The passband of a unidirectional causal filter can be configured according to the size of the object, node height, speed of the moving platform, and environmental noise characteristics. In fixed indoor scenarios, slower envelope changes at lower frequencies can be prioritized, while the upper limit of the passband can be appropriately increased for mobile robots or warehouse inspection scenarios. The filter can adopt Butterworth, Bessel, Chebyshev, or other second-order nodal IIR structures, or an FIR structure that meets the requirements of real-time edge processing.

[0112] Subsequently, the edge processor 106 performs hardware transient truncation processing on the filtered envelope voltage sequence to obtain a truncated voltage sequence. The hardware transient truncation processing can employ either a fixed-length truncation method or a threshold-stabilized truncation method. The fixed-length truncation method discards a fixed number of sampling points at the beginning of each sampling sequence segment. The threshold-stabilized truncation method continuously discards samples when the envelope amplitude exceeds a transient threshold until the sequence returns to a stable range. For the fixed-length truncation method, the first L sampling points of each filtered envelope voltage sequence segment are discarded, where L can be determined jointly based on the stabilization time of the envelope detector 104, the stabilization time of the analog-to-digital conversion reference voltage, the stabilization time of frequency switching, and the sampling rate. For the threshold-stabilized truncation method, when the amplitude of the filtered envelope voltage sequence exceeds the transient threshold, the edge processor 106 continuously discards the corresponding sampling points until the envelope amplitude returns to a preset stable range, after which samples are retained. This reduces the impact of power-on, sleep / wake-up, frequency switching, or analog-to-digital conversion reference voltage stabilization processes on foreign object detection.

[0113] Edge processor 106 performs dynamic normalization calibration on the truncated voltage sequence based on an environmental statistical fingerprint database to obtain an envelope timing transformation sequence. Specifically, it reads the background mean and background scale that match the current edge node, receiving frequency, and spatial location. The background scale can be the standard deviation or a robust scale converted from the absolute median. Then, it performs timing calibration on the truncated voltage sequence based on the background mean and background scale to obtain a timing calibration sequence. Finally, it performs amplitude pruning or piecewise linear compression on the timing calibration sequence to obtain the envelope timing transformation sequence. This invention uses a piecewise linear function to compress anomalous amplitudes.

[0114] The minimum constant in dynamic normalization can be set to 1e. -5 up to 1e -3 To prevent division by zero or excessive values ​​from occurring when the background scale is too small, the amplitude clipping threshold C can be set from 1.5 to 4. In indoor scenes with low ambient noise, a smaller clipping threshold can be used to improve sensitivity, while in scenes with more temporary mobile interference or strong broadcast signal fluctuations, a larger clipping threshold can be used to reduce false alarms.

[0115] The edge processor 106 generates a foreign object presence confidence score based on the envelope timing transformation sequence. Specifically, the envelope timing transformation sequence is segmented according to a preset window length W and a sliding step size H to obtain timing segmentation features. When there is one edge node and one receiving frequency point, each window forms a one-dimensional timing segmentation feature. When there are multiple edge nodes or multiple receiving frequencies points, the node dimension or frequency point dimension is used as the channel dimension to obtain multi-channel timing segmentation features. In mobile platform scenarios, mileage, speed, or path segment number can be used as auxiliary information input to the decision module.

[0116] The envelope temporal transformation sequence is segmented into short time windows matching the passage time of the target, and overlapping temporal segmentation features are generated with a sliding step size smaller than the window length. The window length can be set to the number of sampling points corresponding to 0.5 seconds to 5 seconds, and the sliding step size can be set to 5% to 50% of the window length; the specific values ​​can be adjusted according to the size of the target object, the node installation height, the speed of the moving platform, the sampling rate, and the throughput capability of the edge processor 106.

[0117] The preset window length W can be set according to the duration of the target disturbance, and the sliding step size H can be set according to a certain proportion of the window length. When the sampling rate, target moving speed, or node installation height changes, the window length and sliding step size can be adjusted accordingly to achieve a balance between alarm response speed and discrimination stability. Preferably, a longer window is used in fixed low-speed monitoring scenarios to improve stability, and a shorter window is used in mobile platform scenarios to improve response speed.

[0118] The edge processor 106 inputs the temporal segmentation features into the lightweight foreign object discrimination module to obtain the confidence level of foreign object presence. The lightweight foreign object discrimination module can use statistical thresholds, traditional machine learning, standard one-dimensional convolutional neural networks or combinations thereof to discriminate local ripples, slope changes, short-term energy changes and background deviation, and output the confidence level of foreign object presence.

[0119] In one optional embodiment, the lightweight foreign object detection module employs a standard one-dimensional convolutional neural network. For example... Figure 2 As shown, the input to this network is a temporal segmentation feature. The network consists of three one-dimensional convolutional layers with decreasing feature dimensions. Each one-dimensional convolutional layer is followed by a batch normalization layer, a non-linear activation layer (ReLU), and a max pooling layer to extract multi-scale temporal features. Subsequently, the feature sequence enters a global average pooling layer and a feature flattening layer to map the variable-length features into a fixed-length global feature vector. Finally, the confidence level of the presence of foreign objects on the indoor ground is output through multiple cascaded fully connected layers, a non-linear activation layer (ReLU), and a fully connected layer. Figure 2 This invention only illustrates one specific lightweight network topology embodiment and does not intend to use it for other purposes. Figure 2 The invention is not limited to specific convolutional kernel size, number of channels, or number of fully connected layer nodes. It does not require specific convolutional kernel combinations, specific channel increment methods, specific pooling operators, or specific parameter quantities as essential technical features; when the edge processor 106 has limited capabilities, statistical thresholding, support vector machines, decision trees, or other lightweight classifiers can be used as alternatives.

[0120] When training the lightweight foreign object detection module, the calibration or training samples are collected from both foreign object-free and foreign object-containing states. Foreign object-containing samples can cover categories such as transparent containers, metal sheets, cables, flexible packaging, and low-lying obstacles, and can accommodate different placement locations, ground materials, node heights, and frequency conditions. All samples undergo the same causal filtering, hardware transient truncation, dynamic standardization, and sliding window segmentation process as the deployment end.

[0121] When the training data does not yet cover all foreign object materials and spatial locations, samples of non-target background disturbances, temporary movement interference, minor furniture movement interference, and intensity fluctuations at different frequencies can be collected as negative samples or interference samples to verify the stability of the envelope detection, causal filtering, hardware transient truncation, dynamic standardization, and foreign object discrimination modules in low signal-to-noise ratio envelope recognition; the alarm threshold and module parameters are based on the calibration results of the scattered foreign object samples on the ground.

[0122] To verify the physical effectiveness of the lightweight foreign object detection module of this invention, offline equivalent verification was performed using near-field dynamic multipath perturbation data. During testing, the envelope temporal features of edge nodes under near-range perturbation conditions were extracted and input into the lightweight foreign object detection module for inference. The verification results are shown in Table 1. This verification demonstrates that the link has the ability to identify low-frequency multipath perturbations. The specific detection threshold and performance indicators for ground-scattered foreign object samples can be replaced or supplemented in subsequent scenario-based experiments.

[0123] Table 1. Verification results for various types of scattered foreign matter materials.

[0124]

[0125] The data in Table 1 show that the envelope detection proposed in this invention, combined with unidirectional causal filtering and dynamic calibration algorithms, can generate stable and distinguishable envelope responses to near-field low-frequency multipath disturbances. This verification demonstrates that the hardware-software collaborative link possesses the ability to identify low signal-to-noise ratio envelope disturbances, providing link-layer support for ground-borne foreign object monitoring. The specific detection thresholds and performance indicators for ground-borne foreign object samples can be further calibrated based on the foreign object material, size, placement location, and node installation conditions.

[0126] During deployment, the edge processor 106 outputs the foreign object presence confidence score for each window. In fixed multi-node scenarios, the foreign object presence confidence scores corresponding to consecutive time windows can be weighted and fused to obtain a fused confidence score. The aforementioned fusion weights can be determined by the geometric distance between the node and the candidate grid, antenna orientation, historical stability, current background noise scale, and frequency quality index. In mobile platform scenarios, the confidence scores on continuous path segments can be projected onto the ground grid, and adjacent grids can be merged or non-maximum suppression can be applied.

[0127] The number of consecutive windows M can be set from 3 to 15, the first threshold can be set from 0.5 to 0.7, and the second threshold can be set from 0.75 to 0.95. For robot obstacle avoidance scenarios with high safety requirements, the alarm threshold, i.e., the second threshold, can be appropriately reduced; for fixed monitoring scenarios that are sensitive to false alarms, the second threshold can be increased and the number of consecutive window confirmations can be increased.

[0128] The fusion decision process of the edge processor 106 corresponds one-to-one with the steps of the aforementioned method, and can be quantified using the following formula:

[0129] ;

[0130] The candidate ground grid with the highest fusion confidence score is selected as the target candidate region:

[0131]

[0132] The three-level state decision is completed based on the fused confidence level and the preset threshold, and the time continuity condition is combined to determine whether an alarm is triggered. The decision logic corresponds completely to the dual threshold rule in the previous steps and satisfies the following relationship:

[0133]

[0134] in, Represents candidate ground grids fusion confidence; This represents a non-linear activation function used to map the weighted summation of log odds results to the interval between 0 and 1; , , These are indices for edge nodes, receiving frequencies, and time windows, respectively. Indicates the first The node, the first The frequency point at the first A time window for candidate grids Confidence weights; This indicates the confidence level of the presence of a foreign object in a single node, single frequency point, and single window. This represents the set of all candidate grids within the ground area to be monitored. This represents the candidate ground grid with the highest fusion confidence.

[0135] Indicates the candidate ground grid The alarm judgment value, This indicates an alarm for scattered foreign objects; otherwise, it is 0. The first threshold, It is the second threshold, and satisfies This corresponds exactly to the warning threshold in the aforementioned steps; This is an indicator function that takes the value 1 when the condition within the parentheses is true, and 0 otherwise; t This represents the total number of consecutive windows involved in the time continuity assessment. In the The fusion confidence of each time window; Indicates continuity More than [number] windows are needed Minimum number of windows, This is a consistency condition indicator function. It takes the value 1 when the same candidate ground area satisfies at least one of edge node consistency, receiving frequency consistency, and path segment consistency, and takes the value 0 otherwise, ensuring that the system triggers an alarm only when both temporal continuity and spatial frequency consistency are satisfied.

[0136] Specifically, the three-level decision logic perfectly matches the aforementioned step rules: when the fusion confidence of the target mesh is less than the first threshold. When no foreign objects are found on the ground, the background fingerprint of the corresponding area is allowed to be updated according to the rules; when the fusion confidence is between the first threshold and the second threshold, the system enters the observation state and continues to accumulate the confidence of subsequent time windows; when the fusion confidence is greater than the second threshold... If the alarm conditions are met, the system will output an alarm for foreign objects scattered on the ground and the corresponding candidate area. Otherwise, it will enter a suspected observation state and suspend the background fingerprint update of the corresponding area.

[0137] In this embodiment of the indoor robotic vacuum cleaner, one or more edge nodes are installed at the front or side of the robot, with the receiving antenna 101 facing the ground in front. The robot establishes a path background fingerprint upon initial mapping or when the user confirms the absence of foreign objects on the ground. During normal cleaning, the system combines wheel speed and mileage to map abnormal windows onto candidate ground grids in front. When a candidate ground grid consistently shows a high-confidence response, the robot can slow down, detour, stop, or call upon other sensors for verification.

[0138] In the embodiment of indoor privacy-sensitive areas, multiple edge nodes are installed on both sides of elderly care rooms, wards, dormitories, offices, or corridors. The system does not collect images or actively transmit radio frequency signals. When abnormal objects such as medicine boxes, cables, or plastic packaging appear on the ground, the envelope statistics of multiple edge nodes change relative to the background fingerprint. The system then performs a fusion decision to output candidate regions and can send alarm information to the local gateway or management platform.

[0139] In this embodiment of fixed monitoring of low-level indoor storage racks, edge nodes are installed at the bottom of the racks, low on the wall, or at the edge of the aisle, with the receiving antenna 101 facing the front of the racks or the aisle floor area. The system can select one or more stable frequencies from FM broadcast, terrestrial digital television, or cellular downlink signals, and divide the rack aisle floor into several candidate grids. During initial deployment, a background fingerprint is established after confirming that there are no scattered foreign objects on the aisle floor. During normal operation, if small metal parts, packing tape, cables, packaging film, or cardboard fragments fall into the candidate grid, the system outputs an alarm based on envelope perturbation, background deviation, and continuous window fusion. In this embodiment, the sampling rate, filter passband, window length, and alarm threshold can be configured within the aforementioned parameter range according to the aisle width, rack height, and number of nodes.

[0140] The sampling rate, filter passband, window length, sliding step size, discrimination module type, feature dimension, number of edge nodes, etc. in the above embodiments can all be adjusted according to the scenario and hardware capabilities. While keeping the core processes such as passive common radio frequency envelope sampling, foreign object background fingerprint calibration, causal preprocessing for ground candidate regions, and continuous decision-making at the edge unchanged, the above parameter adjustments will not affect the implementation of the present invention.

[0141] The solution disclosed in this invention, without actively transmitting radio frequency signals, acquiring images, or requiring CSI demodulation or high-frequency IQ sampling, utilizes only the low-frequency envelope perturbation of a common radio frequency signal to construct a hardware dimensionality reduction for the sensing link through envelope detection. It relies on environmental statistical fingerprints to complete dynamic standardization calibration and time series conversion, effectively overcoming the problems of background drift and feature mismatch that easily occur during real-time edge deployment. This truly achieves miniaturized, lightweight, and low-power distributed indoor deployment. Specific advantages are as follows:

[0142] First, the envelope reduction sensing method lowers the deployment threshold at the signal level. By directly stripping the high-frequency carrier through analog domain envelope detection, data acquisition can be completed at a sampling rate far lower than the radio frequency carrier frequency. This eliminates the need for channel state information, high-speed IQ sampling, and complex coherent processing, significantly reducing data throughput and computational overhead. This methodological approach fundamentally supports the miniaturization and low-power edge node implementation. Furthermore, relying on the passive sensing characteristics, the solution has no active radio frequency transmission, is not limited by illumination or transparent objects, and does not infringe on image privacy, making it suitable for long-term, non-intrusive monitoring in various indoor scenarios.

[0143] Second, a dynamic calibration mechanism based on environmental fingerprints systematically suppresses background drift and feature mismatch. Through hardware transient truncation and causal filtering preprocessing, combined with the background mean and scale of the environmental statistical fingerprint database, dynamic standardization calibration is completed. This effectively offsets baseline shifts caused by slow changes in environmental temperature and humidity, common signal strength, and hardware temperature drift, eliminating interference from slow environmental changes on feature extraction. Combined with amplitude clipping, it improves feature consistency across different nodes, frequencies, and scenarios, ensuring the long-term stability of the discrimination model.

[0144] Third, lightweight discrimination and gating fusion decision-making balances real-time performance and detection reliability at the edge. Features are extracted using a sliding time window and confidence scores are output via a lightweight module. Combined with a dual-threshold, three-level decision-making process and a multi-dimensional weighted fusion mechanism, this reduces false alarms through multi-dimensional consistency verification while adapting to limited computing power at the edge. Simultaneously, a background fingerprint gating update rule is designed: in low-confidence scenarios, the background is adaptively updated to adapt to environmental changes, while in high-confidence scenarios, updates are frozen to prevent foreign object features from being absorbed by the background, thus balancing environmental adaptability and anomaly detection sensitivity.

[0145] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting.

[0146] Embodiments of the present invention may be provided as methods, systems, or computer program products. The invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the invention may take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0147] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0148] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0149] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0150] The present invention also discloses a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting.

[0151] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0152] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for detecting scattered foreign objects on the ground based on passive radio frequency envelope fingerprinting, characterized in that, The method includes: Acquire non-cooperative radio frequency signals containing spatial multipath disturbances at the locations of each edge node in the ground area to be monitored; Envelope detection is performed on the non-cooperative radio frequency signal to construct a discrete envelope voltage sequence; Based on the discrete envelope voltage sequence and the environmental statistical fingerprint database, a time series transformation is performed to obtain the envelope time series transformation sequence; The confidence level for the presence of foreign matter is generated based on the envelope time-series transformation sequence. The ground monitoring results are determined based on the confidence level of the presence of the foreign object and the alarm conditions.

2. The ground-scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting according to claim 1, characterized in that, The step of constructing a discrete envelope voltage sequence by envelope detection of the non-cooperative radio frequency signal specifically includes: Frequency selection and filtering out interference signals in the non-cooperative radio frequency signals yields interference-free radio frequency signals near the target receiving frequency. The gain of the interference-free radio frequency signal is adjusted to the effective input range to obtain the radio frequency power signal; The radio frequency power signal is converted into a continuous envelope voltage signal; The continuous envelope voltage signal is acquired using a sampling rate lower than the radio frequency carrier frequency to obtain a discrete envelope voltage sequence.

3. The ground-scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting according to claim 1, characterized in that, The step of performing time-series transformation based on the discrete envelope voltage sequence and the environmental statistical fingerprint database to obtain an envelope time-series transformation sequence specifically includes: The discrete envelope voltage sequence is processed by a filter to obtain a filtered envelope voltage sequence; The filtered envelope voltage sequence is subjected to hardware transient truncation to obtain the truncated voltage sequence; The cutoff voltage sequence is dynamically standardized and calibrated based on the environmental statistical fingerprint database to obtain the envelope timing transformation sequence.

4. The ground-scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting according to claim 3, characterized in that, The dynamic normalization calibration of the truncated voltage sequence based on the environmental statistical fingerprint database to obtain the envelope time series transformation sequence specifically includes: Read the background mean and background scale that match the current node, current receiving frequency, candidate ground grid or path segment from the environmental statistical fingerprint database; Based on the background mean and the background scale, the cutoff voltage sequence is time-calibrated to obtain an envelope time-calibration sequence. The envelope timing calibration sequence is subjected to amplitude clipping or piecewise linear compression to obtain the envelope timing transformation sequence.

5. The ground-scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting according to claim 1, characterized in that, The step of generating a foreign object presence confidence score based on the envelope time-series transformation sequence specifically includes: The envelope time-series transformation sequence is segmented according to a preset window length and sliding step size to obtain time-series segmentation features; The temporal segmentation features are input into the lightweight foreign object detection module to obtain the confidence level of foreign object presence.

6. The ground-scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting according to claim 1, characterized in that, The determination of ground monitoring results based on the confidence level of the presence of the foreign object and the alarm conditions specifically includes: The confidence of foreign object presence is weighted and fused for continuous time windows and for different edge nodes and / or different receiving frequencies to obtain a fused confidence. Determine whether the fusion confidence score is less than a first threshold. If the fusion confidence score is less than the first threshold, output that there are no scattered foreign objects on the ground, and allow the background fingerprints of the corresponding candidate ground grids, path segments, edge nodes, or receiving frequencies to be updated according to the gating update rules. If the fusion confidence score is greater than or equal to the first threshold and less than or equal to the second threshold, enter the observation state and continue to accumulate the confidence score of subsequent time windows. If the fusion confidence score is greater than the second threshold, determine whether the alarm condition is met. If the alarm condition is met, output that there are scattered foreign objects on the ground and the candidate ground area, and issue an alarm. If the alarm condition is not met, enter the suspected observation state and pause the updating of the background fingerprints of the corresponding candidate ground grids, path segments, edge nodes, or receiving frequencies according to the gating update rules. The second threshold is greater than the first threshold.

7. A ground-scattered foreign object monitoring system based on passive radio frequency envelope fingerprinting, characterized in that, The system includes: at least one edge node and an edge processor; each edge node includes: a receiving antenna and an envelope detection construction module; the receiving antenna is connected to the envelope detection construction module and the edge processor. The receiving antenna is used to receive non-cooperative radio frequency signals containing spatial multipath disturbances in the monitoring state; The envelope detection construction module is used to construct the envelope of non-cooperative radio frequency signals to obtain discrete envelope voltage sequences. The edge processor is used to perform time-series transformation based on the discrete envelope voltage sequence and the environmental statistical fingerprint database to obtain an envelope time-series transformation sequence; generate a foreign object presence confidence level based on the envelope time-series transformation sequence; and determine the ground monitoring results based on the foreign object presence confidence level and alarm conditions.

8. The ground debris monitoring system based on passive radio frequency envelope fingerprinting according to claim 7, characterized in that, The envelope detection construction module includes: a frequency selective filter, an RF amplifier, an envelope detector, and an analog-to-digital converter. The receiving antenna is connected in sequence through the frequency selective filter, the RF amplifier, the envelope detector, the analog-to-digital converter, and the edge processor. The frequency-selective filter is used to select the frequency and filter out interference signals in the non-cooperative radio frequency signal to obtain an interference-free radio frequency signal near the target receiving frequency point; the radio frequency amplifier is used to adjust the gain of the interference-free radio frequency signal to the effective input range to obtain a radio frequency power signal; the envelope detector is used to convert the radio frequency power signal into a continuous envelope voltage signal; and the analog-to-digital conversion unit is used to acquire the continuous envelope voltage signal at a sampling rate lower than the radio frequency carrier frequency to obtain a discrete envelope voltage sequence.

9. A computer-readable storage medium, characterized in that, The device stores a computer program, which, when executed by a processor, implements the steps of the ground debris detection method based on passive radio frequency envelope fingerprinting as described in any one of claims 1-6.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the ground scattered foreign object monitoring method based on passive radio frequency envelope fingerprinting as described in any one of claims 1-6.

Images