Parcel jamming identification method and device, electronic equipment and readable storage medium

Abstract

This application discloses a package congestion identification method, apparatus, electronic device, and readable storage medium. The method includes: acquiring a target image; determining the target package congestion state of a target transport component based on the target image; acquiring the historical package congestion state of the target transport component; and determining the target congestion identification result of the target transport component based on the target package congestion state and the historical package congestion state. This method not only relies on the congestion prediction result of the target transport component at the current time point, but also combines the congestion prediction results of the target transport component before the current time point to judge the actual congestion identification result, which improves the richness of information during congestion identification. Furthermore, it eliminates the need for complex operations such as image fusion, requiring only the acquisition of package congestion states of the target transport component at multiple time points, thus ensuring both the accuracy and speed of congestion identification.

Description

Technical Field

[0001] This application relates to the field of logistics technology, specifically to a method, apparatus, electronic device, and readable storage medium for identifying package blockages. Background Technology

[0002] In transit hubs, packages are transported via conveyor belts. Due to the vast size of the facilities, packages need to pass through numerous conveyor belts during transport to the central location and subsequent distribution. In some sections, congestion is highly likely, leading to jams, damage, and reduced loading and unloading efficiency. However, current congestion detection methods, whether manual or machine-based, are not highly accurate. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, and readable storage medium for identifying package congestion, aiming to solve the problem of low accuracy in identifying package congestion in transit stations.

[0004] Firstly, this application provides a method for identifying package blockages, including:

[0005] Acquire the target image;

[0006] Based on the target image, determine the blockage status of the target package for the target transport component;

[0007] Obtain the historical package congestion status of the target transport component;

[0008] Based on the target package blockage status and the historical package blockage status, the target blockage identification result of the target transportation component is determined.

[0009] In one possible implementation of this application, determining the target package blockage status of the target transport component based on the target image includes:

[0010] Determining the target package blockage status of the target transport component based on the target image includes:

[0011] The target image is input into a preset blockage detection model to obtain the image features of the target image;

[0012] The image features are processed by the blockage detection model to predict the blockage status of the target package of the target transport component in the target image.

[0013] In one possible implementation of this application, the preset congestion detection model is obtained through the following steps:

[0014] Acquire training data, wherein the training data includes training images and a first sample blockage state of the first transport component in each training image;

[0015] Each of the training images is input into the initial blockage detection model to obtain the first package blockage state of the first transport component in each of the training images;

[0016] Based on the first package blockage state in each of the training images, a target training image is selected from each of the training images;

[0017] Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

[0018] In one possible implementation of this application, selecting a target training image from the training images based on the first package blockage state in each training image includes:

[0019] The confidence level of the first package blockage state in each of the training images is matched with a preset confidence level range, and the training images with the corresponding confidence level within the confidence level range are set as target training images.

[0020] And / or,

[0021] The training images are arranged according to their respective timestamps to obtain a training image sequence. The first package blockage state of adjacent training images in the training image sequence is compared to obtain the target training image in which the first package blockage state changes abruptly.

[0022] And / or,

[0023] From each of the training images, select the corresponding image to be identified where the first package blockage state is blocked, identify the package blockage area in the image to be identified, and obtain the area of ​​the package blockage area. If the area of ​​the package blockage area is less than or equal to a preset area threshold, then set the image to be identified as the target training image.

[0024] In one possible implementation of this application, obtaining training data includes:

[0025] Acquire training images;

[0026] The initial annotation state of each training image is obtained by predicting the state of each training image using a preset state annotation model.

[0027] The initial annotation states are corrected to determine the first sample blockage state of the first transport component in each training image, thereby obtaining training data.

[0028] In one possible implementation of this application, adjusting the parameters of the initial blockage detection model based on the first sample blockage state and the first package blockage state in the target training image to obtain the preset blockage detection model includes:

[0029] Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

[0030] The preset test image is input into the blockage detection model to be tested to obtain the second package blockage state of the second transport component in the test image;

[0031] Based on the second sample blockage state of the second transport component and the second package blockage state in the test image, determine the model evaluation value of the blockage detection model to be tested;

[0032] If the model evaluation value meets the preset training termination condition, then the blockage detection model to be tested is set to the preset blockage detection model.

[0033] In one possible implementation of this application, selecting a target training image from the training images based on the first package blockage state in each training image includes:

[0034] Based on the first package blockage state corresponding to each training image, select the target training image to be screened from each training image;

[0035] Obtain the similarity between any two images in the training images of the target to be screened;

[0036] If the similarity is greater than a preset similarity threshold, one of the two images is removed from the target training images to be screened, and the target training image is obtained.

[0037] In one possible implementation of this application, determining the target blockage identification result of the target transport component based on the target package blockage status and the historical package blockage status includes:

[0038] The historical package congestion status and the target package congestion status are input into a preset state machine model, and the target congestion identification result of the target transportation component is determined according to the preset state transition conditions.

[0039] The target blockage identification result is output through the preset target terminal.

[0040] Secondly, this application provides a package blockage identification device, comprising:

[0041] The first acquisition unit is used to acquire the target image;

[0042] The determining unit is used to determine the blockage status of the target package of the target transport component based on the target image;

[0043] The second acquisition unit is used to acquire the historical package blockage status of the target transport component;

[0044] The identification unit is used to determine the target blockage identification result of the target transportation component based on the target package blockage status and the historical package blockage status.

[0045] In one possible implementation of this application, the determining unit is further configured to:

[0046] The target image is input into a preset blockage detection model to obtain the image features of the target image;

[0047] The image features are processed by the blockage detection model to predict the blockage status of the target package of the target transport component in the target image.

[0048] In one possible implementation of this application, the determining unit is further configured to:

[0049] Acquire training data, wherein the training data includes training images and a first sample blockage state of the first transport component in each training image;

[0050] Each of the training images is input into the initial blockage detection model to obtain the first package blockage state of the first transport component in each of the training images;

[0051] Based on the first package blockage state in each of the training images, a target training image is selected from each of the training images;

[0052] Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

[0053] In one possible implementation of this application, the determining unit is further configured to:

[0054] The confidence level of the first package blockage state in each of the training images is matched with a preset confidence level range, and the training images with the corresponding confidence level within the confidence level range are set as target training images.

[0055] And / or,

[0056] The training images are arranged according to their respective timestamps to obtain a training image sequence. The first package blockage state of adjacent training images in the training image sequence is compared to obtain the target training image in which the first package blockage state changes abruptly.

[0057] And / or,

[0058] From each of the training images, select the corresponding image to be identified where the first package blockage state is blocked, identify the package blockage area in the image to be identified, and obtain the area of ​​the package blockage area. If the area of ​​the package blockage area is less than or equal to a preset area threshold, then set the image to be identified as the target training image.

[0059] In one possible implementation of this application, the determining unit is further configured to:

[0060] Acquire training images;

[0061] The initial annotation state of each training image is obtained by predicting the state of each training image using a preset state annotation model.

[0062] The initial annotation states are corrected to determine the first sample blockage state of the first transport component in each training image, thereby obtaining training data.

[0063] In one possible implementation of this application, the determining unit is further configured to:

[0064] Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

[0065] The preset test image is input into the blockage detection model to be tested to obtain the second package blockage state of the second transport component in the test image;

[0066] Based on the second sample blockage state of the second transport component and the second package blockage state in the test image, determine the model evaluation value of the blockage detection model to be tested;

[0067] If the model evaluation value meets the preset training termination condition, then the blockage detection model to be tested is set to the preset blockage detection model.

[0068] In one possible implementation of this application, the determining unit is further configured to:

[0069] Based on the first package blockage state corresponding to each training image, select the target training image to be screened from each training image;

[0070] Obtain the similarity between any two images in the training images of the target to be screened;

[0071] If the similarity is greater than a preset similarity threshold, one of the two images is removed from the target training images to be screened, and the target training image is obtained.

[0072] In one possible implementation of this application, the identification unit is further used for:

[0073] The historical package congestion status and the target package congestion status are input into a preset state machine model, and the target congestion identification result of the target transportation component is determined according to the preset state transition conditions.

[0074] The target blockage identification result is output through the preset target terminal.

[0075] Thirdly, this application also provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor calls the computer program in the memory, it executes the steps in any of the package blockage identification methods provided in this application.

[0076] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the package blockage identification methods provided in this application.

[0077] In summary, the package congestion identification method provided in this application includes: acquiring a target image; determining the target package congestion status of a target transport component based on the target image; acquiring the historical package congestion status of the target transport component; and determining the target congestion identification result of the target transport component based on the target package congestion status and the historical package congestion status. Therefore, the package congestion identification method provided in this application does not only rely on the congestion prediction result of the target transport component at the current time point, but also combines the congestion prediction results of the target transport component before the current time point to judge the actual congestion identification result. This improves the richness of information during congestion identification and eliminates the need for complex operations such as image fusion; it only requires acquiring the package congestion status of the target transport component at multiple time points, ensuring both the accuracy of congestion identification and improving identification speed. Attached Figure Description

[0078] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0079] Figure 1 This is a schematic diagram illustrating an application scenario of the package blockage identification method provided in this application embodiment;

[0080] Figure 2 This is a flowchart illustrating a package blockage identification method provided in an embodiment of this application;

[0081] Figure 3 This is a schematic diagram of a training image provided in an embodiment of this application;

[0082] Figure 4 This is a flowchart illustrating the process of determining the target blockage identification result provided in an embodiment of this application;

[0083] Figure 5 This is a schematic diagram of a process for obtaining a target training image provided in an embodiment of this application;

[0084] Figure 6 This is a schematic diagram of an embodiment of the package blockage identification device provided in this application.

[0085] Figure 7 This is a schematic diagram of an embodiment of the electronic device provided in this application. Detailed Implementation

[0086] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0087] In the description of the embodiments of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0088] To enable any person skilled in the art to implement and use this application, the following description is provided. In this description, details are set forth for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be implemented without using these specific details. In other instances, well-known processes will not be described in detail to avoid obscuring the description of the embodiments of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in the embodiments of this application.

[0089] This application provides a method, apparatus, electronic device, and readable storage medium for identifying package blockages. The package blockage identification apparatus can be integrated into an electronic device, which may be a server or a terminal, etc.

[0090] The execution subject of the package congestion identification method in this application embodiment can be the package congestion identification device provided in this application embodiment, or different types of electronic devices such as server equipment, physical host, or user equipment (UE) that integrate the package congestion identification device. The package congestion identification device can be implemented in hardware or software. The UE can be a terminal device such as a smartphone, tablet computer, laptop computer, handheld computer, desktop computer, or personal digital assistant (PDA).

[0091] The electronic device can operate independently or in a cluster.

[0092] See Figure 1 , Figure 1 This is a schematic diagram of a package blockage identification system provided in an embodiment of this application. The package blockage identification system may include an electronic device 100, which integrates a package blockage identification device.

[0093] In addition, such as Figure 1 As shown, the package blockage identification system may also include a memory 200 for storing data, such as text data.

[0094] It should be noted that, Figure 1 The schematic diagram of the package congestion identification system shown is merely an example. The package congestion identification system and scenario described in this application are for the purpose of more clearly illustrating the technical solutions of this application and do not constitute a limitation on the technical solutions provided in this application. As those skilled in the art will know, with the evolution of package congestion identification systems and the emergence of new business scenarios, the technical solutions provided in this invention are also applicable to similar technical problems.

[0095] The following describes the package congestion identification method provided in the embodiments of this application. In the embodiments of this application, an electronic device is used as the execution subject. For the sake of simplicity and ease of description, the execution subject will be omitted in the subsequent method embodiments. The package congestion identification includes: acquiring a target image; determining the target package congestion status of the transport component based on the target image; acquiring the historical package congestion status of the transport component; and determining the target congestion identification result of the transport component based on the target package congestion status and the historical package congestion status.

[0096] Reference Figure 2 , Figure 2 This is a flowchart illustrating a package blockage identification method provided in an embodiment of this application. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here. Specifically, the package blockage identification method may include the following steps 201-204, wherein:

[0097] 201. Obtain the target image.

[0098] The target image is an image containing the target transport component to be identified.

[0099] Transport components refer to components used to transport packages. For example, transport components can be belt conveyors, plate conveyors, trolley conveyors, etc.

[0100] This application does not limit the method for acquiring the target image. For example, the target image can be acquired by any of the methods described below:

[0101] (1) A camera can be integrated on the target transport component to capture video frames or images of the target transport component in real time and use them as the target image.

[0102] (2) A camera can be installed above the target transport component to capture video frames or images of the target transport component in real time. At the same time, the camera above the target transport component can be connected to an electronic device to acquire the video frames or images of the target transport component captured by the camera above the target transport component online and use them as the target image.

[0103] (3) The target transport component image captured by the camera can be read from the relevant storage medium containing the images captured by the camera and used as the target image.

[0104] (4) Read the video frames or images of the target transport component that have been pre-acquired and stored inside the electronic device, and use them as the target image.

[0105] The camera can capture images according to preset shooting methods, such as setting the shooting height, shooting direction, or shooting distance. The specific shooting method can be adjusted according to the camera itself, and is not limited here.

[0106] 202. Based on the target image, determine the blockage status of the target package of the target transport component.

[0107] Package congestion status refers to the predicted congestion status of packages being transported on a transport component. For example, package congestion status can include two states: "predicted congestion" and "predicted no congestion". When it is predicted that packages being transported on a transport component will be congested, the package congestion status of that transport component is "predicted congestion". When it is predicted that packages being transported on a transport component will be transported normally without congestion, the package congestion status is "predicted no congestion".

[0108] The target package congestion status refers to the predicted congestion status of packages transported on the target transport component at the time the target image was captured. Therefore, it is understandable that the target package congestion status may differ from the actual congestion situation on the target transport component. The reasons for this discrepancy can include the following two points:

[0109] (1) The model or algorithm used to predict the blockage status of the target package is not accurate enough. It predicts that the package on the target transport component is blocked when there is no blockage on the package.

[0110] (2) The package on the target transport component may only be stuck by the gaps, protrusions or other parts on the target transport component at the time when the target image is taken. However, the stuck state of the package is eliminated after the time when the target image is taken, and it can continue to be transported on the target transport component. Therefore, if the state of the package at only one time point is used to determine whether it is blocked, it is easy to make a misjudgment and the judgment is not accurate enough.

[0111] In some embodiments, computer vision (CV) technology can be used to process the target image to predict the blockage status of the target package in the target transport component. For example, an initial blockage detection model can be constructed and trained to obtain a preset blockage detection model, and the target image can be input into the preset blockage detection model to predict the blockage status of the target package in the target transport component. The blockage detection model can be composed of convolutional neural networks (CNNs), which obtain image features in the target image through convolution, pooling, and other methods, and predict the blockage status of the target package in the target transport component based on the image features in the target image. For example, all packages in the target image and the target transport component can be obtained based on the image features in the target image, and then the sum of the areas of the packages and the target transport component in the corresponding regions of the target image can be calculated. If the sum of the areas is greater than a preset threshold, it indicates that there are too many packages in the target image, which may be due to package blockage causing a large number of packages to be packed together and unable to be transported forward. Therefore, the obtained target package blockage status is "predicted blockage". Alternatively, all packages and their positions in the target image can be obtained based on the image features in the target image. If the distance between the positions of each package is less than a preset distance threshold, it indicates that the packages on the target transport component may be blocked, causing multiple packages to be close together. Therefore, the blockage status of the target packages is "predicted blockage".

[0112] For example, a pre-defined congestion detection model can be trained in the following way:

[0113] (1) Obtain training data.

[0114] The training data includes training images and a first sample blockage state of the first transport component in each training image. The first sample blockage state refers to the actual blockage state of the first transport component. For example, the first sample blockage state can include two states: "blocked" and "not blocked." When a package being transported on the first transport component is blocked, the first sample blockage state of the first transport component is "blocked." When a package being transported on the first transport component is being transported normally without blockage, the first sample blockage state is "not blocked." There are various methods to obtain the first sample blockage state; for example, the first sample blockage state of the first transport component in the training image can be obtained through any of the following methods:

[0115] (i) Using manual observation, observe whether the package is blocked in each training image to obtain the first sample blockage status of the first transport component in each training image.

[0116] (ii) Using an automatic machine acquisition method, such as obtaining the time point when the operating parameters of the first transport component become abnormal, and matching this time point with the timestamps of each training image, the first sample corresponding to the matched training image is marked as "blocked", and the first sample corresponding to the other training images is marked as "not blocked". The reason for this judgment is that when the package is blocked, the load on the first transport component will increase, so the operating parameters of the first transport component will change compared to when it is not blocked.

[0117] The first transport component may be the same as or different from the target transport component.

[0118] This application does not limit the method for obtaining training images. For example, training images can be obtained by one of the methods described below:

[0119] (1.1) A camera can be integrated on the first transport component to capture video frames or images of the first transport component in real time and use them as training images.

[0120] (1.2) A camera can be installed above the first transport component to capture video frames or images of the first transport component in real time. At the same time, the camera above the first transport component can be connected to an electronic device to acquire the video frames or images of the first transport component captured by the camera above the first transport component online and use them as training images.

[0121] (1.3) The first transport component image captured by the camera can be read from the relevant storage medium storing the images captured by the camera and used as the training image.

[0122] (1.4) Read the video frames or images of the first transport component that have been pre-acquired and stored inside the electronic device, and use them as training images.

[0123] In some embodiments, if training images are labeled automatically by machine, the labels can be corrected after labeling to obtain the correct first sample blockage state. Exemplarily, this may include the following steps: acquiring training images; predicting the initial labeling state of each training image using a preset state labeling model; correcting the initial labeling state to determine the first sample blockage state of the first transport component in each training image, thereby obtaining training data. The preset state labeling model can be an initial blockage detection model. When correcting the initial labeling state, each training image can be corrected manually to obtain the correct label.

[0124] (2) Input each of the training images into the initial blockage detection model to obtain the first package blockage state of the first transport component in each of the training images.

[0125] The initial congestion detection model is a pre-built classification model, which can be constructed using Convolutional Neural Networks (CNNs). It obtains package features from the training image through convolution and pooling methods, and predicts the first package congestion state of the first transport component based on these features. For example, all packages in the training image can be obtained based on image features. Then, the sum of the areas of the packages and the corresponding regions of the first transport component in the training image is calculated. If the sum of the areas is greater than a preset area threshold, it indicates that there are too many packages in the training image, possibly indicating congestion causing a large number of packages to be packed together and unable to move forward. Therefore, the first package congestion state is "predicted congestion." Alternatively, all packages and their positions in the training image can be obtained based on image features. If the distance between the positions of all packages is less than a preset distance threshold, it indicates that the first transport component may be congested, causing multiple packages to be packed together. Therefore, the first package congestion state is also "predicted congestion."

[0126] (3) Select a target training image from each of the training images based on the first package blockage state in each of the training images.

[0127] The target training image refers to the training image from which the initial congestion detection model is prone to making incorrect predictions. Since the initial congestion detection model predicts the congestion state of the first package based on pixel values, pixel distribution, and other information in the training image, situations such as a similar background color to the package on the first transport component, or difficulties in extracting package features, can lead to a congestion state that is the opposite of the actual congestion situation on the first transport component. To improve the prediction accuracy of the initial congestion detection model in these situations, a target training image can be extracted and used to specifically train the initial congestion detection model, resulting in a pre-defined congestion detection model.

[0128] For example, target training images can be selected in three ways, either by using only one method or by using multiple methods simultaneously:

[0129] (A) Match the confidence level of the first package blockage state in each of the training images with a preset confidence level range, and set the training images with the corresponding confidence level within the confidence level range as target training images.

[0130] When the initial congestion detection model predicts the first package congestion state, it simultaneously provides the confidence score for each training image corresponding to the first package congestion state. The confidence scores of each training image can be matched with a preset confidence score range. When the confidence score range includes the confidence score of a training image, that training image is used as the target training image. Confidence score refers to the likelihood of package congestion on the first transport component, and the confidence score range is a preset range used to evaluate the likelihood of package congestion on the first transport component. For example, [0.3, 0.8] can be used as the confidence score range. When the confidence score of a training image is lower than 0.3 or higher than 0.8, that training image is set as the target training image. It can be seen that by setting a confidence score range, images with extremely low and extremely high likelihood of package congestion on the first transport component can be excluded from the training images, i.e., training images that are unlikely to predict incorrectly are excluded, leaving only those that may predict incorrectly. On the one hand, this gradually improves the prediction accuracy of the initial congestion detection model during training; on the other hand, it reduces the number of training images input to the initial congestion detection model, saving training time and improving training efficiency.

[0131] (B) Arrange the training images according to their respective timestamps to obtain a training image sequence, and compare the first package blockage state of adjacent training images in the training image sequence to obtain a target training image in which the first package blockage state changes abruptly.

[0132] A timestamp is used to represent the time when an image was captured. If the timestamp of an image is "December 9, 2021, 17:00", then the image was captured on December 9, 2021, at 17:00. Using the timestamps of each training image, the images can be arranged in order of their capture time to obtain a training image sequence.

[0133] The following example illustrates a comparison process for obtaining target training images: (Reference) Figure 3 , Figure 3The diagram illustrates one scenario for a training image sequence, where training images AE consist of five images arranged sequentially by their capture time. Training image A was captured earliest, and training image E was captured latest. During comparison, the first package blockage state of training image A can be compared with the first package blockage state of training image B, and the first package blockage state of training image B can be compared with the first package blockage state of training image C, and so on. If the first package blockage state of training image A differs from that of training image B, i.e., if the first package blockage state of training image B undergoes a sudden change, then training image B can be used as the target training image. In this embodiment, the reason for obtaining the target training image is that changes in the predicted blockage result may be due to prediction errors. Therefore, it is necessary to train the initial blockage detection model using the first sample blockage state and the first package blockage state of the target training image to reduce the probability of prediction errors. On the one hand, this can gradually improve the prediction accuracy of the initial blockage detection model during training; on the other hand, it can reduce the number of training images input to the initial blockage detection model, saving training time and improving training efficiency.

[0134] (C) From each of the training images, select the corresponding image to be identified in which the first package blockage state is blocked, identify the package blockage area in the image to be identified, and obtain the area of ​​the package blockage area. If the area of ​​the package blockage area is less than or equal to a preset area threshold, then set the image to be identified as the target training image.

[0135] In step (C), the area of ​​the package blockage region in the image to be identified corresponding to the first package blockage state of "predicted blockage" can be compared with a preset area threshold to select the target training image.

[0136] In some embodiments, the package blockage area may refer to the total area corresponding to the first transport component and the package on the first transport component in the image to be identified, or it may refer only to the area corresponding to the package on the first transport component in the image to be identified.

[0137] For example, when the package blockage area refers to the total area corresponding to the first transport component and the package on the first transport component in the image to be identified, a preset area threshold can be determined based on the area of ​​the area corresponding to the first transport component in the image to be identified. For instance, the area of ​​the area corresponding to the first transport component in the image to be identified can be used as the preset area threshold. Thus, images of packages that were not captured but whose first package is blocked can be used as target training images, gradually reducing the possibility of such prediction errors during the initial training of the blockage detection model.

[0138] When the package blockage area refers to the area corresponding to the package on the first transport component in the image to be identified, a preset area threshold can be determined based on the area of ​​the region corresponding to a single package in the image to be identified. For example, the minimum number of packages that appear in the image to be identified when there is no blockage can be preset based on experience, and then the area threshold can be calculated based on the area of ​​the region corresponding to a single package in the image to be identified and the minimum number.

[0139] (4) Adjust the parameters in the initial blockage detection model according to the first sample blockage state in the target training image and the first package blockage state in the target training image to obtain the preset blockage detection model.

[0140] In some embodiments, data augmentation processing can also be performed on the target training image. For example, data augmentation processing such as flipping, random cropping, color jittering, translation, scaling, contrast transformation, and noise perturbation can be performed on the target training image. Then, the parameters in the initial blockage detection model can be adjusted using the obtained image to obtain the preset blockage detection model.

[0141] Furthermore, in this embodiment, the initial training process of the congestion detection model can be completed on the Automl platform. Specifically, the Automl platform is connected to the preset selection algorithm deployed on the front end. When the preset selection algorithm deployed on the front end selects the target training image, it can feed the target training image back to the Automl platform, and the initial training process of the congestion detection model is completed on the Automl platform.

[0142] The Automl platform is a platform for automating machine learning in the cloud.

[0143] In some embodiments, the parameters in the initial clogging detection model can be adjusted based on the first sample clogging state and the first package clogging state of each target training image. Then, the adjusted clogging detection model can be tested based on a preset test image to evaluate its predictive ability.

[0144] For example, the congestion detection model to be tested can be tested through the following steps:

[0145] (1) Based on the first sample blockage state in the target training image and the first package blockage state in the target training image, adjust the parameters in the initial blockage detection model to obtain the preset blockage detection model;

[0146] (2) Input the preset test image into the blockage detection model to be tested to obtain the second package blockage state of the second transport component in the test image;

[0147] The second transport component may be the same as or different from the first transport component.

[0148] (3) Determine the model evaluation value of the blockage detection model to be tested based on the second sample blockage state of the second transport component and the second package blockage state in the test image;

[0149] The second sample blockage status refers to the actual blockage situation of the package on the second transport component in the test image.

[0150] Model evaluation values ​​can be precision, recall, etc., which can be used to evaluate the predictive ability of the blockage detection model under test.

[0151] (4) If the model evaluation value meets the preset training termination condition, the blockage detection model to be tested is set to the preset blockage detection model.

[0152] The training termination condition is a preset condition used to evaluate whether the blockage detection model under test can effectively detect blockages. In some embodiments, "the model evaluation value is greater than a preset evaluation threshold" can be used as the training termination condition, and the evaluation threshold can be set by the operator based on experience.

[0153] For images that are predicted incorrectly in the test images, the reasons for the prediction errors can be collected manually, and suitable images can be selected from the target training images to continue training the blockage detection model to be tested.

[0154] 203. Obtain the historical package blockage status of the target transport component.

[0155] Historical package congestion status refers to the predicted congestion status of packages on the target transport component before the shooting time corresponding to the target image.

[0156] In this embodiment of the application, there may be only one or multiple historical package blockage states.

[0157] For example, when there are multiple historical package congestion states, the historical package congestion state can be obtained using the following method:

[0158] First, within a historical period preceding the time point corresponding to the capture of the target image, the predicted congestion status of the target transport component at multiple historical time points is obtained. Assuming the time point corresponding to the target image is 17:00 on December 9, 2021, then the predicted congestion status of the target transport component at each full minute (16:55, 16:56, 16:57, 16:58, and 16:59) within the 5 minutes preceding 17:00 on December 9, 2021, i.e., from 16:55 to 17:00 on December 9, 2021.

[0159] When there is only one historical package congestion status, the predicted congestion status of the target transport component can be obtained at any historical time point before the time point corresponding to the target image. To ensure accurate identification, the predicted congestion status of the target transport component can be obtained at adjacent historical time points to obtain a single historical package congestion status. For example, if the time point corresponding to the target image is 17:00 on December 9, 2021, the predicted congestion status of the target transport component at 16:59 on December 9, 2021 can be used as the historical package congestion status.

[0160] For example, the historical package congestion status can be obtained from historical images, which can be achieved through the following steps:

[0161] (a1) Obtain historical images of the target transport component. Historical images refer to images taken before the time point corresponding to the target image. For example, historical images can be obtained through one of the following methods:

[0162] (1.1) A camera can be integrated on the target transport component to capture video frames or images of the target transport component in real time and use them as historical images.

[0163] (1.2) A camera can be installed above the target transport component to capture video frames or images of the target transport component in real time. At the same time, the camera above the target transport component can be connected to an electronic device to acquire the video frames or images of the target transport component captured by the camera above the target transport component online and use them as historical images.

[0164] (1.3) The target transport component image captured by the camera can be read from the relevant storage medium storing the images captured by the camera and used as a historical image.

[0165] (1.4) Read the video frames or images of the target transport components that have been pre-acquired and stored inside the electronic device, and use them as historical images.

[0166] (a2) Input the historical image into the preset blockage detection model to obtain the historical package blockage status of the corresponding historical image.

[0167] If there is only one historical package congestion state, then only one historical image is input into the preset congestion detection model to obtain one historical package congestion state. If there are multiple historical package congestion states, then multiple historical images are input into the preset congestion detection model to obtain multiple corresponding historical package congestion states.

[0168] 204. Based on the target package blockage status and the historical package blockage status, determine the target blockage identification result of the target transportation component.

[0169] The congestion identification result is used to characterize the actual congestion status of packages on the transport component. If a package is congested on the transport component, the congestion identification result can be "actual congestion"; if the package is not congested, the congestion identification result can be "actual non-congestion". The difference between the congestion identification result and the package congestion status is as follows: the package congestion status is a predicted congestion status obtained from the image, which may not match the actual congestion status of packages on the transport component due to prediction errors or other reasons. The congestion identification result, on the other hand, integrates the package congestion status of the transport component at the corresponding time point, as well as the package congestion status of packages before that time point. That is, it is a result obtained by combining information from the image taken at a certain time point and the historical images taken before that image. Compared with the package congestion status, the information is richer, so the congestion identification result can be considered to be consistent with the actual congestion status of packages on the transport component.

[0170] The target congestion identification result refers to the actual congestion status of the package on the target transport component at the time the target image was captured. If the package on the target transport component is determined to be congested based on the target package congestion status and historical package congestion status, the target congestion identification result can be "actual congestion". If the package on the target transport component is determined to be uncongested based on the target package congestion status and historical package congestion status, the target congestion identification result can be "actually uncongested".

[0171] The reason why the target package's congestion status does not match the actual congestion situation has been explained above, and will not be repeated here.

[0172] To ensure the accuracy of target congestion identification results, the result is typically determined based on the target package's congestion status and multiple historical package congestion statuses. For example, the following provides a specific judgment logic for determining the target congestion identification result:

[0173] Logic 1: If the target package's blockage status is "predicted blockage" and all historical packages' blockage statuses are also "predicted blockage", then the target blockage identification result is determined to be "actual blockage".

[0174] If the historical package blockage status corresponding to multiple historical images is "predicted blockage", it means that the prediction results obtained based on multiple different images are all "predicted blockage". Therefore, it can be determined that the reason (2) that the target package blockage status does not match the actual blockage situation of the package on the target transport component in step 202 will not occur. As for reason (1), after training the target training images, the probability that the prediction results of multiple images will not match the actual blockage situation through the preset blockage detection model is extremely low. Therefore, it can also be considered that reason (1) will not occur. At this time, if the target package blockage status is "predicted blockage", it can be determined that the target blockage identification result is "actual blockage".

[0175] Logic 2: If the target package's blockage status is "predicted blockage" and at least one of the historical package blockage statuses is "predicted non-blockage", then the target blockage identification result is determined to be "actually non-blockage".

[0176] If at least one of the historical package blockage states corresponding to multiple historical images is "predicted not blocked", it means that at least one of the reasons (1) and (2) for the discrepancy between the target package blockage state and the actual blockage of the package on the target transport component in step 202 may occur. In order to avoid false alarms, the target blockage identification result can be determined to be "actually not blocked".

[0177] In summary, the package congestion identification method provided in this application includes: acquiring a target image; determining the target package congestion status of a target transport component based on the target image; acquiring the historical package congestion status of the target transport component; and determining the target congestion identification result of the target transport component based on the target package congestion status and the historical package congestion status. Therefore, the package congestion identification method provided in this application does not only rely on the congestion prediction result of the target transport component at the current time point, but also combines the congestion prediction results of the target transport component before the current time point to judge the actual congestion identification result. This improves the richness of information during congestion identification and eliminates the need for complex operations such as image fusion; it only requires acquiring the package congestion status of the target transport component at multiple time points, ensuring both the accuracy of congestion identification and improving identification speed.

[0178] In some embodiments, a preset state machine model can be used to determine the target blockage identification result of the target transport component based on the target package blockage status and historical package blockage status. A state machine, short for Finite-state Automaton, is a model for acquiring temporal information. (Reference) Figure 4At this point, the step "determine the target blockage identification result of the target transport component based on the target package blockage status and the historical package blockage status" includes:

[0179] 301. Input the historical package blockage status and the target package blockage status into a preset state machine model, and determine the target blockage identification result of the target transportation component according to the preset state transition conditions.

[0180] State transition conditions refer to the triggering conditions when a state transition occurs in a state machine model. In some embodiments, to avoid the discrepancy between the target package blockage state and the actual blockage situation of the package on the target transport component (2), the state transition condition from "actually not blocked" to "actually blocked" can be set to "n consecutive package blockage states of 'predicted blockage' have been input", and the state transition condition from "actually blocked" to "actually not blocked" can be set to "one package blockage state of 'predicted not blocked' has been input". Therefore, the target transport component is only judged to be blocked when the package blockage state of n consecutive images is 'predicted blockage', thus improving the accuracy of blockage identification.

[0181] For ease of explanation, the basic concepts of the state machine model are given below.

[0182] A state can refer to the output of a state machine model at each point in time. A state machine must contain at least two states. In the embodiments of this application, a state refers to the blockage identification result of the target transport component at each point in time, that is, the state is one of "actual blockage" and "actual non-blockage".

[0183] An event can refer to the input of a state machine model. In this embodiment, an event refers to a package blockage state.

[0184] A transition can refer to a change in state; a change occurs when an event meets the conditions for a state transition. In the embodiments of this application, a transition occurs when the state changes from "actually blocked" to "actually not blocked," or from "actually not blocked" to "actually blocked."

[0185] The following is a concrete example to illustrate the workflow of a state machine. Suppose there are four historical package congestion states, which are arranged in chronological order: the first historical package congestion state, the second historical package congestion state, the third historical package congestion state, and the fourth historical package congestion state. The state transition condition is set to "five consecutive package congestion states with the prediction of congestion have been input". The preset initial state in the state machine model is "actually not congested".

[0186] (1) Input the first historical package blockage state into the preset state machine model;

[0187] (2) The state machine model reads the information contained in the first historical package blockage state. If the first historical package blockage state is "predicted blockage", since the state transition condition of changing the state from "actually not blocked" to "actually blocked" is not met at this time, the state of the historical time point corresponding to the first historical package blockage state is the initial state "actually not blocked".

[0188] (3) Input the second historical package blockage status, the third historical package blockage status, and the fourth historical package blockage status in sequence. If the second historical package blockage status, the third historical package blockage status, and the fourth historical package blockage status are all "predicted blockage", then, since the state transition condition of changing the status from "actually not blocked" to "actually blocked" is not met each time it is input, the status of the historical time points corresponding to the second historical package blockage status, the third historical package blockage status, and the fourth historical package blockage status are all the initial status "actually not blocked".

[0189] (4a) If the target package's congestion status is "predicted congestion," when the target package's congestion status is input, the state transition condition from "actually not congested" to "actually congested" is met. Therefore, the state at the shooting time corresponding to the target package's congestion status changes from the initial state "actually not congested" to the state "actually congested," meaning the target congestion identification result is "actually congested." Subsequently, if a future package congestion status is input again, and the future package congestion status is "predicted not congested," the state transition condition from "actually congested" to "actually not congested" is met. Therefore, the state changes to "actually not congested," meaning that at the future time point corresponding to the future package congestion status, the congestion identification result for the target transport component is "actually not congested."

[0190] (4b) If the target package is in a “predicted not blocked” state, the state machine model will restart counting the number of consecutive “predicted blocked” inputs.

[0191] 302. Output the target blockage identification result through the preset target terminal.

[0192] In some embodiments, to reduce the number of target training images and improve the training speed of the initial blockage detection model, duplicate images selected from the training images can be removed based on the similarity between images to obtain the target training images. (Reference) Figure 5 At this point, the step "selecting a target training image from each of the training images based on the first package blockage state in each of the training images" includes...

[0193] 401. Based on the first package blockage state corresponding to each of the training images, select the target training image to be screened from each of the training images.

[0194] The steps for selecting target training images to be screened can be any of the methods (A), (B), and (C) mentioned above, and will not be elaborated further.

[0195] 402. Obtain the similarity between any two images in the training images of the target to be screened.

[0196] In this embodiment of the application, the similarity between any two images in the training images of the target to be screened can be obtained by using the SSIM (Structural Similarity) algorithm.

[0197] The SSIM algorithm determines the similarity between two images by calculating their mean, variance, and covariance. The algorithm used here to calculate the target similarity between any two images is merely an example; in reality, target similarity can also be calculated using other image similarity algorithms, or future image similarity algorithms.

[0198] 403. If the similarity is greater than a preset similarity threshold, remove one of the two images from the target training images to be screened to obtain the target training image.

[0199] If the similarity between two target training images to be screened is greater than the preset similarity threshold, it means that the two target training images to be screened are duplicates. Even if one of them is removed, it will not have a significant impact on the training effect. Therefore, in order to improve the training speed of the initial blockage detection model, the two target training images to be screened can be deduplicated to obtain the deduplicated target training images.

[0200] To better implement the package blockage identification method in this application embodiment, based on the package blockage identification method, this application embodiment also provides a package blockage identification device, such as... Figure 6 The diagram shown is a structural schematic of one embodiment of the package blockage identification device in this application. The package blockage identification device 500 includes:

[0201] The first acquisition unit 501 is used to acquire the target image;

[0202] The determining unit 502 is used to determine the target package blockage status of the target transport component based on the target image;

[0203] The second acquisition unit 503 is used to acquire the historical package blockage status of the target transport component;

[0204] The identification unit 504 is used to determine the target blockage identification result of the target transportation component based on the target package blockage status and the historical package blockage status.

[0205] In one possible implementation of this application, the determining unit 502 is further configured to:

[0206] The target image is input into a preset blockage detection model to obtain the image features of the target image;

[0207] The image features are processed by the blockage detection model to predict the blockage status of the target package of the target transport component in the target image.

[0208] In one possible implementation of this application, the determining unit 502 is further configured to:

[0209] Acquire training data, wherein the training data includes training images and a first sample blockage state of the first transport component in each training image;

[0210] Each of the training images is input into the initial blockage detection model to obtain the first package blockage state of the first transport component in each of the training images;

[0211] Based on the first package blockage state in each of the training images, a target training image is selected from each of the training images;

[0212] Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

[0213] In one possible implementation of this application, the determining unit 502 is further configured to:

[0214] The confidence level of the first package blockage state in each of the training images is matched with a preset confidence level range, and the training images with the corresponding confidence level within the confidence level range are set as target training images.

[0215] And / or,

[0216] The training images are arranged according to their respective timestamps to obtain a training image sequence. The first package blockage state of adjacent training images in the training image sequence is compared to obtain the target training image in which the first package blockage state changes abruptly.

[0217] And / or,

[0218] From each of the training images, select the corresponding image to be identified where the first package blockage state is blocked, identify the package blockage area in the image to be identified, and obtain the area of ​​the package blockage area. If the area of ​​the package blockage area is less than or equal to a preset area threshold, then set the image to be identified as the target training image.

[0219] In one possible implementation of this application, the determining unit 502 is further configured to:

[0220] Acquire training images;

[0221] The initial annotation state of each training image is obtained by predicting the state of each training image using a preset state annotation model.

[0222] The initial annotation states are corrected to determine the first sample blockage state of the first transport component in each training image, thereby obtaining training data.

[0223] In one possible implementation of this application, the determining unit 502 is further configured to:

[0224] Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

[0225] The preset test image is input into the blockage detection model to be tested to obtain the second package blockage state of the second transport component in the test image;

[0226] Based on the second sample blockage state of the second transport component and the second package blockage state in the test image, determine the model evaluation value of the blockage detection model to be tested;

[0227] If the model evaluation value meets the preset training termination condition, then the blockage detection model to be tested is set to the preset blockage detection model.

[0228] In one possible implementation of this application, the determining unit 502 is further configured to:

[0229] Based on the first package blockage state corresponding to each training image, select the target training image to be screened from each training image;

[0230] Obtain the similarity between any two images in the training images of the target to be screened;

[0231] If the similarity is greater than a preset similarity threshold, one of the two images is removed from the target training images to be screened, and the target training image is obtained.

[0232] In one possible implementation of this application, the identification unit 504 is further configured to:

[0233] The historical package congestion status and the target package congestion status are input into a preset state machine model, and the target congestion identification result of the target transportation component is determined according to the preset state transition conditions.

[0234] The target blockage identification result is output through the preset target terminal.

[0235] In practice, each of the above units can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units, please refer to the previous method embodiments, which will not be repeated here.

[0236] Since the package blockage identification device can perform the steps of the package blockage identification method in any embodiment, it can achieve the beneficial effects that the package blockage identification method in any embodiment of this application can achieve, as detailed in the preceding description, and will not be repeated here.

[0237] Furthermore, to better implement the package blockage identification method in this application embodiment, based on the package blockage identification method, this application embodiment also provides an electronic device, see below. Figure 7 , Figure 7 This illustration shows a structural diagram of an electronic device according to an embodiment of this application. Specifically, the electronic device provided in this embodiment includes a processor 601. The processor 601 is used to execute a computer program stored in a memory 602 to implement each step of the package blockage identification method in any embodiment; or, the processor 601 is used to execute a computer program stored in a memory 602 to implement, for example... Figure 6 The functions of each unit in the corresponding embodiment.

[0238] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 602 and executed by processor 601 to complete the embodiments of this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a computer device.

[0239] The electronic device may include, but is not limited to, processor 601 and memory 602. Those skilled in the art will understand that the illustrations are merely examples of an electronic device and do not constitute a limitation on the device. It may include more or fewer components than illustrated, or combine certain components, or use different components.

[0240] Processor 601 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the electronic device through various interfaces and lines.

[0241] The memory 602 can be used to store computer programs and / or modules. The processor 601 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 602 and by calling data stored in the memory 602. The memory 602 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device (such as audio data, video data, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0242] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the package blockage identification device, electronic device and its corresponding units described above can be referred to the description of the package blockage identification method in any embodiment, and will not be repeated here.

[0243] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions, or by instructions controlling related hardware. These instructions can be stored in a readable storage medium and loaded and executed by a processor.

[0244] Therefore, embodiments of this application provide a readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of the package blockage identification method in any embodiment of this application. For specific operations, please refer to the description of the package blockage identification method in any embodiment, which will not be repeated here.

[0245] The readable storage medium may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0246] Since the instructions stored in the readable storage medium can execute the steps in the package blockage identification method in any embodiment of this application, the beneficial effects that the package blockage identification method in any embodiment of this application can achieve can be realized, as detailed in the preceding description, and will not be repeated here.

[0247] The above provides a detailed description of a package blockage identification method, apparatus, storage medium, and electronic device provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A package jam identification method characterized by, include: Acquire the target image; Based on the target image, the target package blockage status of the target transport component is determined, and the package blockage status is the predicted blockage status of the package being transported on the target transport component. The historical package congestion status of the target transportation component is obtained. The historical package congestion status is the predicted congestion status of the packages transported on the target transportation component before the shooting time point corresponding to the target image. The historical package congestion status and the target package congestion status are input into a preset state machine model, and the target congestion identification result of the target transportation component is determined according to the preset state transition conditions.

2. The package jam identification method of claim 1, wherein, Determining the target package blockage status of the target transport component based on the target image includes: The target image is input into a preset blockage detection model to obtain the image features of the target image; The image features are processed by the blockage detection model to predict the blockage status of the target package of the target transport component in the target image.

3. The package jam identification method of claim 2, wherein, The preset blockage detection model is obtained through the following steps: Acquire training data, wherein the training data includes training images and a first sample blockage state of the first transport component in each training image; Each of the training images is input into the initial blockage detection model to obtain the first package blockage state of the first transport component in each of the training images; Based on the first package blockage state in each of the training images, a target training image is selected from each of the training images; Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model.

4. The package jam identification method of claim 3, wherein, The step of selecting a target training image from each of the training images based on the first package blockage state in each of the training images includes: The confidence level of the first package blockage state in each of the training images is matched with a preset confidence level range, and the training images with the corresponding confidence level within the confidence level range are set as target training images. And / or, The training images are arranged according to their respective timestamps to obtain a training image sequence. The first package blockage state of adjacent training images in the training image sequence is compared to obtain the target training image in which the first package blockage state changes abruptly. And / or, From each of the training images, select the corresponding image to be identified where the first package blockage state is blocked, identify the package blockage area in the image to be identified, and obtain the area of ​​the package blockage area. If the area of ​​the package blockage area is less than or equal to a preset area threshold, then set the image to be identified as the target training image.

5. The package jam identification method of claim 3, wherein, The acquisition of training data includes: Acquire training images; The initial annotation state of each training image is obtained by predicting the state of each training image using a preset state annotation model. The initial annotation states are corrected to determine the first sample blockage state of the first transport component in each training image, thereby obtaining training data.

6. The package jam identification method of claim 3, wherein, The step of adjusting the parameters in the initial blockage detection model based on the first sample blockage state and the first package blockage state in the target training image to obtain the preset blockage detection model includes: Based on the first sample blockage state and the first package blockage state in the target training image, the parameters in the initial blockage detection model are adjusted to obtain the preset blockage detection model. The preset test image is input into the blockage detection model to be tested to obtain the second package blockage state of the second transport component in the test image; Based on the second sample blockage state of the second transport component and the second package blockage state in the test image, determine the model evaluation value of the blockage detection model to be tested; If the model evaluation value meets the preset training termination condition, then the blockage detection model to be tested is set to the preset blockage detection model.

7. The package jam identification method of claim 3, wherein, The step of selecting a target training image from each of the training images based on the first package blockage state in each of the training images includes: Based on the first package blockage state corresponding to each training image, select the target training image to be screened from each training image; Obtain the similarity between any two images in the training images of the target to be screened; If the similarity is greater than a preset similarity threshold, one of the two images is removed from the target training images to be screened, and the target training image is obtained.

8. The package jam identification method of any one of claims 1-7, wherein, The package blockage identification method also includes: The target blockage identification result is output through the preset target terminal.

9. A package jam detection apparatus, characterized by, include: The first acquisition unit is used to acquire the target image; The determining unit is used to determine the blockage status of the target package of the target transport component based on the target image; The package congestion status is the predicted congestion status of the packages being transported on the target transport component; The second acquisition unit is used to acquire the historical package congestion status of the target transportation component. The historical package congestion status is the predicted congestion status of the packages transported on the target transportation component before the shooting time point corresponding to the target image. The identification unit is used to input the historical package blockage status and the target package blockage status into a preset state machine model, and determine the target blockage identification result of the target transportation component according to preset state transition conditions.

10. An electronic device, comprising: The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the package blockage identification method as described in any one of claims 1 to 8.

11. A readable storage medium, characterized by, The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the package blockage identification method according to any one of claims 1 to 8.

Images