CN112257493B - Method, device, equipment and storage medium for identifying abnormal sorting of articles - Google Patents
Method, device, equipment and storage medium for identifying abnormal sorting of articles Download PDFInfo
- Publication number
- CN112257493B CN112257493B CN202010905162.5A CN202010905162A CN112257493B CN 112257493 B CN112257493 B CN 112257493B CN 202010905162 A CN202010905162 A CN 202010905162A CN 112257493 B CN112257493 B CN 112257493B
- Authority
- CN
- China
- Prior art keywords
- identified
- sorting
- articles
- acquiring
- candidate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/44—Event detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The embodiment of the invention provides a method, a device, equipment and a storage medium for identifying abnormal sorting of articles, wherein articles to be identified, which meet the requirement of abnormal sorting identification, in candidate articles are determined; acquiring a video fragment of an object to be identified in a sorting process; acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clips and a preset abnormal sorting identification model; and outputting a judging result to alarm according to the judging result. In this embodiment, the article that may have abnormal sorting is first selected from the candidate articles as the article to be identified, and then the video segment of the sorting process of the article to be identified is processed in a targeted manner, so as to determine whether the abnormal sorting behavior actually exists, thereby reducing the data size of the identification process, reducing the pressure of the server, reducing the requirement on the processing capability of the server, and identifying the abnormal sorting behavior timely and accurately for alarming or other corresponding processing.
Description
Technical Field
The embodiment of the invention relates to the field of warehouse logistics, in particular to a method, a device, equipment and a storage medium for identifying abnormal sorting of articles.
Background
With the rise of online shopping and logistics industry, the quantity of express bill rises rapidly year by year. In the express delivery process, the articles are required to be picked out from a storage area or a sorting area according to varieties, warehouse-in and warehouse-out sequence and the like, and are classified and concentrated to be stacked, so that the smoothness of the whole logistics network is ensured.
At present, the sorting of express delivery is usually performed manually, and abnormal sorting behaviors of the sorters on the articles, such as throwing, falling, stepping, kicking and the like, may exist, and the abnormal sorting behaviors may damage the articles. In order to avoid abnormal sorting, cameras are usually installed in the field of express sorting, video data of a sorting process are collected through the cameras and uploaded to a cloud server, the cloud server processes the video data, and whether abnormal sorting behaviors exist is identified.
In the prior art, as the sites for express sorting are numerous, a plurality of cameras can be installed on the same site, so that massive video data can be generated, the processing of the massive video data has higher requirements on the processing capacity of a cloud server, and the situation that abnormal sorting behaviors cannot be timely and accurately identified can exist.
Disclosure of Invention
The embodiment of the invention provides a method, a device, equipment and a storage medium for identifying abnormal sorting of articles, which are used for solving the problem of insufficient processing capacity of a server in the process of identifying abnormal sorting of articles and improving the efficiency and accuracy of identifying abnormal sorting.
In a first aspect, an embodiment of the present invention provides a method for identifying abnormal sorting of articles, including:
determining articles to be identified, which meet the abnormal sorting and identifying requirements, in the candidate articles;
acquiring a video segment of the sorting process of the object to be identified;
acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video fragment and a preset abnormal sorting identification model;
and outputting the judging result to alarm according to the judging result.
In a second aspect, an embodiment of the present invention provides an apparatus for identifying abnormal sorting of articles, including:
the processing module is used for determining the articles to be identified, which meet the abnormal sorting and identifying requirements, in the candidate articles;
the acquisition module is used for acquiring video clips of the sorting process of the objects to be identified;
the identification module is used for acquiring a judgment result of whether the object to be identified is abnormally sorted or not according to the video fragment and a preset abnormal sorting identification model;
And the output module is used for outputting the judging result so as to alarm according to the judging result.
In a third aspect, an embodiment of the present invention provides a computer apparatus, including:
a memory for storing computer-executable instructions;
a processor for executing computer-executable instructions stored in the memory to implement the method as described in the first aspect.
In a fourth aspect, embodiments of the present invention provide a computer-readable storage medium having stored therein computer-executable instructions for performing the method according to the first aspect when executed by a processor.
The identification method, the device, the equipment and the storage medium for abnormal sorting of the articles provided by the embodiment of the invention are characterized in that the articles to be identified, which meet the requirement of abnormal sorting identification, in the candidate articles are determined; acquiring a video fragment of an object to be identified in a sorting process; acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clips and a preset abnormal sorting identification model; and outputting a judging result to alarm according to the judging result. In this embodiment, the article that may have abnormal sorting is first selected from the candidate articles as the article to be identified, and then the video segment of the sorting process of the article to be identified is processed in a targeted manner, so as to determine whether the abnormal sorting behavior actually exists, thereby reducing the data size of the identification process, reducing the pressure of the server, reducing the requirement on the processing capability of the server, and identifying the abnormal sorting behavior timely and accurately for alarming or other corresponding processing.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure.
Fig. 1 is a system schematic diagram of a method for identifying abnormal sorting of articles according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for identifying abnormal sorting of articles according to an embodiment of the present invention;
FIG. 3 is a flow chart of a method for identifying abnormal sorting of articles according to another embodiment of the present invention;
FIG. 4 is a flowchart of a method for identifying abnormal sorting of articles according to another embodiment of the present invention;
FIG. 5 is a schematic diagram of a model frame of a dual-layer kernel extreme learning machine in an identification method for abnormal sorting of articles according to an embodiment of the present invention;
FIG. 6 is a block diagram of an apparatus for identifying abnormal sorting of articles according to an embodiment of the present invention;
fig. 7 is a block diagram of a computer device according to an embodiment of the present invention.
Specific embodiments of the present disclosure have been shown by way of the above drawings and will be described in more detail below. These drawings and the written description are not intended to limit the scope of the disclosed concepts in any way, but rather to illustrate the disclosed concepts to those skilled in the art by reference to specific embodiments.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
At present, the sorting of the express delivery is usually carried out manually, and abnormal sorting behaviors of the sorters on the articles, such as throwing, falling, stepping, kicking and the like, may exist. In order to avoid abnormal sorting, cameras are usually installed in the field of express sorting, video data of a sorting process are collected through the cameras and uploaded to a cloud server, the cloud server processes the video data, and whether abnormal sorting behaviors exist is identified. Because the sites for express sorting are numerous, a plurality of cameras can be installed on the same site, so that massive video data can be generated, the processing of the massive video data brings high requirements on the processing capacity of the cloud server, and the situation that abnormal sorting behaviors cannot be timely and accurately identified can exist.
According to the technical problems, before the identification processing is carried out on the video data, the articles possibly having abnormal sorting are screened out from the alternative articles to serve as articles to be identified, then video clips of sorting processes corresponding to the articles to be identified are obtained from the massive video data, and further the video clips of the sorting processes of the articles to be identified are processed only to judge whether abnormal sorting behaviors exist truly, so that the massive video data are not required to be processed completely, the data volume is reduced, the condition that the processing capacity of a server is insufficient is avoided, the requirement on the processing capacity of the server is reduced, and the abnormal sorting behaviors can be identified timely and accurately to carry out alarming or other corresponding processing.
The method for identifying abnormal sorting of articles according to the embodiment of the present invention may be applied to a system as shown in fig. 1, where the system specifically includes a server 101, a database 102 and a camera 103, where the camera 103 is disposed at a sorting site and is used for collecting video data during sorting of articles and uploading the video data to the database 102, the server 101 may first determine articles to be identified meeting the requirement of abnormal sorting identification, obtain video segments of the sorting process of the articles to be identified from the database 102, and further obtain a determination result of whether the articles to be identified are abnormal sorted according to the video segments and a preset abnormal sorting identification model, and if it is determined that the articles to be identified are abnormal sorted, the server may display alarm information on the display screen 105 or send the alarm information to the target terminal 104 for alarming, so as to facilitate corresponding measures to process the abnormal sorting actions or prevent the abnormal sorting actions from appearing again.
The following describes the technical scheme of the present invention and how the technical scheme of the present application solves the above technical problems in detail with specific embodiments. The following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. Embodiments of the present invention will be described below with reference to the accompanying drawings.
Fig. 2 is a flowchart of an identification method for abnormal sorting of articles according to an embodiment of the present invention. The embodiment provides a method for identifying abnormal sorting of articles, an execution main body of which can be computer equipment such as a server, and the like, and the method for identifying abnormal sorting of articles comprises the following specific steps:
s201, determining the articles to be identified, which meet the abnormal sorting and identifying requirements, in the candidate articles.
In this embodiment, it may be first determined from a plurality of candidate articles, which candidate articles may be abnormally sorted, or which candidate articles may be damaged when being abnormally sorted, or which candidate articles are more concerned, and the candidate articles are used as articles to be identified, and then the abnormal sorting identification is performed based on the video clips of the sorting process. That is, in this embodiment, the abnormal sorting and identifying requirement is preset, and the subsequent abnormal sorting and identifying process is performed on the candidate articles meeting the abnormal sorting and identifying requirement. Of course, the abnormal sorting identification requirement in the present embodiment is not limited to the above example, and other abnormal sorting identification requirements may be used.
S202, acquiring video clips of the sorting process of the objects to be identified.
In this embodiment, only for the articles to be identified that meet the abnormal sorting identification requirement, the video clips of the sorting process are acquired. Specifically, since the camera shoots in real time when shooting the sorting sites, the obtained video data may be a whole video data stream, and thus it is necessary to intercept video clips of the sorting process for the objects to be identified from the video data.
Alternatively, a time period during which the sorting process is performed on the article to be identified may be first obtained; and then according to the time period, intercepting video fragments of the sorting process of the object to be identified from real-time video data of the scene where the object to be identified is located.
Wherein, optionally, when acquiring the time period of the sorting process of the object to be identified, the first time when the bar code on the object to be identified is scanned before the sorting operation can be acquired; acquiring the advancing time length of the object to be identified on the conveyor belt according to the speed information of the conveyor belt; and then the time period for sorting the objects to be identified can be determined according to the first time and the travelling time.
S203, acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clips and a preset abnormal sorting identification model.
In this embodiment, after a video segment of a sorting process for an object to be identified is obtained, the video segment may be identified in an action manner, and in this embodiment, the video segment is input into a preset abnormal sorting identification model, and the identification of the action manner is performed through the preset abnormal sorting identification model, so as to obtain a determination result of whether the object to be identified is abnormally sorted. In this embodiment, the preset abnormal sorting recognition model may be a deep learning model capable of performing action behavior recognition, for example, an action recognition algorithm model of a double-layer kernel extreme learning machine and a deep learning technology, and may of course be other models.
S204, outputting the judging result to alarm according to the judging result.
In this embodiment, after the judgment result is obtained, the judgment result may be output, for example, output to a display screen or sent to the target terminal device, so that an alarm may be performed according to the judgment result. Optionally, only when the judgment result is that the object to be identified is abnormal sorted, outputting the judgment result to prompt a processor to process the abnormal sorting event.
According to the identification method for abnormal sorting of the articles, the articles to be identified, which meet the requirement for abnormal sorting identification, in the candidate articles are determined; acquiring a video fragment of an object to be identified in a sorting process; acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clips and a preset abnormal sorting identification model; and outputting a judging result to alarm according to the judging result. In this embodiment, the article that may have abnormal sorting is first selected from the candidate articles as the article to be identified, and then the video segment of the sorting process of the article to be identified is processed in a targeted manner, so as to determine whether the abnormal sorting behavior actually exists, thereby reducing the data size of the identification process, reducing the pressure of the server, reducing the requirement on the processing capability of the server, and identifying the abnormal sorting behavior timely and accurately for alarming or other corresponding processing.
On the basis of any of the above embodiments, as shown in fig. 3, the determining the to-be-identified item satisfying the abnormal sorting identification requirement in the candidate item in S201 in this embodiment may specifically include:
s301, acquiring at least one attribute tag of a candidate item;
s302, acquiring a weight corresponding to each attribute tag, and acquiring the total weight of the candidate object according to the weight corresponding to each attribute tag;
and S303, if the total weight of the candidate articles meets a preset range, determining that the candidate articles are the articles to be identified.
In this embodiment, at least one attribute tag may be configured in advance for each candidate item, and may include attribute tags such as package, material, value, and the like, specifically, for example, tags such as "canned", "bottled", "liquid", "fragile", "valuables", "high reliability", and the like, and each attribute tag corresponds to a weight, for example, a weight of "canned" is 0.3, a weight of "bottled" is 0.6, a weight of "high reliability" is-0.2, and the higher the weight is, the higher the risk of damage to the item is, or whether it is more required to be paid attention to abnormal sorting or not. It should be noted that, the weight corresponding to the attribute tag may be fixed, or may be adjusted according to actual situations, for example, if an article having a "bottled" attribute tag in a certain scene is more easily damaged, the weight corresponding to the "bottled" attribute tag may be increased.
In this embodiment, after the weight corresponding to each attribute tag is obtained, the total weight of the candidate item may be obtained according to the weight corresponding to each attribute tag, that is, the total weight is the sum of the weights corresponding to the attribute tags, and then the total weight of the candidate item is compared with the predetermined range. In this embodiment, if the total weight of the candidate articles meets the predetermined range, the candidate articles are considered to meet the abnormal sorting and identifying requirement, and then the candidate articles are determined to be articles to be identified, and the subsequent abnormal sorting and identifying based on the video clips of the sorting process is required.
Based on the above embodiment, considering that the damage risk of different types of articles is different for the same attribute tag, for example, bottled articles, liquid articles such as wines and the like are more easily damaged or more influenced than solid articles such as candies and the like, and whether the articles are abnormally sorted or not needs to be concerned, the articles can be classified according to the basic types of the articles, and the appropriate weight values can be configured for the same attribute tag of different types, and may not be the same.
Alternatively, the articles may be classified into three classes, for example, a first class may include a plurality of classes such as appliances, furniture, stationery and books, sports and fitness products, toys, foods, etc., a second class may include a plurality of classes such as drinks, raw and fresh products, grain and oil, etc., a third class may include a class such as carbonated drinks, general drinks, beer, red wine, white wine, etc., and the weights of the respective attribute tags may be configured for each class of articles in the third class, wherein the same attribute tag in the same class in the third class has the same weight, and the weights of the same attribute tags may be the same or different in other cases.
Further, as shown in fig. 4, the step of obtaining the weight corresponding to each attribute tag in step S302 may specifically include:
s3021, obtaining initial weights of the attribute tags according to the categories of the candidate items.
In this embodiment, since the respective weights are configured for the respective attribute tags of each category, the category of the candidate item may be determined first, and then the weight corresponding to each attribute tag of the candidate item may be obtained as the initial weight according to the category of the candidate item.
And S3022, if the candidate item is determined to be damaged, acquiring a first compensation value.
In this embodiment, when an object is damaged, a detection personnel usually adopts a stricter detection measure to treat the object, so if it is determined that the candidate object is damaged, a first compensation value is obtained to compensate for the initial weight of each attribute tag, that is, to increase the attention degree of the object or the damage risk of the object.
Wherein, optionally, the first compensation value may be obtained according to the damage degree parameter, the damage time and the preset time attenuation parameter of the candidate item
In this embodiment, the damage risk of the article may be increased according to different damage degrees, where the higher the damage degree is, the greater the damage risk is, and the greater the first compensation value is; considering that the attention degree of the article cannot be kept high all the time, especially when abnormal sorting is detected and certain measures are taken, the damage probability of the article is gradually reduced, so that the first compensation value can be correspondingly gradually reduced, namely, the first compensation value can be attenuated according to the damage time, and a time attenuation parameter can be preset.
More specifically, as an example, the damage level may be divided into three levels, such as light damage, medium damage, and severe damage, with the corresponding damage level parameter noted as α i (α i E (0, 1)), where slight damage is alpha 1 Moderate damage to alpha 2 Serious damage to alpha 3 . Each initial attribute of candidate item a is labeled beta j (j represents the attribute tag number of the item) the preset time decay parameter is z, and the time interval between the current time and the damaged time is t-ts, so the first compensation value can be calculated by the following formula:
first compensation value=α i ×e -z(t-ts) 。
And S3023, compensating the initial weight of each attribute tag according to the first compensation value to obtain the weight corresponding to each attribute tag.
In this embodiment, after the first compensation value is obtained, the initial weight of each attribute tag may be compensated, so as to obtain the weight corresponding to each attribute tag.
Specifically, for the above example, the attribute label j corresponds to the weight χ j The method comprises the following steps:
χ j ＝β j +α i ×e -z(t-ts)
through the updating process, the abnormal sorting identification process can consider the factors of the article types, so that whether the articles of different types are abnormal sorted or not can be accurately identified, and the situation that omission or excessive strictness occurs when the articles to be identified are determined to carry out subsequent video recognition is avoided, namely, certain articles which are not determined to be the articles to be identified in the video recognition process, or the articles which are not determined to be the articles to be identified in the video recognition process are avoided. It should be noted that, if the candidate article is not damaged, the above-described compensation process may not be performed.
On the basis of the above embodiment, if a certain candidate article is damaged, other articles with the same attribute labels may be damaged as well, so that the damage risk needs to be increased correspondingly. Thus, in this embodiment, the first compensation value may be multiplexed onto the same attribute tags of other items.
However, considering that the possible damage risks of different types of articles are different, for example, beer with a bottled attribute tag is damaged, the weight of the bottled attribute tag of the beer type can be improved, and the damage risk of candies with bottled attribute tags is not as high as that of beer type, if the candies with bottled attribute tags are obviously unreasonable to be compensated by the same first compensation value, the type relation among articles can be considered when the damage risk is improved, for example, if other articles B and the candidate article A belong to the same type, the first compensation value is directly multiplexed onto the same attribute tag of the other articles B, namely, the weight of the compensated same attribute tag of the other articles B is consistent with the weight value of the same attribute tag of the candidate article A; if the other article B and the candidate article A belong to different categories, a preset coefficient can be configured.
For example, in the above scenario in which the items are classified into three classes, if the other item B and the candidate item a belong to the same third class, the preset coefficient is 1; if the other article B and the candidate article A belong to the same second level but are different from the third level, the preset coefficient is 0.8; if the other article B and the candidate article A belong to the same first level but are different from the second level, the preset coefficient is 0.5; if the other article B and the candidate article A belong to different first levels, the preset coefficient is 0.2. When compensating the initial weight of the label with the same attribute, multiplying the first compensation value by the preset coefficient, and adding the first compensation value with the initial weight.
That is, in this embodiment, for all other articles having at least one attribute tag identical to the candidate article, the initial weight of the attribute tag identical to the candidate article may be updated according to the first compensation value and a preset coefficient, where the preset coefficient is obtained according to a category relationship between the other articles and the candidate article.
It should be noted that, when the initial weight of each attribute tag is obtained according to the category of the candidate item in S3021, it is considered that damage to the candidate item in the category may affect the corresponding weight of the tag of the candidate item in the category, so that the initial weight of each attribute tag needs to adopt the latest weight, that is, the updated weight.
On the basis of any of the above embodiments, evaluation criteria considering the magnitude of the operation action in practice may be different. Taking paper towels as an example, in order to ensure sorting efficiency, the action with larger swing amplitude does not damage articles, otherwise, if bottled white spirit is thrown with smaller swing amplitude, articles are damaged. Therefore, in this embodiment, a plurality of different preset ranges are set, where each preset range corresponds to one preset abnormal sorting recognition model, and each preset abnormal sorting recognition model can recognize different motion amplitudes, so that abnormal sorting recognition by using different evaluation standards can be implemented.
Alternatively, a plurality of different preset ranges may be provided, for example, divided into four according to a stepwise manner: class A, class B, class C, class D, are specifically shown in Table 1 below, wherein δ k Is the total weight of candidate item k.
TABLE 1
The A, B, C corresponds to different preset abnormal sorting recognition models, and adopts different recognition standards, and if the D is satisfied, the abnormal sorting recognition based on the video segment is not performed, that is, the D does not correspond to the preset abnormal sorting recognition model. Where φ and ε are related parameters used to define a predetermined range and can be empirically set.
Further, the step S203 of obtaining a determination result of whether the object to be identified is abnormal sorted according to the video clip and a preset abnormal sorting identification model may specifically include:
determining a target preset range in which the total weight of the object to be identified is located, inputting the video segment into a preset abnormal sorting identification model corresponding to the target preset range, and obtaining a judging result of whether the object to be identified is abnormally sorted.
In this embodiment, it may be determined in which preset range the total weight of the object to be identified is located, where the preset range is the target preset range, and further it may be determined that a preset abnormal sorting identification model corresponding to the target preset range is adopted, and the video clip is input into the preset abnormal sorting identification model corresponding to the target preset range, so as to obtain a determination result of whether the object to be identified is abnormal sorted.
It should be noted that, because each preset abnormal sorting recognition model corresponds to each preset range, each preset abnormal sorting recognition model needs to be trained independently, each model needs to acquire corresponding training data, the training data is a sorting process video segment of the marked articles with the total weight in each preset range, and the corresponding preset abnormal sorting recognition model is trained according to the training data. The training process is not described in detail here.
On the basis of any one of the above embodiments, the preset abnormal sorting recognition model is a double-layer kernel extreme learning machine model. The double-layer kernel extreme learning machine model can adopt a deep learning technology to realize recognition of actions, and whether abnormal sorting is performed or not can be recognized after video clips are input. The model frame diagram of the double-layer core extreme learning machine is shown in fig. 5.
More specifically, step S203 of obtaining a determination result of whether the object to be identified is abnormal sorted according to the video clip and a preset abnormal sorting identification model may specifically include:
extracting a deep learning feature and a manual feature from the video segment through a first layer of the double-layer kernel extreme learning machine model, fusing the deep learning feature and the manual feature to obtain a fused feature, and outputting the deep learning feature, the manual feature and a prediction score corresponding to the fused feature;
and classifying according to the deep learning features, the manual features and the prediction scores corresponding to the fusion features through a second layer of the double-layer kernel extreme learning machine model, and outputting a judging result of whether the object to be identified is abnormally sorted or not.
In this embodiment, the first layer of the two-layer kernel extreme learning machine model may obtain a deep learning feature kernel, a manual feature kernel, and a fusion feature kernel.
For the feature kernel of deep learning, all motion sequences of one image frame need to be acquired, the feature of the video frame time dimension and the feature of the video frame three-color channel dimension are extracted through a C3D (Convolitional 3D) convolution network model, feature fusion is obtained through a Convolution Neural Network (CNN), and specifically, the time information can be modeled through 3D convolution and 3D pooling operation, and then the feature kernel of deep learning can be obtained through a full-connection layer and a Softmax layer. It should be noted that, in this embodiment, other deep learning models may be used to extract the feature kernels of the deep learning, which will not be described in detail here.
For the manual feature kernel, an IDT (Improved Dense Trajectories, modified density track) descriptor of the video is extracted through an IDT algorithm model, wherein the IDT algorithm model is a very classical algorithm model in the field of behavior recognition, and an algorithm basic framework comprises: densely sampling feature points, tracking a feature point track, extracting features based on the track, encoding features and classifying a classifier, wherein the IDT descriptor comprises: trace line, describe static features, pixel absolute motion features, pixel relative motion features. The IDT features are encoded using Fisher vectors. The Fisher vector is processed with the L2 norm, and the linear kernel for each descriptor is computed.
And for the fused feature kernels, calculating the kernel matrix average value of the deep learning feature kernels and the manual feature kernels to obtain the fused feature kernels.
After three feature kernels are obtained, a kernel extreme learning machine is adopted to calculate the prediction scores of different feature kernels.
The second layer of the double-layer kernel extreme learning machine model is a classifier, and after the prediction scores of different feature kernels are input into the classifier, the prediction scores can be mapped to a final judgment result of whether the object to be identified is abnormally sorted.
According to the method for identifying abnormal sorting of the articles, the articles possibly existing abnormal sorting are firstly screened out from the alternative articles to serve as articles to be identified, then the video clips of the sorting process of the articles to be identified are processed in a targeted mode, whether abnormal sorting behaviors exist actually or not is judged, the data size of the identification process is reduced, the pressure of a server is reduced, the requirement on the processing capacity of the server is also reduced, and the abnormal sorting behaviors can be timely and accurately identified to give an alarm or perform other corresponding processing. In addition, the article category is considered, and the preset abnormal sorting recognition models corresponding to a plurality of different preset ranges are set, so that different recognition standards can be adopted for different articles, error recognition caused by the adoption of a single recognition standard is avoided, and the accuracy of abnormal sorting recognition is further improved.
Fig. 6 is a block diagram of an identification device for abnormal sorting of articles according to an embodiment of the present invention. The device for identifying abnormal sorting of articles provided in this embodiment may execute a processing flow provided by the embodiment of the method for identifying abnormal sorting of articles, as shown in fig. 6, where the device 600 for identifying abnormal sorting of articles includes a processing module 601, an obtaining module 602, an identifying module 603, and an output module 604.
A processing module 601, configured to determine an item to be identified that meets an abnormal sorting identification requirement in the candidate items;
an obtaining module 603, configured to obtain a video segment of the sorting process for the object to be identified;
the recognition module 603 is configured to obtain a determination result of whether the object to be recognized is abnormal sorted according to the video clip and a preset abnormal sorting recognition model;
and the output module 604 is configured to output the determination result, so as to alarm according to the determination result.
On the basis of the above embodiment, the processing module 601 is configured, when determining an item to be identified that meets the requirement of abnormal sorting identification in candidate items, to:
acquiring at least one attribute tag of a candidate item;
acquiring a weight corresponding to each attribute tag, and acquiring the total weight of the candidate object according to the weight corresponding to each attribute tag;
And if the total weight of the candidate articles meets a preset range, determining that the candidate articles are the articles to be identified.
On the basis of any one of the foregoing embodiments, when the processing module 601 obtains a weight corresponding to each attribute tag, the processing module is configured to:
acquiring an initial weight of each attribute tag according to the category of the candidate item;
if the candidate article is determined to be damaged, a first compensation value is obtained;
and compensating the initial weight of each attribute tag according to the first compensation value to obtain the weight corresponding to each attribute tag.
On the basis of any of the foregoing embodiments, the processing module 601 is configured to, when acquiring the first compensation value:
and acquiring the first compensation value according to the damage degree parameter, the damage time and the preset time attenuation parameter of the candidate object.
On the basis of any of the foregoing embodiments, the processing module 601 is further configured to, after obtaining the first compensation value:
and updating the initial weight of the label with the same attribute for all other articles with at least one label with the same attribute as the candidate article according to the first compensation value and a preset coefficient, wherein the preset coefficient is obtained according to the category relation between the other articles and the candidate article.
On the basis of any one of the above embodiments, the predetermined range includes a plurality of different preset ranges, wherein each preset range corresponds to a preset abnormal sorting identification model;
the identifying module 603 is configured to, when acquiring a determination result of whether the object to be identified is abnormal sorted according to the video clip and a preset abnormal sorting identification model:
determining a target preset range in which the total weight of the object to be identified is located, inputting the video segment into a preset abnormal sorting identification model corresponding to the target preset range, and obtaining a judging result of whether the object to be identified is abnormally sorted.
On the basis of any one of the above embodiments, the preset abnormal sorting recognition model is a double-layer kernel extreme learning machine model;
the identifying module 603 is configured to, when acquiring a determination result of whether the object to be identified is abnormal sorted according to the video clip and a preset abnormal sorting identification model:
extracting a deep learning feature and a manual feature from the video segment through a first layer of the double-layer kernel extreme learning machine model, fusing the deep learning feature and the manual feature to obtain a fused feature, and outputting the deep learning feature, the manual feature and a prediction score corresponding to the fused feature;
And classifying according to the deep learning features, the manual features and the prediction scores corresponding to the fusion features through a second layer of the double-layer kernel extreme learning machine model, and outputting a judging result of whether the object to be identified is abnormally sorted or not.
On the basis of any one of the foregoing embodiments, the apparatus further includes a training module configured to:
respectively acquiring corresponding training data for a preset abnormal sorting identification model corresponding to each preset range, wherein the training data are sorting process video clips of the marked articles with total weights in each preset range;
and training the corresponding preset abnormal sorting recognition model according to the training data.
On the basis of any one of the foregoing embodiments, the obtaining module 603 is configured to, when obtaining a video segment of the sorting process for the object to be identified:
acquiring a time period of a sorting process of the object to be identified;
and according to the time period, capturing video clips of the sorting process of the object to be identified from real-time video data of the scene where the object to be identified is located.
On the basis of any one of the foregoing embodiments, the obtaining module 603 is configured to, when obtaining a time period for performing a sorting process on the object to be identified:
Acquiring the first time when the bar codes on the objects to be identified are scanned before sorting operation;
acquiring the advancing time of the object to be identified on the conveyor belt according to the speed information of the conveyor belt;
and determining a time period for sorting the objects to be identified according to the first time and the travelling time.
The device for identifying abnormal sorting of articles provided in the embodiment of the present invention may be specifically used to execute the method embodiments provided in fig. 2 to 4, and specific functions are not described herein.
According to the identification device for abnormal sorting of the articles, provided by the embodiment of the invention, the articles to be identified meeting the requirement of abnormal sorting identification in the candidate articles are determined; acquiring a video fragment of an object to be identified in a sorting process; acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clips and a preset abnormal sorting identification model; and outputting a judging result to alarm according to the judging result. In this embodiment, the article that may have abnormal sorting is first selected from the candidate articles as the article to be identified, and then the video segment of the sorting process of the article to be identified is processed in a targeted manner, so as to determine whether the abnormal sorting behavior actually exists, thereby reducing the data size of the identification process, reducing the pressure of the server, reducing the requirement on the processing capability of the server, and identifying the abnormal sorting behavior timely and accurately for alarming or other corresponding processing.
Fig. 7 is a schematic structural diagram of a computer device according to an embodiment of the present invention. The computer device provided by the embodiment of the invention can execute the processing flow provided by the control method embodiment of the laundry device, as shown in fig. 7, the computer device 70 comprises a memory 71, a processor 72 and a computer program; wherein a computer program is stored in the memory 71 and configured to be executed by the processor 72 in the control method of the laundry appliance as described in the above embodiments. The computer device 70 may also have a communication interface 73 for receiving control instructions.
The computer device of the embodiment shown in fig. 7 may be used to implement the technical solution of the above-mentioned method embodiment, and its implementation principle and technical effects are similar, and are not described here again.
In addition, the present embodiment also provides a computer-readable storage medium having stored thereon a computer program that is executed by a processor to implement the control method of the laundry device described in the above embodiments.
In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.
The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform part of the steps of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional modules is illustrated, and in practical application, the above-described functional allocation may be performed by different functional modules according to needs, i.e. the internal structure of the apparatus is divided into different functional modules to perform all or part of the functions described above. The specific working process of the above-described device may refer to the corresponding process in the foregoing method embodiment, which is not described herein again.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the embodiments of the present invention, and are not limited thereto; although embodiments of the present invention have been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention.
Claims (22)
1. An identification method for abnormal sorting of articles, comprising the steps of:
Determining articles to be identified, which meet the abnormal sorting and identifying requirements, in the candidate articles; wherein, abnormal sorting is a sorting behavior which has the risk of causing damage to the articles;
acquiring a video segment of the sorting process of the object to be identified;
acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video fragment and a preset abnormal sorting identification model;
and outputting the judging result to alarm according to the judging result.
2. The method of claim 1, wherein determining the item to be identified among the candidate items that meets the abnormal sort identification requirement comprises:
acquiring at least one attribute tag of a candidate item;
acquiring a weight corresponding to each attribute tag, and acquiring the total weight of the candidate object according to the weight corresponding to each attribute tag;
and if the total weight of the candidate articles meets a preset range, determining that the candidate articles are the articles to be identified.
3. The method of claim 2, wherein the obtaining the weight corresponding to each attribute tag comprises:
acquiring an initial weight of each attribute tag according to the category of the candidate item;
If the candidate article is determined to be damaged, a first compensation value is obtained;
and compensating the initial weight of each attribute tag according to the first compensation value to obtain the weight corresponding to each attribute tag.
4. A method according to claim 3, wherein said obtaining a first compensation value comprises:
and acquiring the first compensation value according to the damage degree parameter, the damage time and the preset time attenuation parameter of the candidate object.
5. The method of claim 3, wherein after the obtaining the first compensation value, further comprising:
and updating the initial weight of the label with the same attribute for all other articles with at least one label with the same attribute as the candidate article according to the first compensation value and a preset coefficient, wherein the preset coefficient is obtained according to the category relation between the other articles and the candidate article.
6. The method of any one of claims 2-5, wherein the predetermined range comprises a plurality of different predetermined ranges, wherein each predetermined range corresponds to a predetermined anomaly sorting recognition model;
the step of obtaining a judging result of whether the object to be identified is abnormally sorted according to the video clips and a preset abnormal sorting identification model comprises the following steps:
Determining a target preset range in which the total weight of the object to be identified is located, inputting the video segment into a preset abnormal sorting identification model corresponding to the target preset range, and obtaining a judging result of whether the object to be identified is abnormally sorted.
7. The method of claim 6, wherein the pre-set anomaly sorting recognition model is a dual-layer kernel extreme learning machine model;
the step of obtaining a judging result of whether the object to be identified is abnormally sorted according to the video clips and a preset abnormal sorting identification model comprises the following steps:
extracting a deep learning feature and a manual feature from the video segment through a first layer of the double-layer kernel extreme learning machine model, fusing the deep learning feature and the manual feature to obtain a fused feature, and outputting the deep learning feature, the manual feature and a prediction score corresponding to the fused feature;
and classifying according to the deep learning features, the manual features and the prediction scores corresponding to the fusion features through a second layer of the double-layer kernel extreme learning machine model, and outputting a judging result of whether the object to be identified is abnormally sorted or not.
8. The method as recited in claim 6, further comprising:
respectively acquiring corresponding training data for a preset abnormal sorting identification model corresponding to each preset range, wherein the training data are sorting process video clips of the marked articles with total weights in each preset range;
and training the corresponding preset abnormal sorting recognition model according to the training data.
9. The method of claim 1, wherein the acquiring a video segment of the sorting process of the item to be identified comprises:
acquiring a time period of a sorting process of the object to be identified;
and according to the time period, capturing video clips of the sorting process of the object to be identified from real-time video data of the scene where the object to be identified is located.
10. The method of claim 9, wherein the acquiring a time period for performing a sorting process on the item to be identified comprises:
acquiring the first time when the bar codes on the objects to be identified are scanned before sorting operation;
acquiring the advancing time of the object to be identified on the conveyor belt according to the speed information of the conveyor belt;
And determining a time period for sorting the objects to be identified according to the first time and the travelling time.
11. An apparatus for identifying abnormal sorting of articles, comprising:
the processing module is used for determining the articles to be identified, which meet the abnormal sorting and identifying requirements, in the candidate articles; wherein, abnormal sorting is a sorting behavior which has the risk of causing damage to the articles;
the acquisition module is used for acquiring video clips of the sorting process of the objects to be identified;
the identification module is used for acquiring a judgment result of whether the object to be identified is abnormally sorted or not according to the video fragment and a preset abnormal sorting identification model;
and the output module is used for outputting the judging result so as to alarm according to the judging result.
12. The apparatus of claim 11, wherein the processing module, when determining an item to be identified among candidate items that meets an abnormal sort identification requirement, is to:
acquiring at least one attribute tag of a candidate item;
acquiring a weight corresponding to each attribute tag, and acquiring the total weight of the candidate object according to the weight corresponding to each attribute tag;
And if the total weight of the candidate articles meets a preset range, determining that the candidate articles are the articles to be identified.
13. The apparatus of claim 12, wherein the processing module, when obtaining the weight corresponding to each attribute tag, is configured to:
acquiring an initial weight of each attribute tag according to the category of the candidate item;
if the candidate article is determined to be damaged, a first compensation value is obtained;
and compensating the initial weight of each attribute tag according to the first compensation value to obtain the weight corresponding to each attribute tag.
14. The apparatus of claim 13, wherein the processing module, when obtaining the first compensation value, is configured to:
and acquiring the first compensation value according to the damage degree parameter, the damage time and the preset time attenuation parameter of the candidate object.
15. The apparatus of claim 13, wherein the processing module, after obtaining the first compensation value, is further configured to:
and updating the initial weight of the label with the same attribute for all other articles with at least one label with the same attribute as the candidate article according to the first compensation value and a preset coefficient, wherein the preset coefficient is obtained according to the category relation between the other articles and the candidate article.
16. The apparatus of any one of claims 12-15, wherein the predetermined range comprises a plurality of different predetermined ranges, wherein each predetermined range corresponds to a predetermined anomaly sorting recognition model;
the identification module is used for acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clip and a preset abnormal sorting identification model, and is used for:
determining a target preset range in which the total weight of the object to be identified is located, inputting the video segment into a preset abnormal sorting identification model corresponding to the target preset range, and obtaining a judging result of whether the object to be identified is abnormally sorted.
17. The apparatus of claim 16, wherein the pre-set anomaly sorting recognition model is a dual-layer kernel extreme learning machine model;
the identification module is used for acquiring a judging result of whether the object to be identified is abnormally sorted or not according to the video clip and a preset abnormal sorting identification model, and is used for:
extracting a deep learning feature and a manual feature from the video segment through a first layer of the double-layer kernel extreme learning machine model, fusing the deep learning feature and the manual feature to obtain a fused feature, and outputting the deep learning feature, the manual feature and a prediction score corresponding to the fused feature;
And classifying according to the deep learning features, the manual features and the prediction scores corresponding to the fusion features through a second layer of the double-layer kernel extreme learning machine model, and outputting a judging result of whether the object to be identified is abnormally sorted or not.
18. The apparatus of claim 16, further comprising a training module to:
respectively acquiring corresponding training data for a preset abnormal sorting identification model corresponding to each preset range, wherein the training data are sorting process video clips of the marked articles with total weights in each preset range;
and training the corresponding preset abnormal sorting recognition model according to the training data.
19. The apparatus of claim 11, wherein the acquisition module, when acquiring a video clip of the sorting process of the item to be identified, is configured to:
acquiring a time period of a sorting process of the object to be identified;
and according to the time period, capturing video clips of the sorting process of the object to be identified from real-time video data of the scene where the object to be identified is located.
20. The apparatus of claim 19, wherein the acquisition module, when acquiring a time period for sorting the item to be identified, is configured to:
Acquiring the first time when the bar codes on the objects to be identified are scanned before sorting operation;
acquiring the advancing time of the object to be identified on the conveyor belt according to the speed information of the conveyor belt;
and determining a time period for sorting the objects to be identified according to the first time and the travelling time.
21. A computer device, comprising:
a memory for storing computer-executable instructions;
a processor for executing computer-executable instructions stored in the memory to implement the method of any one of claims 1-10.
22. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010905162.5A CN112257493B (en) | 2020-09-01 | 2020-09-01 | Method, device, equipment and storage medium for identifying abnormal sorting of articles |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010905162.5A CN112257493B (en) | 2020-09-01 | 2020-09-01 | Method, device, equipment and storage medium for identifying abnormal sorting of articles |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112257493A CN112257493A (en) | 2021-01-22 |
| CN112257493B true CN112257493B (en) | 2023-08-08 |
Family
ID=74223833
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010905162.5A Active CN112257493B (en) | 2020-09-01 | 2020-09-01 | Method, device, equipment and storage medium for identifying abnormal sorting of articles |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112257493B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106897670A (en) * | 2017-01-19 | 2017-06-27 | 南京邮电大学 | A kind of express delivery violence sorting recognition methods based on computer vision |
| CN107358194A (en) * | 2017-07-10 | 2017-11-17 | 南京邮电大学 | A kind of violence sorting express delivery determination methods based on computer vision |
| CN111126115A (en) * | 2018-11-01 | 2020-05-08 | 顺丰科技有限公司 | Violence sorting behavior identification method and device |
| CN111209847A (en) * | 2020-01-03 | 2020-05-29 | 秒针信息技术有限公司 | Violence sorting identification method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2584683A1 (en) * | 2006-04-20 | 2007-10-20 | Optosecurity Inc. | Apparatus, method and system for screening receptacles and persons |
-
2020
- 2020-09-01 CN CN202010905162.5A patent/CN112257493B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106897670A (en) * | 2017-01-19 | 2017-06-27 | 南京邮电大学 | A kind of express delivery violence sorting recognition methods based on computer vision |
| CN107358194A (en) * | 2017-07-10 | 2017-11-17 | 南京邮电大学 | A kind of violence sorting express delivery determination methods based on computer vision |
| CN111126115A (en) * | 2018-11-01 | 2020-05-08 | 顺丰科技有限公司 | Violence sorting behavior identification method and device |
| CN111209847A (en) * | 2020-01-03 | 2020-05-29 | 秒针信息技术有限公司 | Violence sorting identification method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112257493A (en) | 2021-01-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Nam et al. | Pest detection on traps using deep convolutional neural networks | |
| US20190122072A1 (en) | Reverse neural network for object re-identification | |
| CN105426356B (en) | A kind of target information recognition methods and device | |
| CN109255322A (en) | A kind of human face in-vivo detection method and device | |
| CN111340126A (en) | Article identification method and device, computer equipment and storage medium | |
| CN108711093B (en) | Real object leasing intelligent contract method and system based on deep learning | |
| CN111353549B (en) | Image label verification method and device, electronic equipment and storage medium | |
| CN109993201A (en) | A kind of image processing method, device and readable storage medium storing program for executing | |
| CN108197636A (en) | A kind of paddy detection and sorting technique based on depth multiple views feature | |
| CN105809094A (en) | Bar code identification method based on machine vision | |
| Ribeiro et al. | An adaptable deep learning system for optical character verification in retail food packaging | |
| CN111753290A (en) | Software type detection method and related equipment | |
| CN109766962B (en) | Commodity identification method, storage medium and commodity identification system | |
| CN114937179B (en) | Junk image classification method and device, electronic equipment and storage medium | |
| CN106570015A (en) | Image searching method and device | |
| CN107341139A (en) | Multimedia processing method and device, electronic equipment and storage medium | |
| CN111382605A (en) | Video content auditing method and device, storage medium and computer equipment | |
| CN115752683A (en) | Weight estimation method, system and terminal based on depth camera | |
| CN111488939A (en) | Model training method, classification method, device and equipment | |
| CN112257493B (en) | Method, device, equipment and storage medium for identifying abnormal sorting of articles | |
| EP3425559A2 (en) | Image content moderation | |
| PirahanSiah et al. | Adaptive image segmentation based on peak signal-to-noise ratio for a license plate recognition system | |
| CN112784111A (en) | Video classification method, device, equipment and medium | |
| US20210166028A1 (en) | Automated product recognition, analysis and management | |
| CN116188842A (en) | Logistics illegal operation detection method in strong noise environment based on light-weight countermeasure enhancement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |