A delivery whole-process monitoring method applied to a logistics delivery system

Abstract

The application discloses a delivery whole-process monitoring method applied to a logistics distribution system, relates to the technical field of logistics distribution monitoring, solves the technical problem that in the prior art, in the sorting link, only dominant violent sorting damage can be identified, and hidden internal damage of goods cannot be detected, and the application constructs a three-section whole-link monitoring system of sorting hidden internal damage intelligent identification, conveying weight dynamic monitoring and end distribution trajectory accurate correction, which covers core operation links such as logistics sorting, collection conveying and end distribution. Through multi-dimensional sensing data fusion, self-adaptive threshold regulation, space topology verification and multi-scene abnormality classification determination, fine, intelligent and quantitative monitoring of the whole process of logistics distribution is realized, the goods damage and leakage detection rate and the monitoring false alarm rate are greatly reduced, the abnormality traceability data chain is improved, the logistics distribution control precision and operation optimization efficiency are improved, and the method is suitable for various complex logistics operation scenes.

Description

Technical Field

[0001] This invention relates to the field of logistics and distribution monitoring technology, specifically to a method for monitoring the entire distribution process in a logistics and distribution system. Background Technology

[0002] With the rapid development of e-commerce, same-city instant delivery, and bulk logistics, logistics and distribution have formed a standardized full-chain operation system of sorting, pickup and transportation, trunk line and last-mile delivery. The volume of parcel delivery has grown explosively, and users' requirements for the integrity rate of goods, the authenticity of the trajectory, the accuracy of anomaly tracing, and the stability of delivery timeliness in logistics and distribution continue to increase.

[0003] Current end-to-end monitoring technology for logistics and distribution systems remains in a traditional stage characterized by partial monitoring, fixed thresholds, and limited data, failing to form a comprehensive, multi-dimensional, and adaptive intelligent monitoring system. In the sorting stage, only visible damage from rough handling can be identified, while hidden internal damage to goods cannot be detected. In the pickup and transportation stage, traditional fixed weight monitoring thresholds fail to adapt to the physical characteristics of packages and the transportation environment, leading to imbalances in the judgment of abnormalities in soft and hard packages, inability to distinguish between equipment errors and actual losses, and low identification rates for minor structural damage.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to solve the problems mentioned above by proposing a method for monitoring the entire delivery process in a logistics and distribution system.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for monitoring the entire delivery process in a logistics and distribution system is as follows:

[0008] Step 1: Collision detection during sorting. Collision detection is performed on the delivered items during the sorting process to identify the risk of internal damage to the delivered items.

[0009] Step 2: Weight monitoring during delivery. After sorting and collision identification, the delivery items are delivered. The weight of each pickup node in the entire delivery process is monitored, and delivery decisions are made based on the monitoring results.

[0010] Step 3: Delivery process monitoring. After sorting and picking up the packages, the items are delivered. During the delivery phase, the delivery process is monitored to identify the reasons for location data delays and to make delivery process monitoring and decision-making based on the reasons.

[0011] Furthermore, the process of collision identification in step one is as follows:

[0012] The delivery items to be sorted are marked as sorting pieces, and the sorting time period is extracted to obtain the standardized time-domain sound pressure signal of the surface physical coordinate points of the sorting pieces within the sorting time period;

[0013] Perform a fast Fourier transform on each time-domain sound pressure signal to obtain the one-sided power spectral density and the frequency domain result in complex form; take the square of the modulus of the complex spectrum and divide it by the frequency resolution to obtain the two-sided power spectral density, and construct a two-sided power spectral density sequence.

[0014] Two non-overlapping frequency bands are defined within the full frequency band: the low-frequency base band and the initial wideband high-frequency band of interest;

[0015] Integrating these components separately yields the low-frequency energy base and the high-frequency energy lines that retain their complete morphology.

[0016] Furthermore, the low-frequency energy substrate is standardized across the entire field to obtain a dynamic modulation factor; based on this, a unique high-frequency analysis window is dynamically generated for each point; the center frequency of the window drifts linearly with the dynamic modulation factor, and the window bandwidth adaptively widens with the absolute value of the dynamic modulation factor; thus determining the dynamic high-frequency analysis interval of the impact point.

[0017] Furthermore, based on the dynamic high-frequency analysis interval and high-frequency energy spectral lines, dynamic window attenuation slope extraction and intermediate state mapping are performed;

[0018] The first transformation, namely linear fitting, involves selecting pairs of points within the dynamic range in the power spectral density domain and fitting a straight line using univariate linear regression to obtain the dynamic high-frequency attenuation slope.

[0019] The second transformation, namely nonlinear mapping, is to calculate the moving average of all dynamic high-frequency attenuation slopes in the current batch as the dynamic baseline, and to calculate the variance of the slope across the entire field to set the shape parameter α; the dynamic high-frequency attenuation slopes are then substituted into the improved S-shaped mapping function to generate the initial internal loss potential index.

[0020] Furthermore, the initial internal damage potential index for each impact point is extracted, and its value range is nonlinearly compressed to [0,1);

[0021] Based on the initial internal damage potential index, the physical coordinates of the impact point and the preset spatial neighborhood radius are extracted simultaneously.

[0022] For each impact point i, find all neighboring points h in its physical neighborhood; compute the discrete Laplace operator of the potential field at the impact point to quantify the spatial consistency between the point and its neighborhood.

[0023] Based on this difference, an exponential spatial consistency penalty is applied to the initial potential index to obtain the final judgment index; the final internal loss potential index matrix is ​​obtained after spatial topology consistency verification; when the value of the final internal loss potential index matrix of the impact point exceeds the set red line value, the coordinates of the corresponding impact point and the sorting execution time generated by the impact point are extracted and uploaded to the administrator terminal for sorting item detection and sorting execution adjustment.

[0024] Furthermore, the process of monitoring the conveying weight in step two is as follows:

[0025] Identify each pickup node in the entire delivery process and assign a label j; in the pickup stage, obtain the package volume and initial weight, and calculate the density; identify the packaging material visually, assign a structural fragility coefficient, and define a mass inertia factor;

[0026] When a package arrives at node j, obtain the weight of a single package and the median of the weight distribution of packages of the same type in the same batch, calculate the relative deviation rate within the batch, obtain the corrected weight after environmental error concealment, and append the corrected weight value to the historical sequence;

[0027] The upper bound of normal fluctuations for each node is dynamically generated using the physical properties of the package, D.

[0028] Furthermore, based on the corrected weight sequence and dynamic threshold, the cumulative negative drift deviation is recursively calculated; and the hidden crack risk index is collected.

[0029] When the risk index of microcracks is <0.5: maintain routine monitoring;

[0030] When the hidden crack risk index ∈ [0.5, 1.0): trigger outer loop feedback modulation, that is, the weighing sampling rate of subsequent nodes is increased from a single measurement to three repeated measurements, and the variance of the three measurements is included in the confidence interval calculation of the next node;

[0031] When the hidden crack risk index is ≥1.0: trigger abnormal output, lock the current node as a structural break point, automatically increase the re-inspection priority of the package at downstream nodes, and use the hidden crack risk index evolution curve as a quantitative evidence chain for liability determination.

[0032] Furthermore, the process of monitoring the delivery process in step three is as follows:

[0033] The system acquires real-time GPS location data of delivery personnel, specifically their location, and collects inertial trajectory data of equipment, specifically historical delivery trajectories; it identifies abrupt changes, jumps, and breaks in the GPS trajectory and marks them as stationary segments; it records the distribution of stationary and non-stationary segments and analyzes the distributed trajectories.

[0034] By comparing non-stationary segments with the inertial travel trajectory of the equipment, if the overlapping trajectory exceeds the set trajectory distance threshold, the current distribution trajectory is set as an inertial trajectory; otherwise, it is marked as a non-inertial trajectory.

[0035] Furthermore, based on the geofence data of the delivery address business district and the inertial trajectory, the intersection points are identified and marked as obstruction points. If there are obstruction points in the stationary segment of the inertial trajectory, the current delivery is marked as a route obstruction type, a delay warning is issued, and route planning is performed based on the obstruction points. The delivery personnel delivering in the same business district are also given a route adjustment warning. If there are no obstruction points in the stationary segment of the inertial trajectory, the current delivery is marked as a transportation obstruction delivery, and the current delivery is marked as a delivery obstruction type, that is, the current stationary segment has a route deviation risk.

[0036] Furthermore, if the frequency of stagnant segments in the non-inertial trajectory continues to increase, it is inferred that the current GPS positioning data is normal, and a delivery delay warning is issued; conversely, if the frequency of stagnant segments in the non-inertial trajectory does not continue to increase, it is inferred that the current GPS positioning data is abnormal, and the fluctuation of GPS positioning data is determined to be positioning interference.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] 1. By using FFT frequency domain transformation and power spectral density normalization, the influence of sampling parameters and batch differences on the detection data is eliminated, ensuring the comparability of detection data across scenarios and batches. By separating low-frequency macroscopic impact energy and high-frequency microstructural response energy, the system accurately distinguishes between the external force impact state during sorting and the internal structural damage state of goods, achieving internal damage identification from a physical mechanism perspective. Based on the low-frequency energy base, an adaptive high-frequency analysis window is dynamically generated to adapt to sorting scenarios with different impact forces and contact stiffness, solving the problems of poor adaptability and low detection accuracy of fixed analysis windows. Through linear slope fitting and improved S-shaped nonlinear mapping, a normalized internal damage potential index is quantified to achieve quantitative judgment of the degree of damage, rather than the traditional binary judgment of presence or absence. Finally, it can accurately output full-dimensional traceability data such as the coordinates of the damage impact point, damage level, and sorting execution time, and upload it to the backend in real time to complete the dynamic adjustment of sorting operations, reducing hidden cargo damage in the sorting process from the source.

[0039] 2. By fusing equipment accuracy parameters with environmental loss baselines to calculate the upper limit of dynamic fluctuations, the interference of equipment measurement errors and natural environmental losses on weight monitoring is effectively eliminated, significantly improving the accuracy of identifying actual cargo damage anomalies. Through recursive cumulative negative drift deviation calculation, only continuous and trend-based weight losses are judged, while instantaneous random fluctuations are automatically zeroed, further avoiding noise interference. Based on the hidden crack risk index, anomaly risk classification and control are implemented. In low-risk states, data collection accuracy is automatically enhanced, while in high-risk states, abnormal nodes are locked in real time and the priority of re-inspection is increased, forming a graded, precise, and intelligent control mechanism.

[0040] 3. By fusing and verifying GPS positioning data and device inertial walking trajectory data, accurate correction of delivery trajectories in complex urban scenarios is achieved, completely solving the problem of false anomalies caused by GPS signal drift and disconnection; by classifying stationary segments and distinguishing trajectory types, three types of abnormal scenarios are accurately identified: GPS positioning interference, road condition obstruction, and human route deviation, so as to achieve accurate qualitative analysis of delivery anomalies and eliminate invalid warnings and scheduling misjudgments. Attached Figure Description

[0041] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0042] Figure 1 This is a block diagram of the method logic architecture of the present invention;

[0043] Figure 2 This is a flowchart of the method in step one of this invention;

[0044] Figure 3 This is a flowchart of the method for step two in this invention. Detailed Implementation

[0045] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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.

[0046] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0047] Please see Figures 1-3 As shown, a method for monitoring the entire delivery process in a logistics and distribution system is applied. The specific steps of this method are as follows:

[0048] Step 1: Sorting Collision Detection. This step monitors the initial stage of the logistics and distribution system, namely the sorting step. In the sorting stage, the delivered items need to be classified. However, the existing technology only relies on video vision and acceleration overload sensors for sorting detection. Only when the sensor detects that the instantaneous acceleration exceeds the physical threshold will it be judged as violent sorting. For sorting that does not exceed the threshold, it is impossible to identify hidden internal damage, which leads to damage inside the delivered items and cannot be reported in time, increasing the difficulty of traceability in the entire delivery process.

[0049] The delivery items to be sorted are marked as sorting pieces, and the sorting time period is extracted. The standardized time-domain sound pressure signal Snorm(t,i) of the N physical coordinate points Pi on the surface of the sorting piece during the sorting time period is obtained. Specifically, it is the amplitude of the standardized preprocessed time-domain sound pressure signal collected at the i-th physical impact point at the t-th discrete sampling time.

[0050] Perform a Fast Fourier Transform (FFT) on each time-domain sound pressure signal to obtain the one-sided power spectral density P(f,i); the specific process is as follows: based on the standardized time-domain sound pressure signal Snorm(t,i), where t=1,2,…,T (T=1024), the sampling rate is F s =48kHz; Perform a Fast Fourier Transform (FFT) with T points on each normalized time-domain sequence Snorm(t,i) to obtain the complex frequency-domain result X(k,i):

[0051] ;

[0052] Where k is the frequency index, k=1,2,…,T, and the corresponding analog frequency is f(k)=(k−1)×F s / T. This complex result contains the amplitude and phase information of each frequency component; and is labeled as a complex spectrum matrix X(k,i);

[0053] Take the square of the modulus of the complex spectrum X(k,i) and divide it by the frequency resolution Δf=F. s / T, yielding the two-sided power spectral density (in V). 2 / Hz):

[0054] ;

[0055] The physical significance of this step lies in normalizing the square of the complex amplitude (in terms of energy) to a unit frequency bandwidth, so that the integration result no longer depends on the number of FFT points T, thereby ensuring the comparability of subsequent data across batches and sampling parameters; and constructing a two-sided power spectral density sequence. ;

[0056] Based on the conjugate symmetry of real signals, the energy of negative frequency components is folded to positive frequencies. For frequency indices k=1 (corresponding to 0Hz, i.e., the DC component), the original value is retained; for k=2 to k=T / 2 (corresponding to the positive frequency part), its amplitude is multiplied by 2; the negative frequency part of k>T / 2 is discarded. The one-sided power spectral density P(f k The calculation logic for i) is as follows:

[0057] ;

[0058] Among them, f k =(k−1)×F s / T, and f k ∈[0,F s / 2]. After this transformation, integrating P(fk,i) over any frequency band yields a total power that is strictly consistent with the total power of the original time-domain signal (satisfying Parseval's theorem). This yields the one-sided power spectral density P(f,i). Here, f ranges from 0 Hz to the Nyquist frequency (i.e., 24 kHz).

[0059] Subsequently, two non-overlapping frequency bands are defined across the entire frequency band:

[0060] Low-frequency baseband: ;

[0061] Initial wideband and highbandwidth band of interest: ;

[0062] The low-frequency energy substrate E is obtained by integration. Low (i) and the high-frequency energy spectral line P that retains its complete form High (f,i); specifically: low-frequency energy substrate E Low (i): Define the low-frequency extraction interval as f∈[f LL ,f LH Integrating and summing the power spectral density within this frequency band yields a scalar value, namely... This parameter characterizes the macroscopic mechanical energy input to the component by the impact excitation and its local macroscopic impedance response. Its value is not sensitive to internal micro-damage, but is highly sensitive to external coupling states such as impact force and contact stiffness.

[0063] High-frequency energy spectral line P High (f,i): Set the initial high-frequency extraction interval as f∈[fLL ,f LH The power spectral density values ​​corresponding to all discrete frequency points within the frequency band are retained to form a one-dimensional array; this parameter fully records the fine spectral morphology of high-frequency sound waves after propagation and attenuation within the component, containing rich information about the material's microstructure response;

[0064] Low-frequency energy substrate E Low (i) Standardize across the entire field to obtain the dynamic modulation factor δ Low (i): ;

[0065] in, ,

[0066] .

[0067] The δ Low (i) reflects the degree to which the local mechanical impedance at the current impact point deviates from the average state across the entire field.

[0068] Based on this, a unique high-frequency analysis window is dynamically generated for each point i. The center frequency f of the window is... center (i) with δ Low (i) Linear drift, window bandwidth Δf win (i) with δ Low (i) Absolute value adaptive broadening:

[0069] ;

[0070] ;

[0071] Where f0 is the reference center frequency (e.g., 10kHz), Δf dev For the maximum frequency deviation (e.g., 3kHz), k c The global coupling coefficients (estimated in real time by least-squares fitting), Δf base η is the reference bandwidth (e.g., 4kHz), and η is the bandwidth modulation coefficient (e.g., 0.15).

[0072] Determine the dynamic high-frequency analysis interval of the impact point [f] start (i),f end (i)], where:

[0073] f start (i)=f center (i)−Δf win (i) / 2, f end (i)=f center (i)+Δf win (i) / 2.

[0074] Based on the dynamic high-frequency analysis interval [f start (i),f end (i)] and high-frequency energy spectral line P High (f,i) performs dynamic window in-window decay slope extraction and intermediate state mapping;

[0075] First transformation (linear fitting): In the power spectral density domain (vertical coordinate y=10log) 10 P High Within (f,i), in dB), extract the pair of points (f) located within the dynamic range. m ,y m The line y=K is fitted using univariate linear regression (least squares method). High (i)×f+b, yielding the dynamic high-frequency attenuation slope K. High (i): ;

[0076] Second transformation (nonlinear mapping): Calculate all K values ​​in the current batch. High (i) The moving average is used as the dynamic baseline β (update rule: ), and calculate the variance of the slope over the entire field. To set the shape parameter α = α0 / (σ) K +ϵ). Finally, the slope K High (i) Substitute the improved S-shaped mapping function to generate the initial internal loss potential index Γ. init (i): ;

[0077] Extract the initial internal damage potential index Γ at each impact point init (i), whose range is non-linearly compressed to [0,1);

[0078] Based on the initial internal loss potential index Γ init (i) synchronously extract the physical coordinates (x) of the impact point. i ,y i ), the preset spatial neighborhood radius R;

[0079] For each impact point i, find its physical neighborhood. Find all neighboring points h within the potential field. Calculate the discrete Laplace operator for the potential field at that point. This is used to quantify the spatial consistency (degree of distortion) between the point and its neighborhood.

[0080] ;

[0081] Among them, M neighborΓ represents the number of valid points in the neighborhood. Physically, true internal loss causes smooth and continuous field distortion (small Laplace operator value), while random noise forms isolated spikes (large Laplace operator value). Based on this difference, an exponential spatial consistency penalty is applied to the initial potential index to obtain the final judgment index Γ. final (i): ;

[0082] Where λ is a preset penalty intensity coefficient (e.g., λ=0.05); if the absolute value of the Laplace operator at a certain point is extremely large, its exponential term will rapidly decay to 0, thereby effectively suppressing the confidence of isolated noise points.

[0083] The final internal loss potential index matrix Γ after spatial topological consistency verification final (i); when the final internal damage potential index matrix Γ at the impact point final (i) If the value exceeds the set red line value, it indicates a higher probability of hidden internal damage at that point. Extract the coordinates of the corresponding impact point and the sorting execution time generated by the impact point, and upload them to the administrator terminal for sorting item detection and sorting execution adjustment;

[0084] Step 2: Monitor the weight of the package and identify each pickup node in the entire delivery process, assigning a label j. During the pickup stage, obtain the package volume V and the initial weight W0, and calculate the density ρ0 = W0 / V. Visually identify the packaging material and assign a structural fragility coefficient G (high value 1.1–1.2 for soft packaging / woven bags, low value 0.8–0.9 for hard boxes / foam boxes). Define the mass inertia factor: ;where ρ ref For empirical reference density (e.g., 0.3 g / cm³) 3 The physical meaning of D: The higher the density and the more compact the structure, the narrower the allowable range of normal fluctuations should be (the larger D is, the narrower the subsequent μ). j The larger the value, the more lenient the judgment—here defined as the magnitude of "inertia," i.e., the ability to resist change.

[0085] When the package arrives at node j, obtain the single package weight W. j and the median weight distribution of similar packages in the same batch Calculate the intra-batch relative deviation rate: ; Obtain the corrected weight after environmental error hiding: This correction value is added to the historical sequence. ;

[0086] The upper bound μ of normal fluctuations for each node is dynamically generated using the physical properties of the package. j :

[0087] ;

[0088] Where: σsensor The inherent measurement standard deviation of the dynamic scale at this node (equipment static accuracy parameter);

[0089] : The median absolute value of the weight change of all packages in the current batch between two adjacent nodes (reflecting the common loss baseline of the transportation environment).

[0090] kl: Environmental weighting coefficient (empirical value 0.3~0.5);

[0091] μ j With package density ρ0 and environmental baseline Dynamic changes, not a fixed constant. High-density encapsulation achieves smaller μ. j More sensitive to minute losses; low-density soft packaging achieves larger μ j It is not overwhelmed by its own physical deformation readings.

[0092] Based on the corrected weight sequence and dynamic threshold, the cumulative negative drift deviation is recursively calculated: S0=0, ;

[0093] in This represents the weight loss of adjacent nodes; the physical meaning of this recursive formula is that the system requires the cumulative loss to continuously exceed the upper limit of the dynamic allowable fluctuation, μ. j A positive accumulation is formed; otherwise, S j Zeroing the scale prevents frequent false alarms caused by inherent weighing noise.

[0094] Risk index for collecting microcracks: Where θ is the crack triggering threshold base (a fixed constant, e.g., θ = 0.5% × W0). The physical meaning of this index is: after normalizing the cumulative deviation to the triggering threshold, it is divided by the mass inertia factor D—low-density wrapping (small D) in the same S... j To obtain a higher Q j Earlier intervention is needed.

[0095] When Q j <0.5: Maintain routine monitoring, no special treatment required;

[0096] When Q j ∈[0.5,1.0): Trigger outer loop feedback modulation—the weighing sampling rate of subsequent nodes is increased from a single measurement to three repeated measurements, and the variance of the three measurements is included in the next node. Confidence interval calculation (data collection rules are dynamically modulated by the current risk status);

[0097] When Q j ≥1.0: Triggers abnormal output, locking the current node as a structural breakpoint (first Q). jFor nodes with i ≥ 1.0, automatically increase the re-inspection priority of the package on downstream nodes, and set Q... j Evolutionary curves serve as a quantitative chain of evidence for liability determination;

[0098] Step 3: Delivery process monitoring. After sorting and picking up the packages, the items are delivered. During the delivery phase, the delivery process is monitored to avoid positioning deviations that could cause errors in the real-time delivery time estimation. The existing monitoring relies solely on a single GPS positioning data to determine the compliance of the delivery trajectory. It has not been adapted and corrected for signal defects in densely populated urban areas. GPS drift and disconnection lead to a large number of false anomalies in last-mile delivery, making it impossible to accurately reflect the status of last-mile delivery operations.

[0099] Real-time acquisition of delivery personnel's GPS location data, specifically their location, and collection of equipment inertial walking trajectory data, specifically historical delivery trajectories;

[0100] Identify abrupt changes, jumps, and breaks in GPS trajectory segments and uniformly label them as stationary segments; record the distribution of stationary and non-stationary segments and analyze the distributed trajectories:

[0101] By comparing non-stationary segments with the inertial travel trajectory of the equipment, if the overlapping trajectory exceeds the set trajectory distance threshold, the current distribution trajectory is set as an inertial trajectory; otherwise, it is marked as a non-inertial trajectory.

[0102] Based on the geofence data of the delivery address business district and the inertial trajectory, the intersection point is identified and marked as an obstacle point. If there is an obstacle point in the stationary segment of the inertial trajectory, the current delivery is marked as a route obstacle type, a delay warning is issued, and route planning is performed based on the obstacle point. The route adjustment warning is also issued to delivery personnel in the same business district. If there are no obstacle points in the stationary segment of the inertial trajectory, the current delivery is marked as a transportation obstacle delivery, and the current delivery is marked as a delivery obstacle type, that is, the current stationary segment has a risk of route deviation.

[0103] If the frequency of stagnant segments in the non-inertial trajectory continues to increase, it is inferred that the current GPS positioning data is normal, and a delivery delay warning is issued; conversely, if the frequency of stagnant segments in the non-inertial trajectory does not continue to increase, it is inferred that the current GPS positioning data is abnormal, and the fluctuation of GPS positioning data is determined to be positioning interference.

[0104] In summary, this invention, through refined acoustic frequency domain analysis technology, achieves accurate identification of hidden internal micro-damage and structural cracks in the cargo sorting process, filling the technological gap in microscopic internal damage detection in logistics sorting. By introducing package density and material fragility coefficients to construct a mass inertia factor, it accurately matches the physical characteristics and damage resistance of different types of packages, achieving differentiated and accurate judgment of soft-pack and hard-pack packages, eliminating false alarms caused by physical deformation and missed alarms caused by minor structural damage. It also achieves fully dynamic adaptive control of weight anomaly judgment in the logistics transportation process, completely solving the technical problems of poor adaptability and high false alarm / missed alarm rates of traditional fixed threshold judgment.

[0105] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for monitoring the entire delivery process in a logistics distribution system, characterized in that, The full-process monitoring method is as follows: Step 1: Collision detection during sorting. Collision detection is performed on the delivered items during the sorting process to identify the risk of internal damage to the delivered items. Step 2: Weight monitoring during delivery. After sorting and collision identification, the delivery items are delivered. The weight of each pickup node in the entire delivery process is monitored, and delivery decisions are made based on the monitoring results. Step 3: Delivery process monitoring. After sorting and picking up the packages, the items are delivered. During the delivery phase, the delivery process is monitored to identify the reasons for location data delays and to make delivery process monitoring and decision-making based on the reasons.

2. The method for monitoring the entire delivery process in a logistics distribution system according to claim 1, characterized in that, The collision detection process in step one is as follows: The delivery items to be sorted are marked as sorting pieces, and the sorting time period is extracted to obtain the standardized time-domain sound pressure signal of the surface physical coordinate points of the sorting pieces within the sorting time period; Perform a fast Fourier transform on each time-domain sound pressure signal to obtain the one-sided power spectral density and the frequency domain result in complex form; take the square of the modulus of the complex spectrum and divide it by the frequency resolution to obtain the two-sided power spectral density, and construct a two-sided power spectral density sequence. Two non-overlapping frequency bands are defined within the full frequency band: the low-frequency base band and the initial wideband high-frequency band of interest; Integrating these components separately yields the low-frequency energy base and the high-frequency energy lines that retain their complete morphology.

3. The method for monitoring the entire delivery process in a logistics distribution system according to claim 2, characterized in that, The low-frequency energy substrate is standardized across the entire field to obtain the dynamic modulation factor; based on this, a unique high-frequency analysis window is dynamically generated for each point; the center frequency of the window drifts linearly with the dynamic modulation factor, and the window bandwidth adaptively widens with the absolute value of the dynamic modulation factor; thus, the dynamic high-frequency analysis interval of the impact point is determined.

4. The method for monitoring the entire delivery process in a logistics distribution system according to claim 3, characterized in that, Based on the dynamic high-frequency analysis interval and high-frequency energy spectral lines, perform dynamic window decay slope extraction and intermediate state mapping; The first transformation, namely linear fitting, involves selecting pairs of points within the dynamic range in the power spectral density domain and fitting a straight line using univariate linear regression to obtain the dynamic high-frequency attenuation slope. The second transformation, namely nonlinear mapping, is to calculate the moving average of all dynamic high-frequency attenuation slopes in the current batch as the dynamic baseline, and to calculate the variance of the slope across the entire field to set the shape parameter α; the dynamic high-frequency attenuation slopes are then substituted into the improved S-shaped mapping function to generate the initial internal loss potential index.

5. The method for monitoring the entire delivery process in a logistics distribution system according to claim 4, characterized in that, The initial internal damage potential index of each impact point is extracted, and its value range is nonlinearly compressed to [0,1); Based on the initial internal damage potential index, the physical coordinates of the impact point are extracted simultaneously, and the spatial neighborhood radius is preset. For each impact point i, find all neighboring points h within its physical neighborhood; Calculate the discrete Laplace operator of the potential field at the impact point to quantify the spatial consistency between the point and its neighborhood. Based on this difference, an exponential spatial consistency penalty is applied to the initial potential index to obtain the final judgment index. The final internal loss potential index matrix after spatial topology consistency verification; when the value of the final internal loss potential index matrix of the impact point exceeds the set red line value, the corresponding impact point coordinates and the sorting execution time generated by the impact point are extracted and uploaded to the administrator terminal for sorting item detection and sorting execution adjustment.

6. The method for monitoring the entire delivery process in a logistics distribution system according to claim 1, characterized in that, The process of monitoring the conveying weight in step two is as follows: Identify each pickup node in the entire delivery process and assign a label j; in the pickup stage, obtain the package volume and initial weight, and calculate the density; identify the packaging material visually, assign a structural fragility coefficient, and define a mass inertia factor; When a package arrives at node j, obtain the weight of a single package and the median of the weight distribution of packages of the same type in the same batch, calculate the relative deviation rate within the batch, obtain the corrected weight after environmental error concealment, and append the corrected weight value to the historical sequence; The upper bound of normal fluctuations for each node is dynamically generated using the physical properties of the package, D.

7. The method for monitoring the entire delivery process in a logistics distribution system according to claim 6, characterized in that, Based on the modified weight sequence and dynamic threshold, the cumulative negative drift deviation is recursively calculated; and the hidden crack risk index is collected. When the risk index of microcracks is <0.5: maintain routine monitoring; When the hidden crack risk index ∈ [0.5, 1.0): trigger outer loop feedback modulation, that is, the weighing sampling rate of subsequent nodes is increased from a single measurement to three repeated measurements, and the variance of the three measurements is included in the confidence interval calculation of the next node; When the hidden crack risk index is ≥1.0: trigger abnormal output, lock the current node as a structural break point, automatically increase the re-inspection priority of the package at downstream nodes, and use the hidden crack risk index evolution curve as a quantitative evidence chain for liability determination.

8. The method for monitoring the entire delivery process in a logistics distribution system according to claim 1, characterized in that, The process of monitoring the delivery process in step three is as follows: The system acquires real-time GPS location data of delivery personnel, specifically their location, and collects inertial trajectory data of equipment, specifically historical delivery trajectories; it identifies abrupt changes, jumps, and breaks in the GPS trajectory and marks them as stationary segments; it records the distribution of stationary and non-stationary segments and analyzes the distributed trajectories. By comparing non-stationary segments with the inertial travel trajectory of the equipment, if the overlapping trajectory exceeds the set trajectory distance threshold, the current distribution trajectory is set as an inertial trajectory; otherwise, it is marked as a non-inertial trajectory.

9. The method for monitoring the entire delivery process in a logistics distribution system according to claim 8, characterized in that, Based on the geofence data of the delivery address business district and the inertial trajectory, the intersection point is identified and the intersection point is marked as an obstacle point. If there is an obstacle point in the stagnant segment of the inertial trajectory, the current delivery is marked as a route obstacle type, a delay warning is issued, and the route is planned based on the obstacle point. The route adjustment warning is also issued to the delivery personnel in the same business district. If there are no obstruction points in the stationary segments of the inertial trajectory, the current delivery is marked as a transportation obstruction delivery, and the current delivery is marked as a delivery obstruction type, that is, the current stationary segment has a risk of route deviation.

10. The method for monitoring the entire delivery process in a logistics distribution system according to claim 9, characterized in that, If the frequency of stagnant segments in the non-inertial trajectory continues to increase, it is inferred that the current GPS positioning data is normal, and a delivery delay warning is issued; conversely, if the frequency of stagnant segments in the non-inertial trajectory does not continue to increase, it is inferred that the current GPS positioning data is abnormal, and the fluctuation of GPS positioning data is determined to be positioning interference.

Images