Container supervision article verification method based on radio frequency technology

Abstract

The invention provides a container supervision article verification method based on the radio frequency identification technology, and relates to the technical field of container supervision article verification based on the radio frequency identification technology, and the method comprises the steps: transmitting a wireless signal to a tag on the surface of an article through RFID equipment, and enhancing a received reflection signal through a low-noise amplifier, filtering environmental noise by using a band-pass filter, converting a filtered reflected signal into a time domain signal by using an analog-to-digital converter, and extracting frequency domain characteristics by using FFT (Fast Fourier Transform) so as to calculate a waveform characteristic index; establishing a historical fingerprint template library, storing a waveform characteristic index corresponding to each state, setting a dynamic similarity threshold value based on historical data distribution, screening out templates of which the Euclidean distance is smaller than the threshold value and which belong to the same state as candidates, and if the candidate templates are not empty, selecting the state corresponding to the template of which the Euclidean distance is minimum as a judgment result; and if the candidate template is empty, the state is judged to be abnormal, and container opening alarm is triggered.

Description

Technical Field

[0001] The present invention relates to the technical field of container supervision article verification based on wireless radio frequency technology, and in particular to a container supervision article verification method based on wireless radio frequency technology. Background Art

[0002] Monitoring the health of item tags has become increasingly important in modern logistics and supply chain management. These tags are often used for asset tracking, inventory management, and product identification. With the increasing complexity of tag applications and the diversification of usage environments, tags may be affected by physical damage, signal interference, or other unexpected factors, resulting in abnormal status. However, traditional monitoring methods often rely on manual inspections, which are inefficient and prone to errors, and cannot provide accurate real-time feedback on tag status.

[0003] RFID (Radio Frequency Identification) technology is widely used in container monitoring and item verification. By attaching tags to cargo and communicating via radio waves, RFID enables real-time tracking and management of goods. This technology is particularly important in ports, warehouses, and during transportation, as it automatically identifies and records cargo information, improving logistics efficiency and reducing human error. However, with the development of the Internet of Things and smart logistics, RFID systems face the challenge of accurately monitoring and verifying item status in complex environments. In these scenarios, tags may experience signal anomalies or damage due to physical collisions, environmental interference, and other factors, placing higher demands on logistics accuracy and security.

[0004] Existing technologies typically use a basic RFID reader combined with a database to monitor goods. In this system, the RFID reader scans each passing tag and uploads the data to a central database for recording and analysis. This method can monitor the status of goods to a certain extent, but it relies on fixed signal strength and simple anomaly detection algorithms. When the tag is interfered with or damaged, the existing system is often unable to make timely and accurate judgments. In addition, existing technologies lack in-depth analysis of the tag's health status. For example, it cannot effectively distinguish between noise interference and physical damage.

[0005] The shortcoming of existing technology is that it is not accurate enough in judging the tag status in complex environments. Especially in the face of signal interference and tag damage, the probability of misjudgment is high, leading to incorrect cargo management decisions.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide a method for verifying items in container supervision based on wireless radio frequency technology to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The method for selecting a communication path for transmitting data by a device node includes the following specific steps:

[0010] Step 1: Use RFID equipment to transmit wireless signals to the tags on the surface of the items in the container to be monitored. The reflected signal is received by the tag. After receiving the reflected signal, the signal strength is enhanced by a low-noise amplifier, and a bandpass filter is used to filter out environmental noise to retain the tag's effective frequency band signal;

[0011] Step 2: Use an analog-to-digital converter to convert the filtered reflection signal into a time domain signal. Then use FFT to extract the time domain signal to obtain frequency domain features. The frequency domain features include maximum amplitude, peak frequency, and harmonic distribution data. The waveform characteristic index is calculated by combining these three characteristic values.

[0012] Step 3: Collect the reflection signal of the object surface tag in different historical states, and calculate the historical waveform characteristic index according to the method of steps 1 and 2. At the same time, record the state of the corresponding object surface tag, build a historical fingerprint template library, store the historical waveform characteristic index corresponding to each state, and calculate the similarity between the waveform characteristic index and each historical waveform characteristic index in the template library using the Euclidean distance algorithm;

[0013] Step 4: Set a dynamic similarity threshold based on the historical data distribution and exclude templates of different states. Select templates with a Euclidean distance less than the threshold and belonging to the same state as candidate templates. If the candidate templates are not empty, select the state corresponding to the template with the smallest Euclidean distance as the judgment result. If the candidate template is empty, the state is judged to be abnormal and the container opening alarm is triggered.

[0014] Furthermore, the feature is that: the logic of using an RFID device to transmit a wireless signal to a tag on the surface of an item in a container to be monitored and receiving a reflected signal from the tag is as follows:

[0015] The RFID device operates in the ultra-high frequency band, and its transmission frequency range is set to [920MHz, 925MHz] in accordance with China's radio frequency allocation standards. The RFID device adopts an intermittent transmission mode, which is set to transmit a signal every 0.5 seconds, with each transmission lasting 50 milliseconds, for a total of 10 transmissions, each lasting 5 seconds.

[0016] The transmitting antenna and receiving antenna of the RFID device are separated. The transmitting signal covers the space inside the container, and the receiving antenna captures the tag reflected signal S rx (t); Receive reflected signal Srx After (t), the signal strength is enhanced by a low-noise amplifier and the environmental noise is filtered out by a bandpass filter. The logic of retaining the tag's effective frequency band signal is as follows:

[0017] Use a low noise amplifier to increase the signal strength, and the amplified signal S amp The expression of (t) is:

[0018] S amp (t) = G·S rx (t)

[0019] Where g is the amplifier gain coefficient, which is set to G = 10dB;

[0020] Use an analog bandpass filter to filter out non-operating frequency noise, based on the formula:

[0021] S fil (t) = h BP (t)*S amp (t)

[0022] Among them, h BP (t) is the impulse response function of the simulated bandpass filter, whose passband is 920 MHz and 925 MHz. The symbol * represents the time domain convolution operation.

[0023] Furthermore, the logic of using an analog-to-digital converter to convert the filtered reflection signal into a digital time domain signal is:

[0024] According to Nyquist sampling theorem, the sampling frequency f s Must be greater than twice the highest frequency of the signal, that is, f s ≥2·f high , f high =925MHz is the maximum frequency of the signal, select f s =2GHz, sampling period

[0025] The duration of a single transmission is T = 50ms, and the total number of sampling points is

[0026] The filtered signal S fil (t) is uniformly sampled, based on the formula:

[0027] S dis [n]=S fil (t0+n·T s ),n=0,1,…,N-1

[0028] Among them, n is the time index, S dis [n] represents the nth sampling point of the discrete signal, t0 is the initial time, and N is the number of sampling data points;

[0029] S dis The digital signal expression is:

[0030] S dis ={S dis [0],S dis [1],…,S dis [N-1]}

[0031] Where N = 100 × 10 6 , is the number of sampling data points;

[0032] The logic of extracting time domain signals through FFT to obtain frequency domain features is:

[0033] S dis Divide into L frames, set the length of each frame to N fft =2 20 = 1048576 points, calculate the FFT after adding window to each frame signal:

[0034]

[0035] Among them, k is the frequency index, which represents the kth discrete frequency component in the frequency domain. After framing, n′=0,1,…,N in each frame fft -1, S fft [k, l] is a complex number containing the amplitude and phase information of the frequency component, w[n′] is the window function, and l = 0, 1, …, L-1 is the frame index;

[0036] The frequency resolution is calculated as:

[0037]

[0038] Among them, f k is the actual frequency corresponding to the frequency index k, Δf is the frequency resolution of the FFT output; the effective frequency band is limited to 920MHz≤f k k value for ≤925MHz:

[0039] The logic of using the frequency domain features to calculate the maximum amplitude, peak frequency and harmonic distribution data is as follows: for the maximum amplitude of the first frame, in the effective frequency band [k min ,k max ], based on the formula:

[0040]

[0041] Among them, A max [l] represents the maximum amplitude of the l-th frame signal in the effective frequency band; traverse all frequency indexes k of each frame, in the effective frequency band [k min,k max ]Search for the maximum value within;

[0042] For the peak frequency, which is the frequency corresponding to the maximum amplitude, the formula is:

[0043] f peak [l] = k peak [l]·Δf

[0044] Among them, f peak [l] represents the main frequency of the first frame, is the index corresponding to the maximum amplitude of the lth frame;

[0045] For harmonic distribution, in the effective frequency band [k min ,k max ], extract the frequency components with the first K amplitudes as candidate fundamental frequencies, and for each candidate fundamental frequency f cand,p [l]∈{f cand,1 [l],f cand,2 [l],…,f cand,K [l]}, where f cand,p [l] = k p [l]·Δf, k p [l] is the frequency index of the K frequencies before the amplitude sorting of the lth frame, k p ∈[1,K], for each candidate fundamental frequency f cand,p [l], calculate its harmonic score:

[0046]

[0047] Among them, H m [l](f cand,p [l]) is the mth harmonic amplitude of the candidate fundamental frequency of the lth frame, H1[l](f cand,p [l]) is the first harmonic amplitude of the candidate fundamental frequency of the lth frame, m=2,3,…,M is the harmonic order, Scire(f cand,p [1]) is the score of the candidate fundamental frequency of the first frame;

[0048] For the fundamental amplitude: H1[l]=|S fft [k base ,l]|, where k base is the index corresponding to the fundamental amplitude;

[0049] For harmonic amplitudes:

[0050]

[0051] The harmonic order limits are: where f high =925MHz;

[0052] Select the candidate fundamental frequency with the highest score as the final fundamental frequency of the lth frame:

[0053]

[0054] Among them, f base [l] is the fundamental frequency finally determined in the lth frame;

[0055] Combine the above three waveform feature values, take the average of L frame features, and calculate the waveform feature index F ind The formula is:

[0056]

[0057] in, is the average of the maximum amplitude of multiple frames in the signal spectrum, is the average of the maximum amplitude of multiple frames, The corresponding frequency mean, is the mean value of the mth order harmonic amplitude in multiple frames, m -p is the weighting of the harmonic order, A0=1 is the reference amplitude, f0=1 is the reference frequency, p is the harmonic attenuation weight, is the highest harmonic order in multiple frames, and

[0058] Furthermore, the logic of collecting the reflection signal of the tag on the surface of the object in different historical states and calculating the historical waveform characteristic index according to the method of steps 1 and 2 is as follows:

[0059] Collect historical tag reflection signal data in the following states: normal state (the tag is intact and free of interference), damaged state (the tag surface is physically damaged), lost state (the tag is completely detached from the object surface), and interference state (metal obstruction or multipath reflection interference).

[0060] Calculate the historical waveform characteristic index F for each group of signals according to steps 1 and 2 ind,h , build fingerprint template library The storage structure is in is the historical label status of the oth template in the template library, which includes normal, damaged, lost and interference, and O is the number of templates in the template library;

[0061] Calculate the Euclidean distance D from the real-time feature to the oth template in the template library o :

[0062]

[0063] Among them, o is the template index in the template library, F ind is the real-time feature index, is the degree history feature index of the oth template in the template library.

[0064] Furthermore, based on the historical data distribution, a dynamic similarity threshold is set and templates of different states are excluded. The logic of selecting templates with a Euclidean distance less than the threshold and belonging to the same state as candidate templates is as follows:

[0065] For each state z in the template library, the Euclidean distance distribution of similar state templates is statistically analyzed based on historical data, and a dynamic threshold is set.

[0066]

[0067] Among them, μ z is the mean Euclidean distance from all templates in state z to the feature center of the state, σ z is the standard deviation of the distance from all templates to the feature center in state z;

[0068] The state feature center is the feature index F of all templates for state z. ind,h , calculate its mean:

[0069]

[0070] Among them, O z is the number of templates in state z, o z =1,2,…,O z ;

[0071] Calculate the Euclidean distance from each template to the feature center to calculate μ z and σ z :

[0072] The condition for screening candidate templates is: the Euclidean distance is less than the threshold of the state to which it belongs: Template status is the preset valid status: label status z o Including normal, damaged, lost, and interfered;

[0073] The candidate templates for screening are: And z o It is in a valid state;

[0074] If the candidate template is not empty, select the template corresponding to the state with the smallest Euclidean distance as the judgment result:

[0075]

[0076] Among them, * is the template index with the smallest Euclidean distance among the candidate templates, is the label state that the real-time signal is most likely to correspond to;

[0077] If multiple templates have the same minimum distance but different states, they are marked as "state conflict" and require manual review; if the candidate template is empty, that is, all templates do not meet the threshold conditions, it is judged as "unknown anomaly";

[0078] The risk set is set to E, which includes the following risks: tag damage, loss, interference, and unknown anomalies; the alarm condition judgment conditions are:

[0079]

[0080] At that time, when Risk = 1, a container unpacking and inspection request is triggered and abnormal information is recorded. When Risk = 0, the tag is in a normal state, only the status log is updated, and no operation is triggered.

[0081] Compared with the prior art, the present invention has the following beneficial effects:

[0082] This method uses RFID equipment to transmit wireless signals to the tags of items in the container, receives the reflected signals, and then processes the received reflected signals using a signal amplifier and filter. This process ensures signal stability and clarity, minimizes the impact of environmental noise, and lays the foundation for subsequent signal analysis. The signal is converted into time domain signal data through an analog-to-digital converter. Then, an FFT (Fast Fourier Transform) is used to extract the frequency domain characteristics of the time domain signal, including maximum amplitude, peak frequency, and harmonic distribution data. These characteristic values ​​are combined to form a waveform characteristic index, which provides the basic data for accurately identifying the tag status.

[0083] The present invention uses frame-based FFT and candidate template screening mechanisms to cover unknown anomaly detection while ensuring real-time performance. It segments long-time signals and independently calculates frequency domain features for each frame, taking into account both frequency resolution and computational efficiency. It also establishes a historical fingerprint template library and combines it with the setting of dynamic similarity thresholds to enhance the system's adaptability to various states.

[0084] This invention uses Euclidean distance calculation to determine the state of the container. The system can flexibly select the template that best matches the current state, providing a scientific basis for state determination. Furthermore, when a candidate template is empty, the system can promptly trigger an abnormality alarm to ensure container safety. The integration of this series of technical features greatly enhances the performance and risk prevention capabilities of smart containers in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 Schematic diagram of the overall method of the present invention. DETAILED DESCRIPTION

[0086] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0087] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0088] Example:

[0089] See also Figure 1 , the present invention provides a technical solution:

[0090] A method for verifying items under control in a container based on wireless radio frequency technology, comprising the following steps:

[0091] Step 1: Use RFID equipment to transmit wireless signals to the tags on the surface of the items in the container to be monitored. The reflected signal is received by the tag. After receiving the reflected signal, the signal strength is enhanced by a low-noise amplifier, and a bandpass filter is used to filter out environmental noise to retain the tag's effective frequency band signal;

[0092] The logic of using RFID (radio frequency identification) equipment to transmit wireless signals to the tags on the surface of the items in the container to be monitored and receiving the reflected signals from the tags is as follows:

[0093] The RFID device operates in the ultra-high frequency band, and its transmission frequency range is set to [920MHz, 925MHz] in accordance with China's radio frequency allocation standards. The RFID device adopts an intermittent transmission mode, which is set to transmit a signal every 0.5 seconds, with each transmission lasting 50 milliseconds, for a total of 10 transmissions, each lasting 5 seconds.

[0094] RFID devices operate in the ultra-high frequency band, and the transmission frequency range complies with China's radio frequency allocation standards. This range [920MHz, 925MHz] is a legal frequency band for industrial, scientific research, and medical scenarios. The choice of transmission frequency affects the penetration ability and transmission distance of wireless signals. For example, lower frequencies such as 920MHz can provide better penetration than higher frequencies, but the transmission data rate may be slightly lower.

[0095] Intermittent transmission mode can save energy and detect tag responses at certain time intervals. Continuous transmission of 50 milliseconds ensures signal stability, while the 0.5-second transmission interval can reduce device power consumption.

[0096] The transmitting antenna and receiving antenna of the RFID device are separated. The transmitting signal covers the space inside the container, and the receiving antenna captures the tag reflected signal S rx (t); Receive reflected signal S rx After (t), the signal strength is enhanced by a low-noise amplifier and the environmental noise is filtered out by a bandpass filter. The logic of retaining the tag's effective frequency band signal is as follows:

[0097] Use a low noise amplifier to increase the signal strength, and the amplified signal S amp The expression of (t) is:

[0098] S amp (t) = G·S rx (t)

[0099] Among them, G is the amplifier gain coefficient, which is set to G = 10dB; the function of the amplifier is to convert the weak signal amplitude A rx Upgrade to a range that is easy to process while preserving the frequency and phase information of the signal; amp (t) is the amplified signal intensity, which is a function of time t;

[0100] S amp (t) is the instantaneous strength of the amplified signal. The output capacity of the amplified signal directly affects the signal-to-noise ratio of the subsequent processing. The larger its value, the higher the signal strength after amplification. However, it is necessary to avoid distortion caused by amplifier saturation. S amp (t) and S rx (t) is in a linear proportional relationship;

[0101] G is the gain of the linear amplifier, which indicates the amplification factor of the signal amplitude. The gain is usually expressed in decibels. The formula is:

[0102] G dB =10·log 10 (G linear )

[0103] If G linear =10, then G dB =10·log 10 (10) = 10·1 = 10; if G dB =10, then G linear =10 10 / 10 =10;G linear The larger the value, the higher the signal amplitude amplification factor, but it may introduce more noise;

[0104] Glinear Indicates the linear gain of the amplifier, that is, the actual amplification factor of voltage or power, G dB Indicates the decibel gain of the amplifier, which is a logarithmic gain value. The higher the gain, the stronger the signal, but the signal-to-noise ratio may decrease. It is necessary to balance the gain and noise performance.

[0105] Use an analog bandpass filter to filter out non-operating frequency noise, based on the formula:

[0106] S fil (t) = h BP (t)*S amp (t)

[0107] Among them, h BP (t) is the impulse response function of the simulated bandpass filter, whose passband is 920MHz and 925MHz. The symbol * represents the time domain convolution operation, which realizes the band-selective filtering. After filtering, only the effective frequency band signal of the tag is retained, and the out-of-band noise is suppressed.

[0108] S fil (t) The signal strength after bandpass filtering, which only retains the components within the frequency bands of 920 MHz and 925 MHz, is the effective signal purity, that is, the quality of the tag reflection signal after removing the environmental noise. The larger the value, the stronger the signal energy in the target frequency band and the higher the signal-to-noise ratio;

[0109] S fil (t) is S amp (t) and h BP The convolution operation combines the input signal with the filter characteristics through integration to achieve frequency domain selectivity.

[0110] As G increases, the output signal amplitude increases linearly, which may cause amplifier saturation and signal distortion; as G decreases, the output signal amplitude decreases, the signal-to-noise ratio decreases, and weak signals may be submerged by noise;

[0111] When the passband range is widened, such as 900-930MHz, more noise enters the subsequent processing, the signal-to-noise ratio is reduced, and the false positive rate increases. When the passband range is narrowed, such as 922-923MHz, signal energy loss may cause valid signals to be filtered out, and the missed detection rate increases.

[0112] Directly enhance the signal amplitude and improve the signal-to-noise ratio, but the gain and noise introduction need to be balanced. G = 10dB linear gain of 10 times is a typical compromise. Frequency domain selectivity is achieved through convolution operation, retaining the target frequency band signal and suppressing out-of-band noise. The passband strictly matches the device operating frequency band, that is, 920-925MHz, to ensure signal purity.

[0113] Step 2: Use an analog-to-digital converter to convert the filtered reflection signal into a time domain signal. Then use FFT to extract the time domain signal to obtain frequency domain features. The frequency domain features include maximum amplitude, peak frequency, and harmonic distribution data. The waveform characteristic index is calculated by combining these three characteristic values.

[0114] The logic of using an analog-to-digital converter to convert the filtered reflection signal into a digital time domain signal is:

[0115] According to Nyquist sampling theorem, the sampling frequency f s Must be greater than twice the highest frequency of the signal, that is, f s ≥2·f high , f high =925MHz is the maximum frequency of the signal, select f s =2GHz, sampling period

[0116] The duration of a single transmission is T = 50ms, and the total number of sampling points is

[0117] The filtered signal S fil (t) is uniformly sampled, based on the formula:

[0118] S dis [n]=S fil (t0+n·T s ),n=0,1,…,N-1

[0119] Among them, n is the time index, S dis [n] represents the nth sampling point of the discrete signal, t0 is the initial time, and N is the number of sampling data points;

[0120] S dis [n] is the discretized time domain signal, which means the signal at time t0+n·T s The amplitude of the continuous analog signal is converted into a digital signal to facilitate subsequent digital processing, S dis The larger the [n] value is, the higher the instantaneous signal strength is at the corresponding moment;

[0121] As n increases, the sampling point number increases, time moves backward, T s Reduce, that is, f s Increase, the time resolution is improved, but the amount of data increases;

[0122] S dis The digital signal expression is:

[0123] S dis ={S dis [0],S dis [1],…,Sdis [N-1]}

[0124] Where N = 100 × 10 6 , is the number of sampling data points;

[0125] The logic of extracting time domain signals through FFT to obtain frequency domain features is:

[0126] S dis Divide into L frames, set the length of each frame to N fft =2 20 = 1048576 points, calculate the FFT after adding window to each frame signal:

[0127]

[0128] Among them, k is the frequency index, which represents the kth discrete frequency component in the frequency domain. After framing, n′=0,1,…,N in each frame fft -1, S fft [k, l] is a complex number containing the amplitude and phase information of the frequency component, w[n′] is a window function, such as a Hamming window, used to reduce spectrum leakage, l = 1, 2, ..., L-1 is the frame index, the total number of frames

[0129] n′ is the time domain index, which is used to traverse the time domain signal. The value range is n′=0,1,…,N fft -1 is the serial number of the time domain sampling point in a single frame, which is used to traverse all time domain signals in a frame. It is only valid within the summation operation and indicates the traversal of each time domain point in a frame.

[0130] k is the frequency index, which is used to identify the frequency domain result and represents the frequency component position of the FFT output. The value range is k=0,1,…,N fft -1 is the frequency component number of the FFT output, each k corresponds to a frequency f k =k·Δf, representing all possible frequency components, covering the range from DC k=0 to Nyquist frequency k=N fft / 2 frequency band;

[0131] S fft [k,l] is the lth frame signal at frequency f k = The complex spectrum at k·Δf refers to the energy distribution of the signal in the frequency domain, including amplitude and phase information, S fft The larger [k,l] is, the higher the frequency f k The stronger the energy of the component, the higher the value. The window function is used to reduce spectrum leakage, that is, to suppress the frequency domain diffusion caused by the discontinuity of the frame edge signal. After the Hamming window is weighted, the weight of the frame center signal is high and the edge weight is low.

[0132] The frequency resolution is calculated as:

[0133]

[0134] Among them, f k is the actual frequency corresponding to the frequency index k, Δf is the frequency resolution of the FFT output; the effective frequency band is limited to 920MHz≤f k k value for ≤925MHz:

[0135] Δf is the minimum frequency interval that FFT can distinguish. The smaller its value, the higher the frequency resolution, but requires a longer FFT point number. The effective frequency band screening only analyzes the tag operating frequency band, i.e., 920-925MHz, to avoid invalid calculations.

[0136] The logic of using the frequency domain features to calculate the maximum amplitude, peak frequency and harmonic distribution data is as follows: for the maximum amplitude, in the effective frequency band [k min ,k max ], based on the formula:

[0137]

[0138] Among them, A max [l] represents the maximum amplitude of the l-th frame signal in the effective frequency band; traverse all frequency indexes k of each frame, in the effective frequency band [k min ,k max ]Search for the maximum value within;

[0139] A max [l] is the maximum energy intensity of the lth frame signal in the effective frequency band. The larger its value is, the stronger the tag reflection signal is, or the closer the tag is to the antenna;

[0140] For the peak frequency, which is the frequency corresponding to the maximum amplitude, the formula is:

[0141] f peak [l] = k peak [l]·Δf

[0142] Among them, f peak [l] represents the main frequency of the first frame, is the index corresponding to the maximum amplitude of the lth frame;

[0143] f peak [l] is the frequency component with the strongest signal energy in the first frame, which refers to the tag resonant frequency or the main frequency of environmental interference;

[0144] For harmonic distribution, in the effective frequency band [k min ,k max ], extract the frequency components with the first K amplitudes as candidate fundamental frequencies, and for each candidate fundamental frequency fcand,p [l]∈{f cand,1 [l],f cand,2 [l],…,f cand,K [l]}, where f cand,p [l] = k p [l]·Δf, k p [l] is the frequency index of the K frequencies before the amplitude sorting of the lth frame, k p ∈[1,K], for each candidate fundamental frequency f cand,p [l], calculate its harmonic score:

[0145]

[0146] Among them, H m [l](f cand,p [l]) is the mth harmonic amplitude of the candidate fundamental frequency of the lth frame, H1[l](f cand,p [l] is the amplitude of the first harmonic of the candidate fundamental frequency of the lth frame, m=2,3,…,M is the harmonic order, Score(f cand,p [1]) is the score of the candidate fundamental frequency of the first frame;

[0147] Score is the sum of the harmonic energy of the candidate fundamental frequency. The larger the value, the more significant the harmonic component of the candidate fundamental frequency, and the more likely it is to be the true fundamental frequency.

[0148] For the fundamental amplitude: H1[l]=|S fft [k base ,l]|, where k base is the index corresponding to the fundamental amplitude;

[0149] For harmonic amplitudes:

[0150]

[0151] The harmonic order limits are: where f high =925MHz; H m is the energy intensity of the mth-order harmonic, which refers to the periodic characteristics of the signal. For example, damage to the tag structure may cause harmonic anomalies.

[0152] Select the candidate fundamental frequency with the highest score as the final fundamental frequency of the lth frame:

[0153]

[0154] Among them, f base [l] is the fundamental frequency finally determined in the lth frame;

[0155] f cand,pRepresents the p-th candidate frequency in the l-th frame in the candidate fundamental frequency set, that is, one of the frequency components with the top K amplitudes extracted from the valid frequency band;

[0156] Combine the above three waveform feature values, take the average of L frame features, and calculate the waveform feature index F ind The formula is:

[0157]

[0158] in, is the average of the maximum amplitude of multiple frames in the signal spectrum, reflecting the signal energy intensity. is the average of the maximum amplitude of multiple frames, reflecting the main frequency of the signal. The corresponding frequency mean, is the mean value of the mth-order harmonic amplitude in multiple frames, reflecting the periodic distortion or abnormal resonance of the signal, m -p is the weighting of the harmonic order, which controls the contribution attenuation rate of high-order harmonics. A0=1 is the reference amplitude, which eliminates the dimension effect. f0=1 is the reference frequency, which is used for normalization. p is the harmonic attenuation weight. is the highest harmonic order in multiple frames, and

[0159] Molecular part Logarithmic amplitude compresses the numerical range. Logarithmic transformation converts the exponential growth of amplitude into linear growth to avoid extreme values ​​dominating the results. The larger the value, the stronger the signal energy, such as when the tag is closer to the antenna or the reflection efficiency is higher. The smaller the value, the weaker the signal energy, which may be due to tag shedding, obstruction or environmental interference.

[0160] Normalized peak frequency balances amplitude and frequency differences. A larger value indicates that the main frequency is closer to the upper frequency band limit of the device, such as 925 MHz. This may be due to changes in the tag's resonance characteristics or external high-frequency interference. A smaller value indicates that the main frequency is closer to the lower frequency band limit, such as 920 MHz, which may be normal.

[0161] Quantify the harmonic proportion. The higher the harmonic energy, the larger the denominator value, and the suppression characteristic index F ind ; The larger p is, the faster the weight of higher-order harmonics decays. For example, when p = 1, the weight of the second-order harmonic is 0.5, and the third-order is 0.33. When p = 0, all harmonic weights are equal. The larger the value, the more significant the harmonic component, which may indicate tag damage, metal interference, or multipath reflection. The smaller the value, the better the signal periodicity and the less harmonic component, which is a normal state. The constant 1 prevents division by zero errors. When there is no harmonic component, the denominator is 1 to avoid calculation anomalies.

[0162] Find It is the comprehensive health status of the signal, balancing the main signal strength and harmonic interference. The larger the value, the higher the signal strength, the clear main frequency and close to high frequency, and the less harmonic components. ind The smaller the value, the weaker the signal energy or the abnormal main frequency, and the significant harmonic components, which may be due to tag damage or environmental interference.

[0163] Step 3: Collect the reflection signal of the object surface tag in different historical states, and calculate the historical waveform characteristic index according to the method of steps 1 and 2. At the same time, record the state of the corresponding object surface tag, build a historical fingerprint template library, store the historical waveform characteristic index corresponding to each state, and calculate the similarity between the waveform characteristic index and each historical waveform characteristic index in the template library using the Euclidean distance algorithm;

[0164] The logic for collecting the reflected signals of the surface tags of items in different historical states and calculating the historical waveform characteristic index according to the methods in steps 1 and 2 is as follows:

[0165] Collect historical tag reflection signal data in the following states: normal state (the tag is intact and free of interference), damaged state (the tag surface is physically damaged), lost state (the tag is completely detached from the object surface), and interference state (metal obstruction or multipath reflection interference).

[0166] Calculate the historical waveform characteristic index F for each group of signals according to steps 1 and 2 ind,h , build fingerprint template library The storage structure is in is the historical label status of the oth template in the template library, which includes normal, damaged, lost and interference, and O is the number of templates in the template library;

[0167] The mapping relationship between the characteristics and states of historical data. Each template represents a signal feature under a known state and is used as a comparison benchmark for real-time detection. The larger the number of templates O, the more state change scenarios are covered, but the computational complexity increases.

[0168] Calculate the similarity distance D between the real-time feature and each template in the template library o :

[0169]

[0170] Among them, o is the template index in the template library; F ind is the real-time feature index, is the historical feature index of the oth template in the template library, D o is the overall difference between the real-time feature and the oth template in the template library, D o The larger the value, the greater the difference between the real-time feature and the template, and they may belong to different states.o The smaller it is, the more similar the real-time feature is to the template, and they may belong to the same state.

[0171] Step 4: Based on the historical data distribution, a dynamic similarity threshold is set and templates of different states are excluded. Templates of the same state with a Euclidean distance less than the threshold are selected as candidate templates. If the candidate templates are not empty, the state corresponding to the template with the smallest Euclidean distance is selected as the judgment result. If the candidate template is empty, the state is determined to be abnormal and the container opening alarm is triggered.

[0172] The logic for setting a dynamic similarity threshold based on historical data distribution and excluding templates in different states, and selecting templates with a Euclidean distance less than the threshold and belonging to the same state as candidate templates is as follows:

[0173] The same status is defined as candidate templates must belong to the preset valid status, i.e. normal, damaged, lost and interference status, and meet the threshold conditions of their own status. Through the "valid status" restriction, test data or other invalid status templates are excluded;

[0174] For each state z in the template library, the Euclidean distance distribution of similar state templates is statistically analyzed based on historical data, and a dynamic threshold is set.

[0175]

[0176] Among them, μ z is the mean Euclidean distance from all templates in state z to the feature center of the state, σ z is the standard deviation of the distance from all templates to the feature center in state z;

[0177] The state feature center is the feature index F of all templates for state z. ind,h , calculate its mean:

[0178]

[0179] Among them, O z is the number of templates in state z, o z =1,2,…,O z ;

[0180] Calculate the Euclidean distance from each template to the feature center So we can calculate μ z and σ z :

[0181]

[0182] Mean μ z :

[0183]

[0184] Standard deviation σ z :

[0185]

[0186] Reflects the upper limit of the reasonable fluctuation range of templates of the same state. Templates within the threshold are very likely to belong to state z, and templates beyond the threshold may be noise or other states. The larger the value, the more dispersed the distribution of templates of the same state (high variance) or the greater the overall deviation (high mean); μ z The larger it is, the greater the average distance between similar templates. The larger the z The larger the value, the greater the discreteness of the same template. The larger the value, the larger it will be; the 3σ principle covers 99.7% of the data under the normal distribution and excludes outliers;

[0187] The role of the dynamic threshold is to ensure that each template in the candidate template is within the reasonable fluctuation range of its state to avoid cross-state mismatching. For example, if the normal template distance threshold is 2.0 and the distance between the real-time feature and a normal template is 1.8, then it is retained; if the damaged template distance threshold is 2.5 and the distance between the real-time feature and a damaged template is 2.3, then it is retained; if the distance between the real-time feature and a normal template is 2.2 and exceeds the threshold of 2.0, then it is excluded.

[0188] The condition for screening candidate templates is: the Euclidean distance is less than the threshold of the state to which it belongs: Template status is the preset valid status: label status z o Including normal, damaged, lost, and interfered;

[0189] The candidate templates for screening are: And z o It is in a valid state;

[0190] D o is the Euclidean distance between the real-time feature and the template o, z o is the status label of template o, and Select is a set of candidate templates that meet the threshold and have valid status. It filters out templates that may be similar to the real-time features and excludes invalid or noise templates; This is to ensure that the template is within a reasonable fluctuation range of its state. o The valid state is to exclude test data or undefined state;

[0191] If the candidate template is not empty, select the template corresponding to the state with the smallest Euclidean distance as the judgment result:

[0192]

[0193] Among them,* is the template index with the smallest Euclidean distance among the candidate templates; It is the label state that is most likely to correspond to the real-time signal, which means selecting the template state that is most similar to the real-time feature as the judgment result;

[0194] If multiple templates have the same minimum distance but different states, they are marked as "state conflict" and require manual review; if the candidate template is empty, that is, all templates do not meet the threshold conditions, it is judged as "unknown anomaly";

[0195] The risk set is set to E, which includes the following risks: tag damage, loss, interference, and unknown anomalies; the alarm condition judgment conditions are:

[0196]

[0197] At that time, when Risk = 1, a container unpacking and inspection request is triggered and abnormal information is recorded. When Risk = 0, the tag is in normal state and only the status log is updated without triggering any operation.

[0198] Risk is the result of an alarm trigger and can be used to monitor and manage risks in real time. Through this mechanism, automated alarms can be issued for tag status to facilitate response.

[0199] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and reflecting the software to obtain the latest real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0200] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0201] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment. The above description is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of this application.

Claims

1. A method for verifying items in a container based on wireless radio frequency technology, characterized in that: The specific steps include: Step 1: Use RFID equipment to transmit wireless signals to the tags on the surface of the items in the container to be monitored. The reflected signal is received by the tag. After receiving the reflected signal, the signal strength is enhanced by a low-noise amplifier, and a bandpass filter is used to filter out environmental noise to retain the tag's effective frequency band signal; Step 2: Use an analog-to-digital converter to convert the filtered reflection signal into a time domain signal. Then use FFT to extract the time domain signal to obtain frequency domain features. The frequency domain features include maximum amplitude, peak frequency, and harmonic distribution data. The waveform characteristic index is calculated by combining these three characteristic values. Step 3: Collect the reflection signal of the object surface tag in different historical states, and calculate the historical waveform characteristic index according to the method of steps 1 and 2. At the same time, record the state of the corresponding object surface tag, build a historical fingerprint template library, store the historical waveform characteristic index corresponding to each state, and calculate the similarity between the waveform characteristic index and each historical waveform characteristic index in the template library using the Euclidean distance algorithm; Step 4: Set a dynamic similarity threshold based on the historical data distribution and exclude templates of different states. Select templates with a Euclidean distance less than the threshold and belonging to the same state as candidate templates. If the candidate templates are not empty, select the state corresponding to the template with the smallest Euclidean distance as the judgment result. If the candidate template is empty, the state is judged to be abnormal and the container opening alarm is triggered.

2. The method for verifying items in a container based on wireless radio frequency technology according to claim 1, characterized in that: The logic of using RFID equipment to transmit wireless signals to the tags on the surface of the items in the container to be monitored and receiving the reflected signals from the tags is as follows: The RFID device operates in the ultra-high frequency band, and its transmission frequency range is set to [920MHz, 925MHz] in accordance with China's radio frequency allocation standards. The RFID device adopts an intermittent transmission mode, which is set to transmit a signal every 0.5 seconds, with each transmission lasting 50 milliseconds, for a total of 10 transmissions, each lasting 5 seconds. The transmitting antenna and receiving antenna of the RFID device are separated. The transmitting signal covers the space inside the container, and the receiving antenna captures the tag reflected signal S rx (t); Receive reflected signal S rx After (t), the signal strength is enhanced by a low-noise amplifier and the environmental noise is filtered out by a bandpass filter. The logic of retaining the tag's effective frequency band signal is as follows: Use a low noise amplifier to increase the signal strength, and the amplified signal S amp The expression of (t) is: S amp (t)=G·S rx (t) Where G is the amplifier gain coefficient, which is set to G = 10dB; Use an analog bandpass filter to filter out non-operating frequency noise, based on the formula: S fil (t)=h BP (t)*S amp (t) Among them, h BP (t) is the impulse response function of the simulated bandpass filter, whose passband is 920 MHz and 925 MHz. The symbol * represents the time domain convolution operation.

3. The method for verifying items in a container based on wireless radio frequency technology according to claim 2, characterized in that: The logic of using an analog-to-digital converter to convert the filtered reflection signal into a digital time domain signal is: According to Nyquist sampling theorem, the sampling frequency f s Must be greater than twice the highest frequency of the signal, that is, f s ≥2·f high , f high =925MHz is the maximum frequency of the signal, select f s =2GHz, sampling period The duration of a single transmission is T = 50ms, and the total number of sampling points is The filtered signal S fil (t) is uniformly sampled, based on the formula: S dis [n]=S fil (t0+n·T s ),n=0,1,…,N-1 Among them, n is the time index, S dis [n] represents the nth sampling point of the discrete signal, t0 is the initial time, and N is the number of sampling data points; S dis The digital signal expression is: S dis ={S dis [0],S dis [1],…,S dis [N-1]} The logic of extracting time domain signals through FFT to obtain frequency domain features is: S dis Divide into L frames, set the length of each frame to N fft =2 20 = 1048576 points, calculate the FFT after adding window to each frame signal: Among them, k is the frequency index, which represents the kth discrete frequency component in the frequency domain. After framing, n′=0,1,…,N in each frame fft -1,s fft [k,l] is a complex number containing the amplitude and phase information of the frequency component, w[n′] is the window function, and l=1,2,…,L is the frame index; The frequency resolution is calculated as: Among them, f k is the actual frequency corresponding to the frequency index k, Δf is the frequency resolution of the FFT output; the effective frequency band is limited to 920MHz≤f k k value for ≤925MHz: The logic of using the frequency domain features to calculate the maximum amplitude, peak frequency and harmonic distribution data is as follows: for the maximum amplitude of the first frame, in the effective frequency band [k min ,k max ], based on the formula: Among them, A max [l] represents the maximum amplitude of the l-th frame signal in the effective frequency band; traverse all frequency indexes k of each frame, in the effective frequency band [k min ,k max ]Search for the maximum value within; For the peak frequency, which is the frequency corresponding to the maximum amplitude, the formula is: f peak [l]=k peak [l]·Δf Among them, f peak [l] represents the main frequency of the first frame, is the index corresponding to the maximum amplitude of the lth frame; For harmonic distribution, in the effective frequency band [k min ,k max ], extract the frequency components with the first K amplitudes as candidate fundamental frequencies, and for each candidate fundamental frequency f cand,p [l]∈{f cand,1 [l],f cand,2 [l],…,f cand,K [l]}, where f cand,p [l] = k p [l]·Δf, k p [l] is the frequency index of the K frequencies before the amplitude sorting of the lth frame, k p ∈[1,K], for each candidate fundamental frequency f cand,p [l], calculate its harmonic score: Among them, H m [l](f cand,p [l]) is the mth harmonic amplitude of the candidate fundamental frequency of the lth frame, H1[l[(f cand,p [l]) is the first harmonic amplitude of the candidate fundamental frequency of the lth frame, m=2,3,…,M is the harmonic order, Score(f cand,p [1]) is the score of the candidate fundamental frequency of the first frame; For the fundamental amplitude: H1[l]=|S fft [k base ,l]|, where k base is the index corresponding to the fundamental amplitude; For harmonic amplitudes: The harmonic order limits are: where f high =925MHz; Select the candidate fundamental frequency with the highest score as the final fundamental frequency of the lth frame: Among them, f base [l] is the fundamental frequency finally determined in the lth frame; Combine the above three waveform feature values, take the average of L frame features, and calculate the waveform feature index F ind The formula is: in, is the average of the maximum amplitude of multiple frames in the signal spectrum, is the average of the maximum amplitude of multiple frames, The corresponding frequency mean, is the mean value of the mth order harmonic amplitude in multiple frames, m -p is the weighting of the harmonic order, A0=1 is the reference amplitude, f0=1 is the reference frequency, p is the harmonic attenuation weight, is the highest harmonic order in multiple frames, and 4. The method for verifying items in a container based on wireless radio frequency technology according to claim 3, characterized in that: The logic for collecting the reflected signals of the surface tags of items in different historical states and calculating the historical waveform characteristic index according to the methods in steps 1 and 2 is as follows: Collect historical tag reflection signal data in the following states: normal state (the tag is intact and free of interference), damaged state (the tag surface is physically damaged), lost state (the tag is completely detached from the object surface), and interference state (metal obstruction or multipath reflection interference). Calculate the historical waveform characteristic index F for each group of signals according to steps 1 and 2 ind,h , build fingerprint template library The storage structure is in is the historical label status of the oth template in the template library, which includes normal, damaged, lost and interference, and O is the number of templates in the template library; Calculate the Euclidean distance D from the real-time feature to the oth template in the template library o : Among them, o is the template index in the template library, F ind is the real-time feature index, is the degree history feature index of the oth template in the template library.

5. The method for verifying items in a container based on wireless radio frequency technology according to claim 4, characterized in that: The logic for setting a dynamic similarity threshold based on historical data distribution and excluding templates in different states, and selecting templates with a Euclidean distance less than the threshold and belonging to the same state as candidate templates is as follows: For each state z in the template library, the Euclidean distance distribution of similar state templates is statistically analyzed based on historical data, and a dynamic threshold is set. Among them, μ z is the mean Euclidean distance from all templates in state z to the feature center of the state, σ z is the standard deviation of the distance from all templates to the feature center in state z; The state feature center is the feature index F of all templates for state z. ind,h , calculate its mean: Among them, O z is the number of templates in state z, o z =1,2,…,O z ; Calculate the Euclidean distance from each template to the feature center to calculate μ z and σ z : The condition for screening candidate templates is: the Euclidean distance is less than the threshold of the state to which it belongs: Template status is the preset valid status: label status z o Including normal, damaged, lost, and interfered; The candidate templates for screening are: And z o It is in a valid state; If the candidate template is not empty, select the template corresponding to the state with the smallest Euclidean distance as the judgment result: Among them, * is the template index with the smallest Euclidean distance among the candidate templates, is the label state that the real-time signal is most likely to correspond to; If multiple templates have the same minimum distance but different status, they are marked as "status conflict" and require manual review. If the candidate template is empty, that is, all templates do not meet the threshold conditions, it is judged as "unknown anomaly"; The risk set is set to E, which includes the following risks: tag damage, loss, interference, and unknown anomalies; the alarm condition judgment conditions are: At that time, when Risk = 1, a container unpacking and inspection request is triggered and abnormal information is recorded. When Risk = 0, the tag is in a normal state, only the status log is updated, and no operation is triggered.

Images