Case packing and stacking into warehouse whole-process tracing method for cold-chain fresh food category

Abstract

The application discloses a full-process tracing method for case stacking and warehousing of cold-chain fresh categories, and relates to the technical field of cold-chain logistics tracing. The method comprises the following steps: acquiring the arrangement pose image of the box body of the cold-chain fresh category in the stacking layer, and analyzing the box body edge contour; extracting the corner point space coordinates of each box body and calculating the orthographic projection overlap; when the orthographic projection overlap exceeds the preset pose threshold, the bottom temperature measurement label sequence of the corresponding box body and the side wall temperature measurement label sequence of the adjacent box body in the same layer are called to perform temperature difference time sequence pairing, and the in-layer temperature difference offset record is generated; the local hot area position is calibrated according to the in-layer temperature difference offset record, the box body group to be traced is demarcated, and the warehousing time stamp breakpoint of each box body is extracted. The method can accurately locate the local hot area caused by the box body pose abnormality in the stacking layer, associate the warehousing time breakpoint of the hot area box body with the batch code, and realize the traceability of the box body.

Description

Technical Field

[0001] This invention relates to the field of cold chain logistics traceability technology, specifically to a method for tracing the entire process of packing, palletizing, and warehousing for cold chain fresh produce. Background Technology

[0002] In the packing, palletizing, and warehousing of fresh produce in the cold chain, pallets are typically used to carry multiple boxes for bulk storage and transport. Because fresh produce is extremely sensitive to temperature, it is essential to record and trace the temperature status during the warehousing process. Existing traceability methods generally link the temperature records of the entire pallet to pallet labels or batch codes. When quality issues occur, they can only pinpoint the batch level of the pallet, not the specific box. Especially during palletizing, boxes may experience compression, tilting, or other positional shifts, obstructing airflow between the bottom and side walls of the boxes and creating localized heat accumulation areas. For detecting these localized heat areas, existing solutions mostly rely on deploying temperature sensors at fixed locations within the warehouse or performing random infrared temperature measurements on the entire pallet. However, these methods cannot dynamically detect micro-environmental temperature changes within a single box within the pallet layer caused by abnormal posture. Regarding box posture detection, visual analysis of box arrangement has been used to some extent, but traditional visual inspection is often only used to identify load rates or external damage, without linking the overlap of the relative postures of the boxes with the multi-point temperature distribution within the boxes themselves. In terms of traceability chain construction, existing inbound traceability systems mostly simply bind pallet-level batch numbers to inbound time. If individual boxes within a pallet experience localized overheating leading to quality deterioration, the traceability granularity remains at the entire pallet level, making it difficult to quickly identify the specific affected boxes and the actual time period during which they experienced abnormal temperatures. The lack of a comprehensive traceability chain—from identifying abnormal postures to pinpointing localized hot spots, and then to extracting box-level timestamp breakpoints—results in information gaps and blind spots in the inbound traceability of cold chain fresh produce. The challenges to address in the full-process traceability of packing, palletizing, and inbound cold chain fresh produce include: how to trigger the temporal pairing of temperature differences at different locations within the boxes through visual analysis of the box palletizing posture, thereby automatically pinpointing the location of localized hot spots within the palletizing layer when posture abnormalities occur; and how to define traceability groups based on the locations of localized hot spots, extract timestamp breakpoints from the inbound time records of the grouped boxes, and bind this breakpoint information to the pallet batch code to form traceable box-level anchor points. Summary of the Invention

[0003] This invention provides a full-process traceability method for the packing, palletizing, and warehousing of cold chain fresh produce. The purpose is to solve the problems in the prior art where local hot zones caused by abnormal box posture within the palletizing layer cannot be automatically identified and associated with specific boxes, and where the traceability link can only reach the pallet batch level and cannot achieve traceability of warehousing time breakpoints at the box granular level.

[0004] To achieve the above objectives, this invention provides the following technical solution: This invention provides a full-process traceability method for the packing, palletizing, and warehousing of cold chain fresh produce, solving the problems of ineffective identification of abnormal box postures, difficulty in early detection of localized hot zones, and insufficient granularity for warehousing breakpoint traceability when cold chain boxes are tightly stacked. This method uses visual recognition of the spatial relationships between boxes, combined with the temperature difference drift characteristics of multi-position temperature measurement tags, to achieve precise positioning and time breakpoint binding of abnormal boxes within the palletizing layer, forming anchor point records that can be traced box by box.

[0005] In one technical solution of the present invention, a full-frame image of the pallet is acquired in a manner perpendicular to the top of the palletizing layer. After grayscale processing, a gradient amplitude response map is extracted. Candidate box regions are selected from the pixel connected domains where the gradient amplitude is greater than the background noise threshold. Line segment fitting is performed on the candidate box regions, and line segments with an angle less than a set angle deviation with the pallet edge direction are extracted as the box edge contour, thereby eliminating complex reflections and label interference in the palletizing environment and stably obtaining the box edge structure.

[0006] Furthermore, the intersection points of adjacent edge lines are extracted from the box's edge contour as candidate corner points, and these are aggregated into a set of corner point coordinates for each box. The area of ​​the quadrilateral enclosed by the four corner point coordinates is used as the orthographic projection area of ​​the box. The orthographic projection area of ​​the target box is then intersected with the orthographic projection areas of adjacent boxes on the same layer in the pallet plane coordinate system to calculate the orthographic projection overlap. This overlap directly reflects whether there are any posture deviations such as compression, tilting, or misalignment between the boxes, providing a reliable triggering condition for subsequent temperature difference analysis.

[0007] Preferably, when the overlap of the orthographic projection of the target container with any adjacent container exceeds a preset attitude threshold, the first RFID temperature measurement tag on the bottom surface of the target container is activated, and the bottom surface temperature value is read at a fixed sampling interval to form a bottom surface temperature measurement tag sequence. At the same time, the second RFID temperature measurement tag on the side wall of the adjacent container is activated, and the side wall temperature value is read at the same sampling interval to form a side wall temperature measurement tag sequence. The difference between the bottom surface temperature value and the side wall temperature value at the same sampling time is used to obtain a temperature difference time series array, and the array segment that exceeds the temperature difference fluctuation range is marked as an intra-layer temperature difference offset record. By using dual-point temperature measurement on the bottom surface and side wall, the local heat conduction difference caused by abnormal contact can be effectively captured, so that the temperature difference drift reflects both the degree of attitude overlap and the trend of thermal zone evolution.

[0008] As a preferred embodiment of the present invention, the cumulative temperature difference of each box within a continuous time window is calculated from the temperature difference offset record within the layer. The area where the box with the largest cumulative temperature difference value is located is marked as a local hot zone. Taking the box corresponding to this location as the center, all boxes on the pallet with a horizontal and vertical spacing not exceeding two box spacings are included in the box group to be traced. The warehousing time field in the RFID storage area of ​​each box in the group is read, and the time point with a time interval greater than twice the average warehousing time interval of adjacent boxes is extracted as the warehousing timestamp breakpoint. By constraining the hot zone diffusion range and identifying the warehousing time discontinuity points, boxes in the same batch that may have experienced chain breaks or delayed warehousing can be traced back.

[0009] Furthermore, batch identification codes are read from the RFID pallet tags on the pallet edge. These codes are then used to establish key-value pairs of traceability anchor records with the entry timestamps of each container, based on the container number. A container-level traceability anchor is constructed using the container number as the index node and the batch identification code and entry timestamp as node attributes. These container-level traceability anchors are then inserted into the end node of the entry traceability chain in chronological order of the entry timestamps. This extends traceability precision down to each individual container, building upon the pallet batch code, and the anchor insertion order reflects the actual timing of the entry process.

[0010] Preferably, the gradient magnitude response map is obtained by calculating the gradient components in the horizontal and vertical directions using the Sobel operator and then taking the modulus. The background noise threshold is taken as the sum of the mean and standard deviation of the gradient magnitudes of all pixels in the full-frame image of the pallet, to adapt to the edge extraction stability under different illumination variations in cold chain warehouses. The first RFID temperature measurement tag is attached to the geometric center of the bottom surface of the box, and the second RFID temperature measurement tag is attached to the midline position in the height direction of the side wall of the box. The reading timestamps of the bottom surface temperature value and the side wall temperature value are synchronized and calibrated by the local clock of the pallet to ensure the time alignment accuracy of the temperature difference time sequence array. The cumulative temperature difference value is obtained by summing the product of the absolute value of the temperature difference of each sampling point within a continuous time window and the sampling interval. The duration of the continuous time window is adaptively set according to the category code of the cold chain fresh food category, so that the hot zone determination can adapt to the differences in heat capacity of different categories. The inbound traceability chain is stored in the local edge computing node of the pallet carrier in the form of a blockchain ledger. Each box-level traceability anchor is accompanied by a writing timestamp and the writing node's digital signature when it is written, ensuring that the traceability information is tamper-proof and verifiable at the inbound stage.

[0011] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0012] During the palletizing layer box pose detection process, the box edge contours in the arrangement pose image are analyzed to extract the spatial coordinates of corner points. The overlap of each box's orthographic projection on the pallet plane is calculated. When this overlap exceeds a preset pose threshold, the temperature difference time sequence pairing between the corresponding box bottom temperature measurement tag and the adjacent box side wall temperature measurement tag in the same layer is triggered. The temperature difference time sequence array is obtained by subtracting the bottom temperature value and the side wall temperature value at the same sampling time. The array segment exceeding the temperature difference fluctuation range is extracted as the temperature difference offset record within the layer. Using this temperature difference offset record, the location of local hot areas within the palletizing layer can be calibrated, and the spatiotemporal correspondence between box pose anomalies and local temperature offsets can be directly established. Compared with the method of simply relying on the average temperature monitoring of environmental sensors, this bottom and side wall temperature difference pairing driven by the box's own pose overlap state can accurately capture the heat accumulation phenomenon caused by the narrowing of air gaps due to box compression and tilting. It avoids the masking of local overheated areas after the whole layer temperature homogenization process, making the calibration of local hot area locations have higher spatial resolution and real-time response. After determining the location of the local hot zone, a group of boxes to be traced is defined centered on this location. The entry timestamp breakpoints of each box are extracted from the group, and these breakpoints are bound to the batch identification code corresponding to the pallet, generating a box-level traceability anchor point. This anchor point is then written into the entry traceability linked list. By limiting the impact range of the abnormal hot zone to a group no more than two box intervals from the central box, and by segmenting the entry time field in the RFID storage area of ​​each box within this group to extract timestamp breakpoints with time intervals greater than twice the average entry time interval of adjacent boxes, the actual entry time segment points generated during non-continuous palletizing can be filtered out. The resulting traceability anchor point is not a uniform time tag for the entire pallet, but rather an independent time breakpoint sequence accurate to each box affected by the hot zone. This sequence is associated with the batch identification code and stored in a linked list structure. This allows for rapid reconstruction of the actual time trajectory of each box during the entry stage when fresh produce quality abnormalities occur, based on the box-level traceability anchor points. This enables precise traceability from pallet batch to individual box, narrowing the scope of problem investigation and reducing recall costs. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0014] Figure 1 It is a flowchart of a full-process traceability method for packing, palletizing, and warehousing of cold chain fresh produce;

[0015] Figure 2 This is a flowchart of the process for locating local hot zones and extracting warehouse entry timestamp breakpoints;

[0016] Figure 3 It is the probability density distribution curve of the overlap of the orthographic projection of the box;

[0017] Figure 4 This is a schematic diagram of the temperature difference offset record within the layer;

[0018] Figure 5 This is a schematic diagram showing the time interval between adjacent warehouse entries and the breakpoint of the warehouse entry timestamp. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, 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.

[0020] See Figure 1 This invention provides a method for full-process traceability of packing, palletizing, and warehousing for cold chain fresh produce. The method acquires images of the arrangement and pose of the palletizing layers, analyzes the edge contours of each box in the images, extracts the spatial coordinates of the box corners, and calculates the orthographic overlap of the boxes on the pallet plane. When the orthographic overlap exceeds a preset pose threshold, the method retrieves the bottom surface temperature measurement tag sequence of the corresponding box and pairs it with the side wall temperature measurement tag sequences of adjacent boxes in the same layer to generate an intra-layer temperature difference offset record. The intra-layer temperature difference offset record is used to calibrate the location of local hot zones within the palletizing layer. A group of boxes to be traced is defined centered on this location, and the warehousing timestamp breakpoints of each box are extracted from the group. These warehousing timestamp breakpoints are bound to the batch identification code corresponding to the pallet to form box-level traceability anchor points and written into the warehousing traceability linked list, thereby achieving full-process cold chain traceability from packing and palletizing to warehousing.

[0021] Example 1:

[0022] In practical implementation, the process of obtaining the arrangement and pose image of the boxes for cold chain fresh produce in the palletizing layer, and parsing the edge contours of the boxes in the arrangement and pose image, is achieved through the following methods:

[0023] The image acquisition device is fixedly installed perpendicular to the pallet layer to capture a full-frame image of the entire pallet area. The full-frame pallet image is then converted to grayscale to obtain a grayscale image. The Sobel operator is used to calculate the horizontal and vertical gradient components for each pixel in the grayscale image. The horizontal gradient component is denoted as... The vertical gradient component is denoted as ,according to Calculate the gradient magnitude of the corresponding pixel. ,in This represents the gradient intensity of each pixel in the gradient magnitude response map. The gradient magnitude of all pixels is combined to form the gradient magnitude response map.

[0024] Calculate the mean and standard deviation of the gradient magnitudes of all pixels in the full-frame image of the tray, and use the sum of the mean and standard deviations as the background noise threshold. Filter out pixels with gradient magnitudes greater than the background noise threshold from the gradient magnitude response map, and aggregate these spatially adjacent pixels to form pixel connected regions. Calculate the bounding rectangle of each pixel connected region, and use this bounding rectangle as the candidate box region.

[0025] For each candidate container area containing edge pixels, a line segment fitting algorithm is used to extract multiple fitted line segments. Line segments reflecting the pallet edge are identified from the full-frame image of the pallet, and their directions are taken as the pallet edge direction. From the multiple fitted line segments corresponding to the candidate container area, line segments with an angle less than a set angle deviation from the pallet edge direction are extracted. The set angle deviation is a pre-configured angle parameter, the value of which is determined based on the allowable angular offset of the pallet layers. The extracted set of line segments is used as the container edge contour.

[0026] Example 2:

[0027] In practical implementation, the process of extracting the spatial coordinates of the corner points of each box based on the edge contour of the box, and calculating the overlap of the orthographic projection of each box on the pallet plane, is achieved in the following way:

[0028] From the extracted box edge contour, a set of straight line segments constituting the complete box boundary is obtained. Each straight line segment in the set is represented by the coordinates of its two endpoints. For any two distinct straight line segments in the set, it is determined whether they intersect in their respective extension directions. If two straight line segments intersect on their respective extensions and the intersection point is within the neighborhood of their respective endpoints, the intersection point is identified as a candidate corner point. The radius of the neighborhood is determined based on the image acquisition resolution and the positioning error of the box edge extraction. The radius of the neighborhood is a circular area centered at the endpoints of the straight line segments with a radius of 5 pixels.

[0029] All candidate corner points belonging to the same container are grouped into a corner point coordinate set according to the container number. The corner point coordinate set contains the coordinates of the four corner points of the corresponding container in the pallet plane coordinate system. The four corner coordinates are denoted as follows: , , , ,in This represents the x-axis coordinate of the first corner point in the pallet plane coordinate system. This represents the vertical coordinate of the first corner point in the pallet plane coordinate system. This represents the x-axis coordinate of the second corner point in the pallet plane coordinate system. This represents the vertical coordinate of the second corner point in the pallet plane coordinate system. This represents the x-axis coordinate of the third corner point in the pallet plane coordinate system. This represents the vertical coordinate of the third corner point in the pallet plane coordinate system. This represents the x-axis coordinate of the fourth corner point in the pallet plane coordinate system. This represents the vertical coordinate value of the fourth corner point in the pallet plane coordinate system.

[0030] Optionally, the coordinates of the four corner points of the target box are read from the corner coordinate set, and then... , , , The area of ​​the quadrilateral enclosed by the coordinates of its four corner points is calculated using the polygon area formula, and the area of ​​the quadrilateral is denoted as . , The calculation expression is:

[0031]

[0032] in, Indicates the corner index number. The value of is an integer from 1 to 4. hour The value is 1. Indicates the first The x-axis coordinates of each corner point in the pallet plane coordinate system. Indicates the first The vertical coordinates of each corner point in the pallet plane coordinate system. Indicates the first The x-axis coordinates of each corner point in the pallet plane coordinate system. Indicates the first The vertical coordinates of each corner point in the pallet plane coordinate system. The calculated area of ​​the quadrilateral. The projected area of ​​the target box.

[0033] It is understandable that the coordinates of the four corner points of adjacent boxes on the same floor are read in the same way, and the orthographic projection area of ​​adjacent boxes on the same floor is calculated. The quadrilateral region corresponding to the orthographic projection area of ​​the target box and the quadrilateral region corresponding to the orthographic projection area of ​​the adjacent box on the same floor are intersected in the pallet plane coordinate system. The intersection region is the overlapping part of the two quadrilateral regions. The area value of the intersection region is calculated and used as the orthographic projection overlap.

[0034] See Figure 3 In the figure, the horizontal axis represents the orthographic overlap of the box on the pallet plane, in pixels, and the vertical axis represents the probability density of the corresponding orthographic overlap. In the legend, solid lines represent the probability density distribution of overlap for all boxes, dashed lines represent the probability density distribution of overlap for abnormal boxes, and dotted lines represent preset attitude thresholds. According to the description of Embodiment 2, the orthographic overlap is obtained by calculating the intersection area of ​​the quadrilateral regions corresponding to the orthographic projection areas of the target box and the adjacent boxes on the same layer.

[0035] As shown in the figure, the probability density of the orthographic overlap of all box pairs has a high peak in the low value range (approximately 0 to 2000 pixels). As the overlap value increases, the probability density decreases rapidly, indicating that the overlap of most box pairs is small, which is consistent with normal stacking arrangement. The preset attitude threshold is set at approximately 2000 pixels as the critical value for judging abnormal box attitude.

[0036] The probability density of the orthographic overlap of the abnormal box pair is mainly distributed in the range to the right of the preset attitude threshold (greater than 2000 pixels). The probability density curve maintains a high and slowly decreasing trend in this range, indicating that the abnormal box pair has a large amount of orthographic overlap, which means that there is obvious overlap between the boxes, exceeding the normal placement tolerance range.

[0037] Example 3:

[0038] In practical implementation, when the overlap of the orthographic projection exceeds a preset attitude threshold, the bottom surface temperature measurement tag sequence of the corresponding box is retrieved, and the bottom surface temperature measurement tag sequence is paired with the side wall temperature measurement tag sequence of the adjacent box in the same layer to generate an intra-layer temperature difference offset record. This process is achieved in the following way:

[0039] The overlap of the target box's orthographic projection with each adjacent box on the same floor is compared one by one with a preset attitude threshold. This preset attitude threshold is a pre-defined area value, determined as a percentage of the box's bottom surface area. Five percent of the bottom surface area is chosen as the specific value for the preset attitude threshold, ensuring that subsequent temperature difference data acquisition is triggered when there is an overlap exceeding five percent of the bottom surface area between the boxes. An activation command is generated when the overlap of the target box's orthographic projection with any adjacent box on the same floor exceeds the preset attitude threshold.

[0040] The first RFID temperature measurement tag on the bottom surface of the target container is activated according to the activation command. The first RFID temperature measurement tag is attached to the geometric center of the bottom surface of the container. The RFID reader sends a temperature reading command to the first RFID temperature measurement tag, and reads the bottom surface temperature value at a fixed sampling interval. The fixed sampling interval is set to 30 seconds. The reason for choosing a fixed sampling interval of 30 seconds is that the temperature change rate of cold chain fresh food can capture effective temperature difference fluctuation characteristics at a time granularity of 30 seconds. The bottom surface temperature values ​​read in each reading are arranged in chronological order according to the reading timestamp to form a bottom surface temperature measurement tag sequence.

[0041] Simultaneously, the second RFID temperature measurement tag on the adjacent side wall of the enclosure is activated according to the activation command. The second RFID temperature measurement tag is attached to the midline position in the height direction of the side wall of the enclosure. The RFID reader sends a temperature reading command to the second RFID temperature measurement tag, and reads the side wall temperature value at the same fixed sampling interval. The side wall temperature values ​​read in each reading are arranged in chronological order according to the reading timestamp to form a side wall temperature measurement tag sequence.

[0042] It is understandable that the timestamps for reading the bottom surface temperature value and the side wall temperature value are synchronized and calibrated by the local clock of the tray. The local clock of the tray is synchronized and calibrated by periodically aligning the local clock of the tray with the clock source of the RFID reader. A clock synchronization operation is performed before each temperature reading, and the synchronization accuracy is controlled at the millisecond level.

[0043] The difference between the bottom surface temperature and the side wall temperature at the same sampling time is recorded as the bottom surface temperature. The sidewall temperature value is recorded as ,in Indicates the sampling time sequence number. The value of is a positive integer starting from 1. Indicates the first The bottom surface temperature value read at each sampling time. Indicates the first The sidewall temperature value read at the sampling time. Calculate the... absolute value of temperature difference at each sampling time absolute value of temperature difference The calculation formula is:

[0044]

[0045] Arrange the absolute values ​​of temperature differences corresponding to all sampling times according to the sampling time sequence to obtain the temperature difference time series array.

[0046] Optionally, the upper and lower limits of the temperature difference fluctuation range are determined based on the allowable temperature fluctuation range of the cold chain fresh food category. The upper limit of the temperature difference fluctuation range is the absolute value of the difference between the maximum and minimum allowable temperature fluctuation values, and the lower limit is zero. Each absolute value of the temperature difference in the temperature difference time series array is compared with the temperature difference fluctuation range. When the absolute value of the temperature difference exceeds the temperature difference fluctuation range, the corresponding sampling time and the consecutive array segment consisting of the sampling time before and after it are marked as an intra-layer temperature difference offset record. The intra-layer temperature difference offset record includes the sampling time number where the temperature difference exceeds the fluctuation range, the corresponding bottom surface temperature value, the corresponding side wall temperature value, and the absolute value of the temperature difference.

[0047] See Figure 4 In the graph, the horizontal axis represents the sampling time number q, ranging from 1 to 500, and the vertical axis represents the absolute temperature difference ΔTq, in degrees Celsius (°C), ranging from 0 to 6°C. The solid curve represents the absolute temperature difference ΔTq corresponding to each sampling time, calculated as the absolute difference between the bottom surface temperature and the sidewall temperature. The dashed lines in the graph represent the upper and lower limits of the temperature difference fluctuation range, with the lower limit at 0°C and the upper limit at 3°C. These limits are determined based on the allowable temperature fluctuation range for cold chain fresh produce. Sampling points marked with red crosses in the graph record the intra-layer temperature difference offset, i.e., the sampling time when the absolute temperature difference exceeds the preset fluctuation range.

[0048] From the curve trends, between sampling time numbers 1 and 300, the absolute value of the temperature difference ΔTq mostly remained between 0 and 3℃, with no red markings, indicating that the temperature difference was within the allowable fluctuation range and no abnormal fluctuations occurred. Between sampling time numbers approximately 300 and 500, the absolute value of the temperature difference showed a significant upward trend, frequently exceeding the upper limit of the 3℃ fluctuation range, accompanied by a large number of temperature difference offset records within the red layer, indicating significant abnormal temperature fluctuations during this stage, triggering the generation of temperature difference offset records.

[0049] Example 4:

[0050] In specific implementation, please refer to Figure 2 The process of determining the location of local hot zones within the palletizing layer based on the temperature difference offset record, defining a group of boxes to be traced with the location of the local hot zones as the center, and extracting the entry timestamp breakpoint of each box from the group of boxes to be traced is achieved in the following way:

[0051] Extract the temperature difference time series array corresponding to each chamber from the intra-layer temperature difference offset record. The temperature difference time series array contains the absolute value of the temperature difference at several sampling times, and the time interval between the sampling times is a fixed sampling interval, denoted as . , The value is 30 seconds. The duration of the continuous time window is adaptively set based on the category code of the cold chain fresh produce. The adaptive setting method is as follows: a mapping table between the category code and the duration of the continuous time window is pre-built. The mapping table stores the continuous time window duration values ​​corresponding to different category codes. The category code of the cold chain fresh produce is written into the storage area of ​​the RFID tag by the cold chain management system when the container enters the warehouse. The category code uses a four-digit code, with each digit representing a different category attribute level. In the mapping table, the correspondence between the first two digits of the category code and the duration of the continuous time window is as follows: the first two digits range from 01 to 99. When the first two digits of the category code are in the range of 01 to 20, the continuous time window duration is 600 seconds; when they are in the range of 21 to 40, the duration is 900 seconds; when they are in the range of 41 to 60, the duration is 1200 seconds; when they are in the range of 61 to 80, the duration is 1500 seconds; and when they are in the range of 81 to 99, the duration is 1800 seconds. The above values ​​are based on the fact that the respiratory heat release rate varies significantly among different category codes for fresh produce. Categories with higher respiratory heat release rates require shorter continuous time window durations to more sensitively capture the cumulative temperature effect. When calculating the cumulative temperature difference, the duration of the continuous time window is obtained from the mapping table based on the category code corresponding to the current enclosure, and is denoted as... .

[0052] For each chamber's temperature difference time series array, the temperature difference time series array is divided into continuous time window durations. The window length is used to determine the fixed sampling interval. To define a sliding step size, multiple continuous time windows are extracted. For each continuous time window, the cumulative temperature difference is calculated by summing the products of the absolute values ​​of the temperature differences at each sampling point within the continuous time window and the fixed sampling interval. The cumulative temperature difference is denoted as... The calculation formula is:

[0053]

[0054] in, This represents the total number of sampling points within a continuous time window. ; Indicates the sequence number of the sampling point within the continuous time window. The value range is 1 to Integers; Indicates the number of consecutive time windows. The absolute value of the temperature difference at each sampling point is extracted from the temperature difference time series array in the order of sampling time. For fixed sampling intervals.

[0055] The cumulative temperature difference values ​​of each chamber are compared across all continuous time windows. The maximum cumulative temperature difference value of each chamber is selected. Then, the maximum cumulative temperature difference values ​​of all chambers are compared. The chamber with the largest cumulative temperature difference value is identified as the heat source center chamber, and the area where the heat source center chamber is located is marked as the local heat zone location.

[0056] It can be understood that, taking the box corresponding to the local hot zone as the center, all boxes on the pallet that are no more than two box spacings apart in both the horizontal and vertical directions from the central box are included in the box group to be traced. Box spacing refers to the distance between the geometric centers of two adjacent boxes in the horizontal or vertical direction, which is obtained from the box arrangement pose image by calculating the distance between the corner coordinates of adjacent boxes. Boxes that are one box spacing apart in the horizontal direction from the central box are horizontally adjacent boxes, and boxes that are two box spacings apart in the horizontal direction from the central box are horizontally spaced boxes, and the same applies in the vertical direction. The central box itself, horizontally adjacent boxes, horizontally spaced boxes, vertically adjacent boxes, vertically spaced boxes, and boxes in the diagonal direction that satisfy the condition that the horizontal and vertical spacing does not exceed two box spacings are all included in the box group to be traced.

[0057] Optionally, the RFID storage area of ​​each box in the group of boxes to be traced is read to record the entry time field, which contains the timestamp of the box passing through the entry scanning point. The entry time fields of each box in the group are sorted chronologically, and the time interval between the entry time fields of adjacent boxes is calculated. The arithmetic mean of all adjacent time intervals is obtained to get the average entry time interval between adjacent boxes. The time intervals of the entry time fields of all adjacent boxes are iterated. When a time interval is greater than twice the average entry time interval between adjacent boxes, the entry time field of the next box corresponding to that time interval is extracted as an entry timestamp breakpoint. An entry timestamp breakpoint indicates a significant interruption or delay in the box entry process before or after that time point.

[0058] Example 5:

[0059] In practice, the process of binding the inbound timestamp breakpoint with the batch identification code corresponding to the pallet to form a box-level traceability anchor point, and writing the box-level traceability anchor point into the inbound traceability chain list, is achieved through the following method:

[0060] An RFID pallet tag is installed on the edge of the pallet. The internal storage area of ​​the RFID pallet tag contains a pre-set batch identification code. This batch identification code is written by the cold chain management system before the pallet is used. The encoding format of the batch identification code is a combination of the pallet number and the batch number. An RFID reader sends a read command to the RFID pallet tag to retrieve the batch identification code from its storage area.

[0061] A mapping relationship is established between the read batch identification code and the entry timestamp breakpoint of each box, based on the box number. The box number is a unique identifier for each box in the palletizing layer, automatically generated according to the box's row and column position on the pallet plane when the boxes are stacked. The mapping relationship is organized in key-value pair format, where the key is the box number and the value is a structure composed of the batch identification code and the entry timestamp breakpoint. Each key-value pair is recorded as a traceability anchor record, with the format: {box number:{batch identification code, entry timestamp breakpoint}}. When multiple entry timestamp breakpoints exist for the same box, the earliest entry timestamp breakpoint is taken as the entry timestamp breakpoint field value in the traceability anchor record.

[0062] Optionally, container-level traceability anchors can be constructed using the container number in each traceability anchor record as the index node and the batch identification code and the entry timestamp breakpoint as node attributes. The data structure of the index node includes a node identifier field, a batch identification code field, an entry timestamp breakpoint field, and a pointer field. The node identifier field stores the container number, the batch identification code field stores the batch identification code read from the RFID pallet tag, the entry timestamp breakpoint field stores the time value of the entry timestamp breakpoint corresponding to the container, and the pointer field is used to point to the next node in the entry traceability chain.

[0063] It can be understood that the inbound traceability list is a chained data structure. The inbound traceability list contains a head node, intermediate nodes, and a tail node. The head node does not store traceability data; its pointer field points to the first intermediate node that stores traceability data. The container-level traceability anchor points are inserted after the tail node of the inbound traceability list according to the order of their inbound timestamp breakpoints. The specific insertion process is as follows: Create a new node, and write the container number, batch identification code, and inbound timestamp breakpoint contained in the container-level traceability anchor point into the node identifier field, batch identification code field, and inbound timestamp breakpoint field of the new node, respectively. Compare the inbound timestamp breakpoint field value of the node to be inserted with the inbound timestamp breakpoint field values ​​of existing nodes in the inbound traceability list, determining the insertion position according to time order. When the inbound timestamp breakpoint time of the node to be inserted is later than any existing node in the inbound traceability list, link the new node after the current tail node of the inbound traceability list, update the pointer field of the new node to null, and set the pointer field of the original tail node to point to the new node. The new node then becomes the new tail node.

[0064] In practice, the inbound traceability list is stored as a blockchain ledger on the local edge computing node of the pallet carrier. The local edge computing node, deployed on the pallet carrier, includes a processor, memory, and communication module. The blockchain ledger adopts a blockchain structure, with each block containing a block header and a block body. The block header contains the hash value of the previous block, the timestamp of the current block, and the Merkle root. The block body stores multiple container-level traceability anchors. When a container-level traceability anchor is written to the inbound traceability list, the local edge computing node obtains the timestamp of the current write operation as the write timestamp. The local edge computing node uses its private key to digitally sign the written content, generating a digital signature for the writing node. The written content includes the container number, batch identification code, inbound timestamp breakpoint, and write timestamp. The write timestamp and the writing node's digital signature are attached to the container-level traceability anchor record and written into the block body of the inbound traceability list. Each container-level traceability anchor record in the block contains a container number field, a batch identification code field, an inbound timestamp breakpoint field, a write timestamp field, and a digital signature field for the write node.

[0065] See Figure 5 In the graph, the horizontal axis represents the sequence number of adjacent container pairs, and the vertical axis represents the time interval between the entry of adjacent containers, in seconds. The curves use solid black lines to represent the trend of the time interval between entry of each pair of adjacent containers, dashed blue lines to represent the average time interval between entry of all adjacent containers, and dashed red lines to represent twice the threshold of this average. The red squares in the graph mark the entry timestamp breakpoints, indicating key nodes where the time interval between entry of adjacent containers increases abnormally.

[0066] As can be seen from the curve trend, the time intervals between the entry of most adjacent containers are distributed below the blue dashed line of the average entry time interval, and most values ​​do not exceed the threshold of twice the average entry time interval marked by the red dashed line, indicating that the entry status of the containers is normal and continuous within this time range. Several black solid line peaks in the figure clearly exceed the threshold of the red dashed line, and the corresponding positions are marked with red squares, which conforms to the rule in Example 4 regarding the determination of the entry time stamp breakpoint, that is, when the time interval between the entry of adjacent containers is greater than twice the average entry time interval, the entry time field of the container after that time interval is extracted as the entry time stamp breakpoint.

[0067] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce, characterized in that: The method includes: Obtain the arrangement and pose image of the boxes for cold chain fresh products in the palletizing layer, and analyze the edge contour of the boxes in the arrangement and pose image; The corner coordinates of each box are extracted based on the edge contour of the box, and the overlap of the orthographic projection of each box on the pallet plane is calculated based on the corner coordinates. When the amount of orthographic projection overlap exceeds the preset attitude threshold, the bottom surface temperature measurement tag sequence of the corresponding box is retrieved, and the bottom surface temperature measurement tag sequence is paired with the side wall temperature measurement tag sequence of the adjacent box in the same layer to generate an intra-layer temperature difference offset record. Based on the temperature difference offset record within the layer, the location of the local hot zone within the palletizing layer is calibrated. The group of boxes to be traced is delineated with the location of the local hot zone as the center, and the entry timestamp breakpoint of each box is extracted from the group of boxes to be traced. The inbound timestamp breakpoint is bound to the batch identification code corresponding to the pallet to form a box-level traceability anchor point, and the box-level traceability anchor point is written into the inbound traceability chain list.

2. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 1, characterized in that, The specific steps of obtaining the arrangement pose image of the boxes for cold chain fresh produce in the palletizing layer and parsing the edge contours of the boxes in the arrangement pose image are as follows: A full-frame image of the pallet is acquired in a manner perpendicular to the top of the pallet stacking layer. After grayscale processing of the full-frame image of the pallet, a gradient magnitude response map is extracted. Pixel connected regions with gradient magnitudes greater than the background noise threshold are selected from the gradient magnitude response map, and the bounding rectangle of the pixel connected regions is used as a candidate box region. Line segment fitting is performed on the candidate box area, and line segments with an angle less than a set angle deviation with the pallet edge direction are extracted as the box edge contour.

3. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 2, characterized in that, The specific steps for extracting the corner coordinates of each box based on the edge contour of the box, and calculating the overlap of the orthographic projection of each box on the pallet plane based on the corner coordinates are as follows: Extract the intersection of two adjacent edge lines from the edge contour of the box as corner candidate points, and collect the corner candidate points into a corner coordinate set according to the box number; The coordinates of the four corner points of the target box are read from the set of corner point coordinates, and the area of ​​the quadrilateral formed by the coordinates of the four corner points is the orthographic projection area of ​​the box. The intersection area of ​​the target box and the adjacent box on the same layer is calculated in the pallet plane coordinate system to obtain the orthogonal projection overlap.

4. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 3, characterized in that, When the orthographic overlap exceeds a preset attitude threshold, the specific steps for retrieving the bottom surface temperature measurement tag sequence of the corresponding box and pairing the bottom surface temperature measurement tag sequence with the side wall temperature measurement tag sequence of the adjacent box in the same layer to generate an intra-layer temperature difference offset record are as follows: When the overlap of the orthographic projection of the target box with any adjacent box on the same layer is greater than the preset attitude threshold, the first RFID temperature measurement tag on the bottom surface of the target box is activated, and the bottom surface temperature value is read at a fixed sampling interval to form the bottom surface temperature measurement tag sequence. Simultaneously activate the second RFID temperature measurement tag on the side wall of the adjacent box, and read the side wall temperature value at the same fixed sampling interval to form the side wall temperature measurement tag sequence; The temperature difference time series array is obtained by subtracting the bottom surface temperature value and the side wall temperature value at the same sampling time. The array segments in the temperature difference time series array that exceed the temperature difference fluctuation range are marked as the temperature difference offset records within the layer.

5. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 4, characterized in that, The steps of determining the location of local hot zones within the palletizing layer based on the intra-layer temperature difference offset record, defining a group of boxes to be traced with the location of the local hot zones as the center, and extracting the entry timestamp breakpoint of each box from the group of boxes to be traced are as follows: Extract the temperature difference time series array corresponding to each box from the temperature difference offset record in the layer, calculate the cumulative temperature difference value of each temperature difference time series array in the continuous time window, and mark the area where the box with the largest cumulative temperature difference value is located as the local hot zone location. Taking the box corresponding to the location of the local hot zone as the center, all boxes on the pallet that are no more than two box spacings apart from the central box in both the horizontal and vertical directions are included in the box group to be traced. Read the entry time field of the RFID storage area of ​​each box in the group of boxes to be traced, and extract the time points in the entry time field whose time interval is greater than twice the average entry time interval of adjacent boxes as the entry timestamp breakpoints.

6. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 5, characterized in that, The specific steps for binding the inbound timestamp breakpoint with the batch identification code corresponding to the pallet to form a box-level traceability anchor point, and writing the box-level traceability anchor point into the inbound traceability chain list are as follows: Read the batch identification code from the RFID pallet tag installed on the edge of the pallet, establish a mapping relationship between the batch identification code and the entry timestamp of each box according to the box number, and generate a traceability anchor record in the form of key-value pairs; Using the container number in each traceability anchor record as the index node, and the batch identification code and the warehouse entry timestamp breakpoint as the node attributes, construct container-level traceability anchors; The container-level traceability anchor points are inserted after the last node of the inbound traceability chain in the order of the inbound timestamp breakpoints.

7. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 2, characterized in that, The gradient magnitude response map is obtained by calculating the gradient components in the horizontal and vertical directions using the Sobel operator and then taking the modulus. The background noise threshold is the sum of the mean and standard deviation of the gradient magnitudes of all pixels in the full-frame image of the tray.

8. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 4, characterized in that, The first RFID temperature measurement tag is attached to the geometric center of the bottom surface of the box, and the second RFID temperature measurement tag is attached to the midline position of the side wall of the box in the height direction. The reading timestamps of the bottom surface temperature value and the side wall temperature value are synchronized and calibrated by the local clock of the tray.

9. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 5, characterized in that, The cumulative temperature difference value is obtained by summing the product of the absolute value of the temperature difference at each sampling point within a continuous time window and the sampling interval. The duration of the continuous time window is adaptively set according to the category code of the cold chain fresh food category.

10. The method for full-process traceability of packing, palletizing, and warehousing of cold chain fresh produce according to claim 6, characterized in that, The inbound traceability chain is stored in the local edge computing node of the pallet carrier in the form of a blockchain ledger, and each box-level traceability anchor is accompanied by a writing timestamp and the writing node's digital signature when it is written.

Images