A non-standard sample container anomaly identification method and system

Abstract

The application provides a non-standard sample container anomaly identification method and system, the method comprises: by generating a container state sequence, superimposing multiple frames of the same digital identity container continuous image, generating a motion trajectory geometric envelope, extracting geometric feature parameters from the geometric envelope, calculating the speed and acceleration of each container based on the container state sequence, obtaining the expected dynamic sequence determined by the geometry according to the geometric feature parameters and the current speed and acceleration, obtaining the residual abnormal feature sequence through the difference between the actual dynamic sequence and the expected dynamic sequence, calculating the cross-correlation coefficient between the expected dynamic sequence and the actual dynamic sequence, obtaining the geometric envelope offset of the container through the residual abnormal feature sequence, calculating the comprehensive abnormal score and determining the abnormal level based on the cross-correlation coefficient, the damping offset and the label recognition confidence, for the container geometric form diversity problem in the sample transfer between departments or hospitals, reducing the misjudgment rate of cross-institution sample identification, and ensuring the stability of detection.

Description

Technical Field

[0001] This application relates to the field of anomaly identification technology for non-standard sample containers, specifically a method and system for anomaly identification of non-standard sample containers. Background Technology

[0002] Non-standard sample containers are containers that do not meet the established specifications or expected requirements of the laboratory or medical testing environment. Non-standard sample containers vary in form or properties.

[0003] In real-world medical scenarios, there are situations where samples are transported across departments or hospitals. However, in these scenarios, the sample containers used by different medical institutions vary significantly in size and shape. A uniform static reference standard is difficult to adapt to various non-standard containers, which can easily lead to unreasonable anomaly detection thresholds and a high false positive rate. How to achieve adaptive recognition and anomaly detection for sample containers with different geometric shapes while ensuring detection stability has become a problem that current technologies urgently need to solve. Summary of the Invention

[0004] To overcome the aforementioned problems in the prior art, this application provides a method and system, which adopts the following technical solution:

[0005] Firstly, this application provides a method for identifying anomalies in non-standard sample containers, including:

[0006] The system collects visual and displacement data of the target container in real time during the transportation process, aligns the visual and displacement data in time, and generates a predicted position. In the current frame, a matching cost function is constructed based on preset constraints, and the predicted position is associated with the historical trajectory through the optimal matching algorithm to generate a container state sequence.

[0007] Multiple frames of continuous images of the same digital identity container are superimposed to generate a geometric envelope of the motion trajectory, and geometric feature parameters are extracted from the geometric envelope.

[0008] The velocity and acceleration of each container are calculated based on the container state sequence. The expected dynamic sequence determined by geometry is obtained based on the geometric feature parameters and the current velocity and acceleration. The remaining abnormal feature sequence that is independent of geometry is obtained by the difference between the measured dynamic sequence and the expected dynamic sequence.

[0009] Calculate the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence;

[0010] The geometric envelope offset of the container is obtained by using the remaining abnormal feature sequence. The damping offset is calculated based on the cross-correlation coefficient and the label recognition confidence is obtained. Then, the comprehensive abnormal score is calculated and the abnormality level is determined.

[0011] Furthermore, the visual data and displacement data are time-aligned to generate predicted locations, including:

[0012] Visual data and displacement data are mapped to the same reference time for time alignment. Based on the container state vector confirmed in the previous time step and the aligned displacement data, the state transition equation of the container is constructed. The predicted state vector is obtained based on the state transition equation, and the predicted state vector contains the predicted position.

[0013] Furthermore, the preset constraints include the image spatial distance between the predicted position and the measured centroid, and the lateral offset of the measured centroid relative to the centerline of the transport track.

[0014] Furthermore, in the current frame, a matching cost function is constructed based on preset constraints, namely: ,in , Indicates the first Predicted location of a historical trajectory, Indicates the first The centroid detected in the current frame, Indicates the coordinates of the centerline of the transport track. Indicates the offset weight. This represents the matching cost function.

[0015] Furthermore, multiple frames of consecutive images of the same digital identity container are superimposed to generate a geometric envelope of the motion trajectory. Geometric feature parameters are extracted from the geometric envelope, including:

[0016] For the same digital identity container, the centroid coordinates of each frame are extracted from N consecutive frames of images. A two-dimensional motion trajectory region is generated by superimposing the centroid trajectory. The convex hull is calculated as the geometric envelope of the container motion through the trajectory region. Geometric feature parameters are calculated based on the geometric envelope.

[0017] Furthermore, based on the geometric characteristic parameters and the current velocity and acceleration, the expected dynamic sequence of geometric determinations is obtained, including:

[0018] Based on the container's equivalent moment of inertia and dimensions, the equivalent mass of the container is determined according to a preset geometric equivalence relationship. The equivalent stiffness and equivalent damping are obtained according to a preset structural proportion relationship. These equivalent mass, equivalent stiffness, and equivalent damping are used as parameters of a preset dynamic model. Velocity and acceleration are used as external excitation inputs. By solving the preset dynamic model in the time domain, the theoretical displacement sequence of the container under the corresponding motion state is obtained. and theoretical velocity sequence The expected dynamic sequence is then... .

[0019] Furthermore, by using the difference between the measured dynamic sequence and the expected dynamic sequence, a geometrically independent residual anomaly feature sequence is obtained, namely: ,in These are residual abnormal features. This is a measured dynamic sequence. For the expected dynamic sequence, This represents the measured displacement of the container at time t. This represents the measured velocity of the container at time t.

[0020] Secondly, this application also provides a non-standard sample container anomaly identification system, including:

[0021] The container state sequence generation module is used to collect visual and displacement data of the target container in real time during the transportation process, align the visual and displacement data in time, and generate the predicted position. In the current frame, a matching cost function is constructed based on preset constraints, and the predicted position is associated with the historical trajectory through the optimal matching algorithm to generate the container state sequence.

[0022] The geometric feature parameter extraction module is used to overlay multiple frames of continuous images of the same digital identity container to generate a geometric envelope of motion trajectory, and extract geometric feature parameters from the geometric envelope.

[0023] The residual anomaly feature acquisition module is used to calculate the velocity and acceleration of each container based on the container state sequence, obtain the expected dynamic sequence determined by geometry based on the geometric feature parameters and the current velocity and acceleration, and obtain the residual anomaly feature sequence that is independent of geometry by the difference between the measured dynamic sequence and the expected dynamic sequence.

[0024] The cross-correlation coefficient acquisition module is used to calculate the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence.

[0025] The anomaly detection module is used to obtain the geometric envelope offset of the container through the remaining anomaly feature sequence, calculate the damping offset based on the cross-correlation coefficient and obtain the label recognition confidence, calculate the comprehensive anomaly score and determine the anomaly level.

[0026] Thirdly, this application provides an electronic device, comprising:

[0027] One or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the device, cause the device to perform the method as described in the first aspect.

[0028] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the method described in the first aspect.

[0029] Fifthly, this application provides a computer program that, when executed by a computer, performs the method described in the first aspect.

[0030] In one possible design, the program in the fifth aspect can be stored wholly or partially on a storage medium packaged with the processor, or it can be stored wholly or partially on a memory not packaged with the processor.

[0031] This application has the following beneficial effects:

[0032] 1. This application collects visual and displacement data of the target container in real time during the transportation process, aligns the visual and displacement data in time, and generates a predicted position. In the current frame, a matching cost function is constructed based on preset constraints, and the predicted position is associated with the historical trajectory through the optimal matching algorithm to generate a container state sequence. The container state sequence in this application can bind the physical spatial position of the container with a unique digital identity, and maintain the continuity of identity even under short-term occlusion or track interference, providing a reliable basis for the stable identification and dynamic anomaly judgment of non-standard containers.

[0033] 2. This application generates a geometric envelope of motion trajectory by overlaying multiple frames of continuous images of the same digital identity container, and extracts geometric feature parameters from the geometric envelope. Based on the container state sequence, the velocity and acceleration of each container are calculated. The expected dynamic sequence determined by geometry is obtained based on the geometric feature parameters and the current velocity and acceleration. By comparing the measured dynamic sequence with the expected dynamic sequence, a residual anomaly feature sequence independent of geometry is obtained. The residual anomaly feature sequence of this application can eliminate the influence of container shape differences on vibration response, retaining only the true dynamic anomaly information. Even when facing non-standard containers with different sizes, eccentricities, or odd shapes, it can accurately extract anomaly features, providing stable input for anomaly judgment and classification, and reducing the risk of misjudgment.

[0034] 3. This application calculates the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence; obtains the geometric envelope offset of the container through the remaining abnormal feature sequence; calculates the damping offset based on the cross-correlation coefficient and obtains the label recognition confidence level; and then calculates a comprehensive anomaly score and determines the anomaly level. This application calculates a comprehensive anomaly score using geometric envelope offset, damping offset, and label recognition confidence level, while simultaneously quantifying container morphology, dynamic response, and visual recognition information to achieve multi-dimensional fusion judgment. By mapping the score to anomaly level, it accurately distinguishes between normal and non-standard abnormal containers, improving the reliability of non-standard abnormal sample container identification. For the problem of diverse container geometric shapes during cross-departmental or cross-hospital sample transportation, it reduces the misjudgment rate of cross-institutional sample identification and ensures the stability of the detection. Attached Figure Description

[0035] Figure 1 This is a flowchart of the non-standard sample container anomaly identification method according to an embodiment of this application;

[0036] Figure 2 This is a flowchart illustrating the re-authentication process for digital identity in an embodiment of this application.

[0037] Figure 3 This is a flowchart illustrating the geometric feature parameter extraction process according to an embodiment of this application.

[0038] Figure 4 This is a flowchart illustrating the expected dynamic sequence acquisition process in an embodiment of this application.

[0039] Figure 5 This is a system flowchart of an embodiment of this application. Detailed Implementation

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0041] 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 this application. 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.

[0042] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0043] Please refer to Figure 1 The following is a flowchart of a non-standard sample container anomaly identification method provided in this application embodiment. The specific implementation process is as follows:

[0044] Step 101: Real-time acquisition of visual and displacement data of the target container during the transportation process; time alignment of visual and displacement data to generate predicted position; in the current frame, construction of a matching cost function based on preset constraints; association of predicted position with historical trajectory through optimal matching algorithm to generate container state sequence.

[0045] In this embodiment of the application, visual data is obtained by a continuous image sequence from a high frame rate camera, which is used to extract the centroid coordinates of the container; displacement data is obtained by a real-time displacement amount from a conveyor belt encoder, which is used to characterize the physical movement distance of the container.

[0046] In this embodiment of the application, visual data and displacement data are time-aligned to generate a predicted position, including:

[0047] Visual data and displacement data are mapped to the same reference time for time alignment, based on the container state vector confirmed at the previous time step. Based on the aligned displacement data, construct the state transition equation for the container: ,in Estimated by the difference between the current displacement increment and the previous velocity, where Indicates the velocity at the previous moment. Indicates the sampling time. This represents the output predicted state vector, where The predicted location is .

[0048] In this embodiment of the application, the preset constraints include the image spatial distance between the predicted position and the measured centroid, and the lateral offset of the measured centroid relative to the center line of the transport track.

[0049] In this embodiment of the application, in the current frame, a matching cost function is constructed based on preset constraints, namely:

[0050]

[0051] in , Indicates the first Predicted location of a historical trajectory, Indicates the first The centroid detected in the current frame, Indicates the coordinates of the centerline of the transport track. Indicates the offset weight. This represents the matching cost function.

[0052] This represents the Euclidean distance between the predicted and measured centroids. This indicates the lateral offset of the measured centroid relative to the centerline of the transport track.

[0053] The predicted location is associated with the historical trajectory using an optimal matching algorithm, generating a container state sequence, including: all Composition of cost matrix The optimal matching is solved using the Hungarian algorithm. To minimize the total cost: Once a match is successful, the historical trajectory will be recorded. Digital identity assigned to detected target .

[0054] Candidate target locations are generated based on predicted locations. Within the candidate target location area, when no target satisfying the spatial consistency constraint with the predicted state is detected, it is determined that the container is occluded. The container's state vector is updated using displacement data and state transition equations. When the target reappears, the consistency of the newly detected target state with the state vector obtained during the occlusion period is verified, completing the re-authentication of the digital identity. The container state sequence corresponding to the digital identity is then updated. ,in .in Representing the digital identity of a container, This represents the state vector of the container, describing the container's motion state at the current moment.

[0055] In this embodiment, the newly detected target state is compared with the state vector obtained during the occlusion period to verify consistency, thus completing the re-authentication of the digital identity. Please refer to [link / reference]. Figure 2 ,include:

[0056] Step 21: When the occlusion ends and the container visual features are detected again, a consistency determination is made based on the residual and prediction covariance between the predicted state vector and the measured state.

[0057] Step 22: When the judgment result meets the preset constraints, confirm that the visual target is consistent with the container state before occlusion, and then retain the corresponding digital identity.

[0058] In this embodiment, when the occlusion ends and the container's visual features are detected again, a consistency determination is made based on the residual and prediction covariance between the predicted state vector and the measured state, i.e., the residual vector between the measured centroid and the predicted position is calculated: Using the predicted covariance matrix Calculate Mahalanobis distance ,when At that time, it is determined that the visual target is consistent with the state of the container before occlusion, where This indicates the preset information level threshold.

[0059] In the embodiments of this application, the predicted covariance matrix is... It is based on the covariance of the previous time step. and process noise The calculation results are as follows: ,in It is the Jacobian matrix of the state transition equation. It is the process noise covariance matrix.

[0060] It should be noted that time alignment can be achieved by mapping visual data and displacement data to the same reference time through linear interpolation or nearest neighbor interpolation.

[0061] Step 102: Superimpose multiple frames of continuous images of the same digital identity container to generate a geometric envelope of motion trajectory, and extract geometric feature parameters from the geometric envelope.

[0062] For details, please refer to Figure 3 ,include:

[0063] Step 31: For the same digital identity container, extract the centroid coordinates of each frame from N consecutive frames of images, and generate a two-dimensional motion trajectory region by superimposing the centroid trajectory.

[0064] It should be noted that when extracting the centroid coordinates of each frame from N consecutive frames of images for the same digital identity container, and generating a two-dimensional motion trajectory region by superimposing the centroid trajectory, the superimposed image is binarized to obtain the container pixel distribution, and the geometric envelope of the container motion trajectory is generated by the convex hull algorithm.

[0065] Step 32: Calculate the geometric envelope of the convex hull as the container's motion using the trajectory region. Then, calculate the geometric characteristic parameters based on the geometric envelope, including the aspect ratio, eccentricity, and equivalent moment of inertia. The aspect ratio... eccentricity Equivalent moment of inertia The aspect ratio is the length of the longest side of the smallest bounding rectangle of the convex hull. With the shorter side The ratio of to describes the directional extension of the container's trajectory in a two-dimensional plane. Eccentricity quantifies the degree of elliptical stretching of the trajectory, reflecting the asymmetry of the motion. Under normal, stable motion, the eccentricity of a container's trajectory remains within a certain range; deviations from this range may indicate abnormal container shape or motion. Here, represents the length of the ellipse's major axis. The length of the minor axis is . For the first convex hull The distance from the point to the centroid. For a unit mass, the equivalent moment of inertia is used to extract geometry-independent anomalies. This represents the total number of convex hull points.

[0066] Step 103: Calculate the velocity and acceleration of each container based on the container state sequence, obtain the expected dynamic sequence determined by geometry based on the geometric feature parameters and the current velocity and acceleration, and obtain the remaining abnormal feature sequence that is independent of geometry by the difference between the measured dynamic sequence and the expected dynamic sequence.

[0067] In this embodiment of the application, calculating the velocity and acceleration of each container based on the container state sequence includes: calculating the velocity based on the centroid coordinate difference of consecutive frames in the container state sequence, i.e.

[0068] When the sampling interval is When the velocity is approximately equal to: The speed is Acceleration is obtained by differentially analyzing the velocity, i.e.: ,but , Then the acceleration is Velocity and acceleration are used as parameters of the container's motion state.

[0069] In this embodiment, the expected dynamic sequence of geometric determination is obtained based on geometric feature parameters and the current velocity and acceleration. Please refer to [reference needed]. Figure 4 ,include:

[0070] Step 41: Based on the equivalent moment of inertia and dimensions of the container, determine the equivalent mass of the container according to the preset geometric equivalence relationship.

[0071] Step 42: Obtain the equivalent stiffness and equivalent damping according to the preset structural proportions, and use the equivalent mass, equivalent stiffness, and equivalent damping as parameters of the preset dynamic model.

[0072] Step 43: Using velocity and acceleration as external excitation inputs, the theoretical displacement sequence and theoretical velocity sequence of the container in the corresponding motion state are obtained by solving the preset dynamic model in the time domain. The expected dynamic sequence is obtained through the theoretical displacement sequence and theoretical velocity sequence.

[0073] It should be noted that the equivalent quality is ,in For equivalent rotational inertia, For feature size, These are equivalent coefficients related to the container's geometry, used to characterize the mass distribution assumption. The equivalent stiffness and equivalent damping, obtained according to a pre-defined structural proportion, can be expressed as: , ,in For equivalent stiffness, For equivalent damping, , These are proportionality coefficients related to the structure type. The preset dynamic model is... ,in This is the second derivative of displacement with respect to time. This is the first derivative of displacement with respect to time. This is the equivalent displacement.

[0074] External stimulus input is ,in This represents the instantaneous acceleration of the container during transport. Based on the current state, the initial displacement and initial velocity are set as follows: , Then, by solving the preset dynamic model time-by-time, we obtain... The displacement values ​​obtained from continuous solutions are arranged according to a unified time base to form a theoretical displacement time series. The theoretical speed is The expected dynamic sequence can then be represented as: .

[0075] In this embodiment of the application, a geometrically independent residual anomaly feature sequence is obtained by the difference between the measured dynamic sequence and the expected dynamic sequence, i.e.: ,in These are residual abnormal features. This is a measured dynamic sequence. For the expected dynamic sequence, This represents the measured displacement of the container at time t. This represents the measured velocity of the container at time t.

[0076] Step 104: Calculate the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence.

[0077] In this embodiment of the application, the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence is calculated, i.e. ,in Describing covariance, and These represent the standard deviation and cross-correlation coefficient of the corresponding sequences, respectively. Used to determine the consistency between the measured dynamic sequence and the expected dynamic sequence. This is a measured dynamic sequence. This is the expected dynamic sequence.

[0078] Step 105: Obtain the geometric envelope offset of the container through the remaining abnormal feature sequence, calculate the damping offset based on the cross-correlation coefficient and obtain the label recognition confidence, calculate the comprehensive abnormal score and determine the abnormal level, and generate the corresponding control command according to the abnormal level.

[0079] In this embodiment, obtaining the geometric envelope offset of a container through the remaining abnormal feature sequence includes: performing frequency domain analysis on the remaining abnormal feature sequences of the same digital identity, converting them into corresponding geometric offsets through a preset mapping function, and obtaining the geometric envelope offset of the container in the time domain through inverse transformation processing, i.e. .in This is the geometric envelope offset. This indicates the operation of inverse transformation to the time domain. The remaining abnormal feature sequence, This is a mapping function from anomalous features to geometric offset, which can be obtained through amplitude or frequency domain analysis.

[0080] In this embodiment of the application, calculating the damping offset based on the cross-correlation coefficient includes: comparing the cross-correlation coefficient with a preset threshold to obtain the degree of deviation, and obtaining the corresponding damping offset based on the degree of deviation, i.e.: ,in The damping offset mapping function can be used to map the cross-correlation function into the damping change based on the calibration curve or empirical formula. This represents the damping offset.

[0081] In this embodiment, obtaining the label recognition confidence score includes: processing container image data with the same digital identity using a pre-trained image recognition model, and outputting the predicted probability vector of the label to which the container belongs. The probability value corresponding to the identification label is used as the confidence level. Then the confidence level .

[0082] In this embodiment of the application, a comprehensive anomaly score is calculated and an anomaly level is determined, wherein the anomaly score is: ,in , , These are the weighting coefficients.

[0083] The abnormality level is determined as follows: when the abnormality score is less than or equal to the first preset threshold, it is considered normal; when the abnormality score is greater than the first preset threshold but less than or equal to the second preset threshold, it is considered level two abnormal; and when the abnormality score is greater than the second preset threshold, it is considered level one abnormal.

[0084] In the embodiments of this application, please continue to refer to Figure 5 An anomaly identification system for non-standard sample containers provided in this application embodiment includes:

[0085] The container state sequence generation module 501 is used to collect visual data and displacement data of the target container in real time during the transportation process, align the visual data and displacement data in time, and generate the predicted position. In the current frame, a matching cost function is constructed based on preset constraints, and the predicted position is associated with the historical trajectory through the optimal matching algorithm to generate the container state sequence.

[0086] The geometric feature parameter extraction module 502 is used to overlay multiple frames of continuous images of the same digital identity container to generate a motion trajectory geometric envelope, and extract geometric feature parameters from the geometric envelope.

[0087] The residual anomaly feature acquisition module 503 is used to calculate the velocity and acceleration of each container based on the container state sequence, obtain the expected dynamic sequence determined by geometry based on the geometric feature parameters and the current velocity and acceleration, and obtain the residual anomaly feature sequence that is independent of geometry by the difference between the measured dynamic sequence and the expected dynamic sequence.

[0088] The cross-correlation coefficient acquisition module 504 is used to calculate the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence.

[0089] The anomaly detection module 505 is used to obtain the geometric envelope offset of the container through the remaining anomaly feature sequence, calculate the damping offset based on the cross-correlation coefficient and obtain the label recognition confidence, calculate the comprehensive anomaly score and determine the anomaly level.

[0090] This application generates a container state sequence, overlays multiple frames of continuous images of the same digital identity container to generate a motion trajectory geometric envelope, extracts geometric feature parameters from the geometric envelope, calculates the velocity and acceleration of each container based on the container state sequence, obtains the expected dynamic sequence determined by geometry based on the geometric feature parameters and the current velocity and acceleration, obtains the remaining abnormal feature sequence by the difference between the measured dynamic sequence and the expected dynamic sequence, calculates the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence, obtains the geometric envelope offset of the container through the remaining abnormal feature sequence, calculates the damping offset and obtains the label recognition confidence based on the cross-correlation coefficient, calculates the comprehensive abnormality score and determines the abnormality level, reduces the cross-institution sample recognition misjudgment rate and ensures the stability of detection for the problem of container geometric morphology diversity in cross-department or cross-hospital sample transportation.

[0091] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A method for identifying anomalies in non-standard sample containers, characterized in that, include: Real-time acquisition of visual and displacement data of the target container during the transportation process; time alignment of visual and displacement data to generate predicted position; In the current frame, a matching cost function is constructed based on preset constraints, and the predicted position is associated with the historical trajectory through the optimal matching algorithm to generate a container state sequence; Multiple frames of continuous images of the same digital identity container are superimposed to generate a geometric envelope of the motion trajectory, and geometric feature parameters are extracted from the geometric envelope. The velocity and acceleration of each container are calculated based on the container state sequence. The expected dynamic sequence determined by geometry is obtained based on the geometric feature parameters and the current velocity and acceleration. The remaining abnormal feature sequence that is independent of geometry is obtained by the difference between the measured dynamic sequence and the expected dynamic sequence. Calculate the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence; The geometric envelope offset of the container is obtained by using the remaining abnormal feature sequence. The damping offset is calculated based on the cross-correlation coefficient and the label recognition confidence is obtained. Then, the comprehensive abnormal score is calculated and the abnormality level is determined.

2. The method for identifying anomalies in non-standard sample containers according to claim 1, characterized in that, The visual data and displacement data are time-aligned to generate predicted locations, including: Visual data and displacement data are mapped to the same reference time for time alignment. Based on the container state vector confirmed in the previous time step and the aligned displacement data, the state transition equation of the container is constructed. The predicted state vector is obtained based on the state transition equation, and the predicted state vector contains the predicted position.

3. The method for identifying anomalies in non-standard sample containers according to claim 1, characterized in that, The preset constraints include the image spatial distance between the predicted position and the measured centroid, and the lateral offset of the measured centroid relative to the centerline of the transport track.

4. The method for identifying anomalies in non-standard sample containers according to claim 3, characterized in that, In the current frame, a matching cost function is constructed based on preset constraints, namely: ,in , Indicates the first Predicted location of a historical trajectory, Indicates the first The centroid detected in the current frame. Indicates the coordinates of the centerline of the transport track. Indicates the offset weight. This represents the matching cost function.

5. The method for identifying anomalies in non-standard sample containers according to claim 1, characterized in that, Multiple frames of consecutive images of the same digital identity container are superimposed to generate a geometric envelope of the motion trajectory. Geometric feature parameters are extracted from the geometric envelope, including: For the same digital identity container, the centroid coordinates of each frame are extracted from N consecutive frames of images. A two-dimensional motion trajectory region is generated by superimposing the centroid trajectory. The convex hull is calculated as the geometric envelope of the container motion through the trajectory region. Geometric feature parameters are calculated based on the geometric envelope.

6. The method for identifying anomalies in non-standard sample containers according to claim 1, characterized in that, Based on geometric characteristic parameters and current velocity and acceleration, the expected dynamic sequence of geometric determinations is obtained, including: Based on the container's equivalent moment of inertia and dimensions, the equivalent mass of the container is determined according to a preset geometric equivalence relationship. The equivalent stiffness and equivalent damping are obtained according to a preset structural proportion relationship. These equivalent mass, equivalent stiffness, and equivalent damping are used as parameters of a preset dynamic model. Velocity and acceleration are used as external excitation inputs. By solving the preset dynamic model in the time domain, the theoretical displacement sequence of the container under the corresponding motion state is obtained. and theoretical velocity sequence The expected dynamic sequence is then... .

7. The method for identifying anomalies in non-standard sample containers according to claim 1, characterized in that, By using the difference between the measured dynamic sequence and the expected dynamic sequence, a geometrically independent residual anomaly feature sequence is obtained, namely: ,in These are residual abnormal features. This is a measured dynamic sequence. For the expected dynamic sequence, This represents the measured displacement of the container at time t. This represents the measured velocity of the container at time t. For theoretical displacement sequence, This is the theoretical velocity sequence.

8. A non-standard sample container anomaly identification system, used to implement the method according to any one of claims 1-7, characterized in that, include: The container state sequence generation module is used to collect visual and displacement data of the target container in real time during the transportation process, align the visual and displacement data in time, and generate the predicted position. In the current frame, a matching cost function is constructed based on preset constraints, and the predicted position is associated with the historical trajectory through the optimal matching algorithm to generate the container state sequence. The geometric feature parameter extraction module is used to overlay multiple frames of continuous images of the same digital identity container to generate a geometric envelope of motion trajectory, and extract geometric feature parameters from the geometric envelope. The residual anomaly feature acquisition module is used to calculate the velocity and acceleration of each container based on the container state sequence, obtain the expected dynamic sequence determined by geometry based on the geometric feature parameters and the current velocity and acceleration, and obtain the residual anomaly feature sequence that is independent of geometry by the difference between the measured dynamic sequence and the expected dynamic sequence. The cross-correlation coefficient acquisition module is used to calculate the cross-correlation coefficient between the expected dynamic sequence and the measured dynamic sequence; The anomaly detection module is used to obtain the geometric envelope offset of the container through the remaining anomaly feature sequence, calculate the damping offset based on the cross-correlation coefficient and obtain the label recognition confidence, calculate the comprehensive anomaly score and determine the anomaly level.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-7.

Images