CN114220011A

CN114220011A - Goods quantity identification method and device, electronic equipment and storage medium

Info

Publication number: CN114220011A
Application number: CN202111534478.9A
Authority: CN
Inventors: 夏柏麟; 魏献巍
Original assignee: Guangzhou Lanqiao Software Technology Co ltd
Current assignee: Guangzhou Lanqiao Software Technology Co ltd
Priority date: 2021-12-15
Filing date: 2021-12-15
Publication date: 2022-03-22

Abstract

The application discloses a cargo quantity identification method, a cargo quantity identification device, electronic equipment, a storage medium and a program product, and relates to the technical field of image identification and deep learning. The method comprises the following steps: carrying out image feature extraction on the cargo image by using a pre-trained quantity recognition model, wherein the quantity recognition model is provided with a plurality of convolution layers; determining an object candidate frame set in the cargo image according to image characteristics output by at least two layers of convolutional layers in the quantity identification model, wherein the image characteristics at least comprise an object category and positioning coordinates of the object candidate frame; screening and de-duplicating object candidate frames in the object candidate frame set according to the object categories in the image characteristics and the positioning coordinates of the object candidate frames; and taking the number of the target object candidate frames obtained after the screening and the de-weighting as the number of the goods in the goods image. The technical scheme of this application can improve the precision of goods quantity discernment.

Description

Goods quantity identification method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of image recognition and deep learning technologies, and in particular, to a method and an apparatus for recognizing a quantity of goods, an electronic device, a storage medium, and a program product.

Background

In order to prevent the goods from being damaged in the transportation process and being unable to determine responsibility, the goods need to be checked when being loaded and unloaded respectively. For some goods with more rules, such as steel pipes, the goods can be checked by image recognition technology. And taking a photo during loading, taking another photo during unloading, respectively uploading the photos to the logistics management system, checking the number of the steel pipes by the logistics management system by using an image recognition technology, and comparing the number of the steel pipes, wherein if the number of the two photos is inconsistent, the fact that the goods are lost during transportation can be proved.

In the prior art, the method for identifying the quantity of goods is generally as follows: firstly, converting a steel pipe picture into a gray image, enabling the pixel value of the gray image to be between 0 and 255, and then setting a threshold value, enabling the value larger than the threshold value to be 1 and the value smaller than the threshold value to be 0; then obtaining the outer contour of the steel pipe through an edge detection algorithm; removing noise from the value of the outer contour through expansion and corrosion, enhancing the existing contour display, and detecting the circular shape of the section of the steel pipe based on the contour line; and finally, counting the occurrence frequency of the circles so as to obtain the number of the steel pipes in the picture.

However, the above-mentioned method for identifying the number of goods in the prior art depends heavily on whether the threshold is set reasonably, the contour detection accuracy, and other factors, and thus the accuracy of identification is not high.

Disclosure of Invention

The application provides a goods quantity identification method, a device, an electronic device, a storage medium and a program product, which aim to solve the problem of low precision in the goods quantity identification method in the prior art.

In a first aspect, the present application provides a cargo quantity identification method, including:

carrying out image feature extraction on the cargo image by using a pre-trained quantity recognition model, wherein the quantity recognition model is provided with a plurality of convolution layers;

determining an object candidate frame set in the cargo image according to image characteristics output by at least two layers of convolutional layers in the quantity identification model, wherein the image characteristics at least comprise an object category and positioning coordinates of the object candidate frame;

screening and de-duplicating object candidate frames in the object candidate frame set according to the object categories in the image characteristics and the positioning coordinates of the object candidate frames;

and taking the number of the target object candidate frames obtained after the screening and the de-weighting as the number of the goods in the goods image.

In a second aspect, the present application further provides a device for identifying a quantity of goods, the device comprising:

the image feature extraction module is used for extracting image features of the cargo image by utilizing a pre-trained quantity recognition model, wherein the quantity recognition model is provided with a plurality of layers of convolution layers;

an object candidate frame set determining module, configured to determine an object candidate frame set in the cargo image according to image features output by at least two convolutional layers in the quantity recognition model, where the image features at least include an object category and location coordinates of an object candidate frame;

the screening and de-duplication module is used for screening and de-duplicating the object candidate frames in the object candidate frame set according to the object categories in the image characteristics and the positioning coordinates of the object candidate frames;

and the cargo quantity determining module is used for taking the quantity of the target object candidate frames obtained after the screening and the de-weighting as the cargo quantity in the cargo image.

In a third aspect, the present application further provides an electronic device, including:

one or more processors;

a storage device for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the cargo quantity identification method as described above.

In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the cargo quantity identification method as described above.

In a fifth aspect, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the cargo quantity identification method as described above.

According to the technical scheme, the image features are extracted by adopting a neural network model with multilayer convolution, the image features output by at least two layers of convolution layers in the model are selected to determine an object candidate frame set in the goods image, then the finally identified target object candidate frames are determined from the object candidate frame set through screening and de-duplication, and the number of the target object candidate frames is used as the number of the goods in the goods image. The method comprises the steps of selecting image characteristics output by at least two layers of convolution layers in a model, enabling the model to identify objects with different positions and sizes in a goods image, finding various targets in the image, accurately identifying goods in the image and then improving the accuracy of the model in identifying the quantity of the goods.

Drawings

Fig. 1 is a flowchart of a cargo quantity identification method according to a first embodiment of the present application;

fig. 2 is a flowchart of a cargo quantity identification method according to a second embodiment of the present application;

fig. 3 is a flowchart of a cargo quantity identification method in the third embodiment of the present application;

fig. 4 is a schematic structural diagram of a cargo quantity identification device in a fourth embodiment of the present application;

fig. 5 is a schematic structural diagram of an electronic device in a fifth embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the following drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the application and are not limiting of the application. It should be further noted that, for the convenience of description, only some of the structures related to the present application are shown in the drawings, not all of the structures.

Example one

Fig. 1 is a flowchart of a cargo quantity identification method according to an embodiment of the present application, which is applicable to image identification of a cargo image and a situation of acquiring a quantity of cargo in the image, and relates to the technical field of image identification and deep learning. The method may be performed by a cargo quantity identification device, which may be implemented in software and/or hardware, and is preferably configured in an electronic device, such as a computer device or a server. As shown in fig. 1, the method specifically includes:

s101, carrying out image feature extraction on the cargo image by using a pre-trained quantity recognition model, wherein the quantity recognition model has a plurality of convolution layers.

The quantity recognition model is a neural network model with a multilayer convolution structure, which is trained in advance by using a machine learning method. In the training process of the quantity recognition model, a large number of images of the goods to be recognized are obtained as training samples, and the goods in the images are labeled manually, for example, a rectangular frame is used for marking a circle of the cross section of the goods; then, inputting the training samples into a model, and predicting the goods in the training samples by the model; calculating the error of the predicted value according to the labeled value of the training sample; calculating the gradient of the error value relative to the convolution parameter, and optimizing the model parameter; and obtaining a final quantity identification model based on the convolutional neural network through iteration and tuning of the process.

S102, determining an object candidate frame set in the cargo image according to image characteristics output by at least two layers of convolution layers in the quantity identification model, wherein the image characteristics at least comprise object categories and positioning coordinates of the object candidate frames.

The trained quantity recognition model can perform feature extraction on the cargo image, and each convolution layer can output image features including object categories and positioning coordinates of object candidate frames. And the candidate frames of each object on the cargo image can be identified by the model according to the image characteristics. The object class may include a target and a background, the target refers to an object in a foreground of the image, and the background refers to a background portion of the image. It is understood that the goods to be identified are objects in the image, the category of which is the target. The positioning coordinates are used to position the object candidate frame, and may include, for example, X, Y coordinates at the top left corner and X, Y coordinates at the bottom right corner of the object candidate frame, and since the candidate frame is usually rectangular, the candidate frame may be positioned according to X, Y coordinates at the top left corner and the bottom right corner of the candidate frame. Of course, the positioning coordinates may also be coordinates of other points on the candidate frame, which is not limited in this embodiment of the present application.

In the embodiment of the application, image features output by at least two layers of convolution layers in a model are selected, and an object candidate frame set in a goods image is determined. The benefits of this are: since the sizes of the object candidate frames corresponding to the image features output by different convolutional layers are different, the sizes of the objects with the types as targets are different; considering that the goods in the goods image are not necessarily consistent in size, and even if the goods are the same in size, the front and back distances of the goods displayed in the image are inconsistent due to the placing mode, namely the sizes of the goods displayed in the image are inconsistent; therefore, the image characteristics output by at least two layers of convolution layers in the model are selected to determine the object candidate frame set, and the objects with different positions and sizes in the goods image can be obtained, so that the model has higher recall rate, and the accuracy of the model in identifying the quantity of the goods is improved.

In one embodiment, the number of convolutional layers in the number recognition model is greater than 2, and the at least two convolutional layers include the last convolutional layer and at least one convolutional layer in front of the last convolutional layer. For example, if the number of convolutional layers in the number recognition model may be 7, then the image features output by the 5 th, 6 th, and 7 th convolutional layers may be selected. Of course, in other embodiments, the image features output by the 4 th, 5 th, 6 th, and 7 th convolutional layers, or the image features output by the 6 th and 7 th convolutional layers may be selected, and this may be configured according to actual needs, which is not limited in this embodiment.

S103, screening and de-duplicating the object candidate frames in the object candidate frame set according to the object categories in the image characteristics and the positioning coordinates of the object candidate frames.

Each object candidate frame in the acquired object candidate frame set is an image feature output from at least two layers of convolutional layers, and some object types in the image features are targets and some object types are backgrounds, so that screening needs to be performed according to the object types. Meanwhile, it is possible that image features output by different convolutional layers point to the same object, and therefore, the object candidate frames need to be deduplicated.

Specifically, object candidate frames corresponding to image features of which the object types are targets may be retained, and then, in each object candidate frame after the screening, it is determined which object candidate frames have a high degree of coincidence, and then, the object candidate frames having a high degree of coincidence refer to the same object, and then, the object candidate frames are de-duplicated. For example, the positional relationship between different object candidate frames may be determined from the positioning coordinates of the object candidate frames, and then the degree of coincidence thereof may be determined based on the positional relationship.

And S104, taking the number of the target object candidate frames obtained after screening and de-weighting as the number of the goods in the goods image.

And obtaining target object candidate frames after screening and de-weighting, wherein the object in each target object candidate frame is the finally identified goods in the image, so that the number of the target object candidate frames can be used as the number of the goods in the goods image.

In the technical scheme of the embodiment of the application, the neural network model with multilayer convolution is adopted to extract image features, the image features output by at least two layers of convolution layers in the model are selected to determine an object candidate frame set in the goods image, then, through screening and de-duplication, a finally identified target object candidate frame is determined from the object candidate frame set, and the number of the target object candidate frames is used as the number of goods in the goods image. The method comprises the steps of selecting image characteristics output by at least two layers of convolution layers in a model, enabling the model to identify objects with different positions and sizes in a goods image, finding various targets in the image, accurately identifying goods in the image and then improving the accuracy of the model in identifying the quantity of the goods. In addition, the goods identification is realized based on the multilayer convolutional neural network, and the features with different degrees and different shapes can be extracted from the target in the image, not only the edge feature and the binarization feature, so that the model has strong robustness compared with the prior art; meanwhile, the method can be suitable for identifying different types and shapes of goods, and is more convenient for subsequent expansion.

Example two

Fig. 2 is a flowchart of a cargo quantity identification method according to a second embodiment of the present application, and the present embodiment is further optimized based on the foregoing embodiments. As shown in fig. 2, the method includes:

s201, carrying out image feature extraction on the cargo image by using a pre-trained quantity recognition model, wherein the quantity recognition model has a plurality of convolution layers.

S202, determining an object candidate frame set in the cargo image according to image characteristics output by at least two layers of convolutional layers in the quantity identification model, wherein the image characteristics at least comprise object categories and positioning coordinates of the object candidate frames.

In one embodiment, at least one convolution layer in the quantity recognition model is connected with a deconvolution layer, and the image features output by the at least two convolution layers are deconvolution-processed image features. Wherein the deconvolution layer is to: and expanding the original image characteristics output by the connected convolutional layer through deconvolution, and adding the expanded image characteristics with the image characteristics output by the convolutional layer in front of the connected convolutional layer.

For example, if there are 7 convolutional layers in the number-recognition model, then at least one of the 3 rd to 7 th convolutional layers may be connected with an anti-convolutional layer. Since the characteristic size of the image output by the 2 nd layer convolution layer is too large, the addition of deconvolution is likely to cause a problem of a decrease in calculation efficiency due to too high calculation amount, and therefore, the deconvolution may not be added after the 2 nd layer convolution layer.

Taking the example that the 7 th convolutional layer is connected with the deconvolution layer, the 7 th convolutional layer outputs an original image feature through feature extraction, then the original image feature is expanded through deconvolution, the expanded image feature is added with the image feature output by the 6 th convolutional layer, and the added result is used as the image feature finally output by the 7 th convolutional layer, namely the image feature selected in the embodiment of the application.

The effect of adding the deconvolution layer is as follows: the features extracted by the convolutional layer are recombined through a deconvolution structure and added with the image features output by the convolutional layer on the upper layer of the convolutional layer, so that more detailed image features can be obtained. That is, the bottom-layer features can be combined with the top-layer features, and prediction is performed based on the top-layer features, so that the model can more accurately identify the goods in the goods image, and mark the positions of the goods in the image through the candidate frames, and the accuracy of the model for identifying the goods is improved.

S203, selecting the object candidate frame with the object type as the target from the object candidate frame set.

And S204, determining the area and the position of the object candidate frame with the object type as the target according to the positioning coordinates.

S205, determining the coincidence degree between different object candidate frames according to the area and the position, and performing de-coincidence on the object candidate frame of which the object type is the target.

The background is removed from the set of candidate frames in step S204, and then the candidate frames representing the same subject are deduplicated in step S205, so as to obtain the final identified candidate frame of the coworker subject. In specific implementation, a coincidence degree threshold may be configured, and if it is determined that the coincidence degree of any two object candidate frames is greater than the threshold according to the position and the area, it is considered that the two object candidate frames correspond to the same object.

And S206, taking the number of the target object candidate frames obtained after screening and de-weighting as the number of the goods in the goods image.

According to the technical scheme, the image features are extracted by adopting a neural network model with multilayer convolution, and the image features output by at least two layers of convolution layers in the model are selected to determine the object candidate frame set in the goods image, so that the model can identify objects in different positions and different sizes in the goods image, and various targets in the image can be found. And then, through screening and de-duplication, the finally identified target object candidate frame is determined from the object candidate frame set, and the number of the target object candidate frames is used as the number of the goods in the goods image, so that the identification precision of the number of the goods is improved. Meanwhile, the deconvolution network is added into the model structure, so that the model can detect targets with different positions and sizes in the image, the accuracy of identifying the goods in the image is further improved, the model has higher recall rate, and the accuracy of identifying the quantity of the goods is further improved.

EXAMPLE III

Fig. 3 is a flowchart of a cargo quantity identification method provided in the third embodiment of the present application, and the present embodiment is further optimized based on the foregoing embodiments. As shown in fig. 3, the method includes:

s301, image feature extraction is carried out on the cargo image by utilizing a pre-trained quantity recognition model, wherein the quantity recognition model has a plurality of layers of convolution layers.

S302, determining an object candidate frame set in the cargo image according to image characteristics output by at least two layers of convolutional layers in the quantity identification model, wherein the image characteristics at least comprise object categories and positioning coordinates of the object candidate frames.

S303, screening and de-duplicating the object candidate frames in the object candidate frame set according to the object categories in the image characteristics and the positioning coordinates of the object candidate frames.

S304, taking the number of the candidate frames of the target object obtained after screening and de-weighting as the number of the goods in the goods image.

S305, determining the center point of each target object candidate frame according to the positioning coordinates of the target object candidate frames.

And S306, displaying the center point of the target object candidate frame on the cargo image.

In order to realize visualization of cargo quantity identification and facilitate management of management personnel, the cargo image may be displayed in the corresponding management system, and simultaneously, the center point of each identified target object candidate box is also displayed in the cargo image, for example, the center point is identified by a green dot. Through these central points that show, can directly see whether goods discernment is accurate, whether have the goods of omission and not discern, or mark the position of central point and be not the goods of waiting to discern.

And S307, responding to the click operation of the central point of the target object candidate frame displayed on the goods image, and reducing the number of the goods in the goods image.

And S308, responding to the clicking operation of other objects except the target object represented by the center point on the goods image, and increasing the number of the goods in the goods image.

When the identified cargo has an error, correction can be performed through the above steps S307, S308. For example, if the identified center point is not a good, the manager may click on the displayed green center point to eliminate it, thereby reducing the number of goods in the goods image. If the goods are not identified, the position of the goods which are not identified can be directly clicked on the goods image, the goods can be determined according to the clicked position, and the number of the goods in the goods image is increased. Meanwhile, the center point of the cross section of the manually selected cargo can be displayed on the image. Through the interaction process, the recognition results of the goods and the quantity of the goods can be visually checked, meanwhile, the initial recognition result can be corrected, and the efficiency of goods management is improved.

In addition, the method of the embodiment of the application further comprises the following steps: acquiring a new image training sample according to the clicking operation and the reduced or increased number of the goods; and updating the quantity recognition model by using the new image training sample. That is, after the initial recognition result is corrected, new training data can be generated according to the correction condition in order to continuously improve the accuracy and recall rate of the model, and thus, the model is optimized through continuous iteration and training.

According to the technical scheme of the embodiment of the application, the neural network model with multilayer convolution is adopted to extract the image characteristics, and the robustness is strong; and selecting image characteristics output by at least two layers of convolution layers in the model to determine an object candidate frame set in the goods image, so that the model can identify objects with different positions and sizes in the goods image, thereby finding various targets in the image and improving the recall rate of the model; and then, through screening and de-duplication, the finally identified target object candidate frame is determined from the object candidate frame set, and the number of the target object candidate frames is used as the number of the goods in the goods image, so that the identification precision of the number of the goods is improved. Meanwhile, the identification result can be corrected through human-computer interaction, more accurate cargo quantity is further obtained, and cargo management efficiency is improved.

Example four

Fig. 4 is a schematic structural view of the cargo quantity recognition apparatus in the present embodiment. The embodiment can be suitable for carrying out image recognition on the goods image, obtains the condition of the quantity of goods in the image, and relates to the technical field of image recognition and deep learning. The device can realize the cargo quantity identification method in any embodiment of the application. As shown in fig. 4, the apparatus 400 specifically includes:

an image feature extraction module 401, configured to perform image feature extraction on a cargo image by using a pre-trained quantity recognition model, where the quantity recognition model has multiple convolutional layers;

an object candidate frame set determining module 402, configured to determine an object candidate frame set in the cargo image according to image features output by at least two convolutional layers in the quantity recognition model, where the image features at least include an object category and location coordinates of an object candidate frame;

a screening and de-duplication module 403, configured to screen and de-duplicate object candidate frames in the object candidate frame set according to the object categories in the image features and the location coordinates of the object candidate frames;

a cargo quantity determining module 404, configured to use the quantity of the candidate target object frames obtained after the screening and the deduplication as the cargo quantity in the cargo image.

Optionally, an deconvolution layer is further connected behind at least one layer of the number identification model, and image features output by the at least two layers of the convolution layers are image features subjected to deconvolution;

wherein the deconvolution layer is for: and expanding the original image characteristics output by the connected convolutional layer through deconvolution, and adding the expanded image characteristics with the image characteristics output by the convolutional layer in front of the connected convolutional layer.

Optionally, the number of convolutional layers in the number recognition model is greater than 2, and the at least two convolutional layers include a last convolutional layer and at least one layer before the last convolutional layer.

Optionally, the object category includes a target and a background;

accordingly, the screening and de-weighting module 403 includes:

a screening unit configured to select an object candidate box of which the object type is a target from the set of object candidate boxes;

an area and position determining unit configured to determine an area and a position of an object candidate frame of which the object type is a target, according to the positioning coordinates;

and the duplication removing unit is used for determining the coincidence degree between different object candidate frames according to the area and the position and carrying out duplication removal on the object candidate frames of which the object types are targets.

Optionally, the apparatus further includes a display module, specifically configured to:

determining the central point of each target object candidate frame according to the positioning coordinates of the target object candidate frames;

and displaying the center point of the target object candidate frame on the cargo image.

Optionally, the apparatus further includes a first modification module, configured to:

and reducing the number of cargos in the cargo image in response to a click operation on the central point of the target object candidate frame displayed on the cargo image.

Optionally, the apparatus further includes a second modification module, configured to:

and responding to the clicking operation of other objects except the target object represented by the central point on the goods image, and increasing the number of the goods in the goods image.

Optionally, the apparatus further includes a model training update module, configured to:

acquiring a new image training sample according to the clicking operation and the reduced or increased number of the goods;

updating the quantity recognition model with the new image training samples.

Optionally, the positioning coordinates include the top left corner X, Y coordinate and the bottom right corner X, Y coordinate of the object candidate box.

The goods quantity identification device provided by the embodiment of the application can execute the goods quantity identification method provided by any embodiment of the application, and has corresponding functional modules and beneficial effects of the execution method.

EXAMPLE five

Fig. 5 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present application. FIG. 5 illustrates a block diagram of an exemplary electronic device 12 suitable for use in implementing embodiments of the present application. The electronic device 12 shown in fig. 5 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.

As shown in FIG. 5, electronic device 12 is embodied in the form of a general purpose computing device. The components of electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.

Bus 18 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example, such architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, micro-channel architecture (MAC) bus, enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

Electronic device 12 typically includes a variety of computer system readable media. Such media may be any available media that is accessible by electronic device 12 and includes both volatile and nonvolatile media, removable and non-removable media.

The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)30 and/or cache memory 32. The electronic device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 5, and commonly referred to as a "hard drive"). Although not shown in FIG. 5, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the application.

A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally perform the functions and/or methodologies of the embodiments described herein.

Electronic device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), with one or more devices that enable a user to interact with electronic device 12, and/or with any devices (e.g., network card, modem, etc.) that enable electronic device 12 to communicate with one or more other computing devices. Such communication may be through an input/output (I/O) interface 22. Also, the electronic device 12 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet) via the network adapter 20. As shown, the network adapter 20 communicates with other modules of the electronic device 12 via the bus 18. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

The processing unit 16 executes various functional applications and data processing, such as implementing the cargo quantity identification method provided in the embodiment of the present application, by executing the program stored in the system memory 28.

EXAMPLE six

The sixth embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the cargo quantity identification method provided in the sixth embodiment of the present application.

The computer storage media of the embodiments of the present application may take any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations for aspects of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

Furthermore, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the cargo quantity identification method as provided by any of the embodiments above.

It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present application and the technical principles employed. It will be understood by those skilled in the art that the present application is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the application. Therefore, although the present application has been described in more detail with reference to the above embodiments, the present application is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. A cargo quantity identification method, comprising:

2. The method according to claim 1, wherein a deconvolution layer is further connected behind at least one layer of the number recognition model, and the image features output by the at least two layers of the convolution layers are deconvoluted image features;

3. The method of claim 1, wherein the number of convolutional layers in the number-identifying model is greater than 2, and the at least two convolutional layers comprise a last convolutional layer and at least one preceding convolutional layer.

4. The method of claim 1, wherein the object classes include a target and a background;

correspondingly, the screening and the de-duplication of the object candidate frames in the object candidate frame set according to the object category in the image feature and the location coordinates of the object candidate frames includes:

selecting an object candidate box of which the object type is a target from the object candidate box set;

determining the area and the position of an object candidate frame of which the object type is a target according to the positioning coordinates;

and determining the coincidence degree between different object candidate frames according to the area and the position, and performing de-duplication on the object candidate frame of which the object type is the target.

5. The method of claim 1, further comprising:

6. The method of claim 5, further comprising:

7. The method of claim 5, further comprising:

8. The method of claim 6 or 7, further comprising:

updating the quantity recognition model with the new image training samples.

9. The method of claim 1, wherein the positioning coordinates comprise an upper left corner X, Y coordinate and a lower right corner X, Y coordinate of the object candidate box.

10. A cargo quantity identification device, comprising:

11. An electronic device, comprising:

one or more processors;

a storage device for storing one or more programs,

when executed by the one or more processors, cause the one or more processors to implement the cargo quantity identification method of any of claims 1-9.

12. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, is adapted to carry out the method for identifying a quantity of goods according to any one of claims 1 to 9.

13. A computer program product comprising a computer program, characterized in that the computer program realizes the cargo quantity identification method according to any one of claims 1-9 when executed by a processor.