WO2024044947A1 - 缺陷检测的方法、装置和计算机可读存储介质 - Google Patents

缺陷检测的方法、装置和计算机可读存储介质 Download PDF

Info

Publication number
WO2024044947A1
WO2024044947A1 PCT/CN2022/115774 CN2022115774W WO2024044947A1 WO 2024044947 A1 WO2024044947 A1 WO 2024044947A1 CN 2022115774 W CN2022115774 W CN 2022115774W WO 2024044947 A1 WO2024044947 A1 WO 2024044947A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
feature map
detected
defect detection
coordinate information
Prior art date
Application number
PCT/CN2022/115774
Other languages
English (en)
French (fr)
Inventor
王智玉
王晞
江冠南
Original Assignee
宁德时代新能源科技股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 宁德时代新能源科技股份有限公司 filed Critical 宁德时代新能源科技股份有限公司
Priority to PCT/CN2022/115774 priority Critical patent/WO2024044947A1/zh
Priority to CN202280006731.XA priority patent/CN117957438A/zh
Priority to EP22871168.5A priority patent/EP4357765A4/en
Priority to US18/356,232 priority patent/US20240070840A1/en
Publication of WO2024044947A1 publication Critical patent/WO2024044947A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component

Definitions

  • the present application relates to the field of image processing technology, and in particular to a defect detection method, device and computer-readable storage medium.
  • This application provides a defect detection method, device and computer-readable storage medium, which can improve the sensitivity of the detection neural network to spatial position, thereby enhancing the detection accuracy of the detection neural network for some characteristic defect types, and improving the efficiency of defect detection. Accuracy.
  • a method of defect detection including: obtaining an image to be detected; and obtaining a feature map of the image to be detected based on the image to be detected, where the feature map of the image to be detected includes spatial position coordinate information. the feature map of the image to be detected; perform defect detection on the image to be detected according to the feature map of the image to be detected.
  • the feature map of the spatial position coordinate information is extracted simultaneously during the detection process, so that the neural network used for defect detection is sensitive to the spatial position, thereby improving the detection neural network. Sensitivity to spatial position, thereby enhancing the detection accuracy of the detection neural network for some specific defect types, and improving the accuracy of defect detection.
  • the feature map of the image to be detected also includes a feature map of image information
  • obtaining the feature map of the image to be detected based on the image to be detected includes: based on the image to be detected Image, obtain the feature map of the image information; generate the feature map of the spatial position coordinate information according to the feature map of the image information, the dimension of the feature map of the spatial position coordinate information is consistent with the feature map of the image information dimensions are the same.
  • the dimensions of the two types of feature maps are the same, thereby facilitating subsequent image processing and improving the accuracy of defect detection.
  • generating a feature map of the spatial position coordinate information based on the feature map of the image information includes: generating a linear value corresponding to the spatial position coordinate information; according to the linear value, Generate a first coordinate network; expand the dimensions of the first coordinate network according to the feature map of the image information to generate a feature map of the spatial position coordinate information.
  • the feature map of the spatial position coordinate information is generated by generating a coordinate network corresponding to the spatial position coordinate information, thereby splicing the position feature map into the feature map of the image information, and combining the two to improve the sensitivity of the spatial information. characteristics, which is conducive to improving the accuracy of defect detection.
  • performing defect detection on the image to be detected based on the feature map of the image to be detected includes: using a filter model to detect defects on the image to be detected based on the feature map of the image to be detected.
  • the inspection image is used for defect detection, wherein the filter model includes a filter for processing the feature map of the spatial position coordinate information.
  • corresponding filters are added to the feature map corresponding to the spatial position coordinate information in subsequent detection, thereby facilitating defect detection and improving the accuracy of defect detection.
  • the filter is a filter of the first convolutional layer.
  • the feature map of the spatial position coordinate information includes at least one of an x-axis direction coordinate information feature map, a y-axis direction coordinate information feature map, and a z-axis direction coordinate information feature map.
  • the above embodiment provides a variety of spatially sensitive types by defining the feature map of spatial position coordinate information as a feature map of at least one method among x, y, and z, which facilitates defect detection and improves the accuracy of defect detection.
  • the method is used for defect detection of pole tabs and/or pole pieces.
  • the above embodiments can improve the production efficiency of power batteries when used for defect detection of tabs/pole pieces in power batteries.
  • the defect detection of the tab includes defect detection of the folding characteristics of the tab.
  • obtaining the feature map of the image to be detected based on the image to be detected includes: inputting the image to be detected into a neural network; performing defect characteristics through the backbone network of the neural network Vector extraction and coordinate information feature extraction corresponding to the defect feature vector are used to obtain the feature map of the image to be detected.
  • the above embodiment can improve the accuracy of defect detection by simultaneously extracting the feature vector and corresponding coordinate features of the defect.
  • a device for defect detection including: an acquisition unit for acquiring an image to be detected; a processing unit for acquiring a feature map of the image to be detected based on the image to be detected, the image to be detected being The feature map of the detected image includes a feature map of spatial position coordinate information; the processing unit is further configured to perform defect detection on the image to be detected based on the feature map of the image to be detected.
  • the feature map of the spatial position coordinate information is extracted simultaneously during the detection process, so that the neural network used for defect detection is sensitive to the spatial position, thereby improving the detection neural network. Sensitivity to spatial position, thereby enhancing the detection accuracy of the detection neural network for some specific defect types, and improving the accuracy of defect detection.
  • the feature map of the image to be detected also includes a feature map of image information
  • the processing unit is configured to: obtain the feature map of the image information according to the image to be detected; according to the The feature map of the image information is used to generate the feature map of the spatial position coordinate information, and the dimension of the feature map of the spatial position coordinate information is the same as the dimension of the feature map of the image information.
  • the dimensions of the two types of feature maps are the same, thereby facilitating subsequent image processing and improving the accuracy of defect detection.
  • the processing unit is configured to: generate a linear value corresponding to the spatial position coordinate information; generate a first coordinate network based on the linear value; and expand the feature map of the image information based on the linear value.
  • the dimensions of the first coordinate network are used to generate a feature map of the spatial position coordinate information.
  • the feature map of the spatial position coordinate information is generated by generating a coordinate network corresponding to the spatial position coordinate information, thereby splicing the position feature map into the feature map of the image information, and combining the two to improve the sensitivity of the spatial information. characteristics, which is conducive to improving the accuracy of defect detection.
  • the processing unit is configured to: use a filter model to perform defect detection on the image to be detected according to the feature map of the image to be detected, wherein the filter model includes: The feature map of the spatial position coordinate information is processed by a filter.
  • corresponding filters are added to the feature map corresponding to the spatial position coordinate information in subsequent detection, thereby facilitating defect detection and improving the accuracy of defect detection.
  • the filter is a filter of the first convolutional layer.
  • the feature map of the spatial position coordinate information includes at least one of an x-axis direction coordinate information feature map, a y-axis direction coordinate information feature map, and a z-axis direction coordinate information feature map.
  • the above embodiment provides a variety of spatially sensitive types by defining the feature map of spatial position coordinate information as a feature map of at least one method among x, y, and z, which facilitates defect detection and improves the accuracy of defect detection.
  • the method is used for defect detection of pole tabs and/or pole pieces.
  • the above embodiments can improve the production efficiency of power batteries when used for defect detection of tabs/pole pieces in power batteries.
  • the defect detection of the tab includes defect detection of the folding feature of the tab.
  • the processing unit is configured to: input the image to be detected into a neural network; perform defect feature vector extraction and coordinate information feature extraction corresponding to the defect feature vector through the backbone network of the neural network, so as to Obtain the feature map of the image to be detected.
  • the above embodiment can improve the accuracy of defect detection by simultaneously extracting the feature vector and corresponding coordinate features of the defect.
  • a device for defect detection including a processor and a memory.
  • the memory is used to store a program.
  • the processor is used to call and run the program from the memory to execute the above first aspect or The defect detection method in any possible implementation of the first aspect.
  • a computer-readable storage medium including a computer program.
  • the computer program When the computer program is run on a computer, it causes the computer to execute the above-mentioned first aspect or any possible implementation manner of the first aspect. method of defect detection.
  • a fifth aspect provides a computer program product containing instructions that, when executed by a computer, cause the computer to perform the defect detection method in the above-mentioned first aspect or any possible implementation of the first aspect.
  • Figure 1 is a schematic structural diagram of the system architecture provided by this application.
  • Figure 2 is an image of the pole lug of a power battery to be inspected
  • Figure 3 is a schematic flow chart of a defect detection method disclosed in an embodiment of the present application.
  • Figure 4 is a schematic block diagram of a defect detection method disclosed in an embodiment of the present application.
  • Figure 5 is a schematic structural block diagram of a defect detection device disclosed in an embodiment of the present application.
  • Figure 6 is a schematic diagram of the hardware structure of a defect detection device disclosed in an embodiment of the present application.
  • Embodiments of the present application may be applied to image processing systems, including but not limited to products based on infrared imaging.
  • the defect detection system can be applied to various electronic devices with defect detection devices.
  • the electronic devices can be personal computers, computer workstations, smartphones, tablets, smart cameras, media consumption devices, wearable devices, set-top boxes, game consoles, Augmented reality (AR), AR/virtual reality (VR) equipment, vehicle-mounted terminals, etc.
  • AR Augmented reality
  • VR virtual reality
  • the size of the sequence numbers of each process does not mean the order of execution.
  • the execution order of each process should be determined by its functions and internal logic, and should not be used in the embodiments of the present application.
  • the implementation process constitutes any limitation.
  • this embodiment of the present application provides a system architecture 100.
  • data acquisition device 160 is used to acquire samples of known defects.
  • the sample image with known defects may include a sample image with one or more defects, and the type of defect in the sample image is known.
  • the data collection device 160 After collecting the known defective sample images, the data collection device 160 stores these known defective sample images into the database 130 , and the training device 120 trains the target model/rule 101 based on the known defective sample images maintained in the database 130 .
  • the above target model/rule 101 can be used to implement the defect detection method in the embodiment of the present application.
  • the target model/rule 101 in the embodiment of this application may specifically be a neural network.
  • the known defective sample images maintained in the database 130 may not all be collected by the data collection device 160, and may also be received from other devices.
  • the training device 120 may not necessarily train the target model/rules 101 based entirely on the known defective sample images maintained by the database 130. It may also obtain known defective sample images from the cloud or other places for model training, as described above. The description should not be used as a limitation on the embodiments of the present application.
  • the target model/rules 101 trained according to the training device 120 can be applied to different systems or devices, such as to the execution device 110 shown in Figure 1.
  • the execution device 110 can be a terminal, such as a mobile phone terminal, a tablet computer, Laptops, etc., or servers or clouds, etc.
  • the execution device 110 is configured with an input/output (I/O) interface 112 for data interaction with external devices.
  • the user can input data to the I/O interface 112 through the client device 140.
  • the input data may include: the image to be detected input by the client device 140 .
  • the client device 140 may be the same device as the above-mentioned execution device 110.
  • the client device 140 and the above-mentioned execution device 110 may both be terminal devices.
  • the client device 140 may be a different device from the above-mentioned execution device 110.
  • the client device 140 is a terminal device, and the execution device 110 is a cloud, server, or other device.
  • the client device 140 may communicate via any communication mechanism/ The communication network of the communication standard interacts with the execution device 110.
  • the communication network may be a wide area network, a local area network, a point-to-point connection, or any combination thereof.
  • the computing module 111 of the execution device 110 is used to perform processing according to the input data (such as the image to be detected) received by the I/O interface 112 .
  • the execution device 110 can call the data, codes, etc. in the data storage system 150 for corresponding processing, or can also use the data, instructions, etc. obtained by the corresponding processing. stored in the data storage system 150.
  • the I/O interface 112 returns the processing result, such as the defect classification result obtained above, to the client device 140, thereby providing it to the user.
  • the training device 120 can generate corresponding target models/rules 101 based on different training data for different goals or different tasks, and the corresponding target models/rules 101 can be used to achieve the above goals or complete the The above tasks, thereby providing the user with the desired results.
  • the user can manually enter the input data, and the manual input can be operated through the interface provided by the I/O interface 112 .
  • the client device 140 can automatically send input data to the I/O interface 112. If requiring the client device 140 to automatically send input data requires the user's authorization, the user can set corresponding permissions in the client device 140.
  • the user can view the results output by the execution device 110 on the client device 140, and the specific presentation form may be display, sound, action, etc.
  • the client device 140 can also be used as a data collection end to collect the input data of the input I/O interface 112 and the output results of the output I/O interface 112 as new sample data, and store them in the database 130 .
  • the I/O interface 112 directly uses the input data input to the I/O interface 112 and the output result of the output I/O interface 112 as a new sample as shown in the figure.
  • the data is stored in database 130.
  • Figure 1 is only a schematic diagram of a system architecture provided by an embodiment of the present application.
  • the positional relationship between the devices, devices, modules, etc. shown in the figure does not constitute any limitation.
  • the data The storage system 150 is an external memory relative to the execution device 110. In other cases, the data storage system 150 can also be placed in the execution device 110.
  • a target model/rule 101 is obtained by training according to the training device 120.
  • the target model/rule 101 in this embodiment of the present application may be a neural network.
  • the neural network in this embodiment of the present application may be a convolutional neural network.
  • Network convolutional neuron network, CNN
  • region CNN region CNN
  • RCNN region convolutional neural network
  • this application does not specifically limit this.
  • Power batteries can not only be used in primary power systems such as water conservancy, wind power, thermal power and solar power stations, but can also be widely used in electric vehicles such as electric bicycles, electric motorcycles and electric cars, as well as in many fields such as military equipment and aerospace. .
  • power batteries can also be widely used in electric vehicles such as electric bicycles, electric motorcycles and electric cars, as well as in many fields such as military equipment and aerospace. .
  • defects may occur in multiple processes, such as anode die-cutting, cathode die-cutting, winding and gravure, etc.
  • Vision technology is required to automatically detect, that is, defect detection.
  • Defect detection usually refers to the detection of surface defects of items.
  • Surface defect detection uses advanced machine vision detection technology to detect defects such as spots, pits, scratches, color differences, and defects on the surface of the workpiece.
  • Figure 2 shows an image of a pole lug of a power battery to be detected. When it is necessary to detect the folding defect of the pole lug, you first need to find the local features of the folding defect in the image, as shown in Figure 2.
  • It includes two basically identical local features 210 and 220, where the local feature 210 is located at the tail of the image to be detected, which is caused by the preparation process of the tab and belongs to the normal form, while the local feature 220 is located in the middle of the image to be detected.
  • the lines gradually disappear, indicating that there is a folding defect in the middle of the pole lug. Therefore, there are two basically the same local features 210 and 220.
  • One is a normal form and does not need to be detected, while the other is a defect and needs to be detected. This When , the folding defect of the tab is position-sensitive.
  • the types of defects may also be different in different locations.
  • the traditional detection method requires a combination of neural network and logical post-processing technology to meet the requirements of defect position sensitivity, and the detection efficiency is low. For example, since the neural network cannot learn the same local feature, which locations should be detected and which locations should not be detected, the neural network can detect them all, and then manually analyze the shape of the defect. Annotation and logical post-processing are performed, so the detection efficiency of this method is low.
  • embodiments of the present application provide a method for defect detection.
  • the feature map of the spatial position coordinate information is extracted simultaneously during the detection process, so that the neural network used for defect detection is spatially accurate.
  • Position sensitivity can improve the sensitivity of the detection neural network to spatial position, thereby enhancing the detection accuracy of the detection neural network for some specific defect types and improving the accuracy of defect detection.
  • the defect detection methods and devices provided by the embodiments of this application can be, but are not limited to, defect detection for power batteries, and can also be applied to defect detection of other types of products in modern industrial manufacturing.
  • the main process of the defect detection method in the embodiment of the present application will be introduced below with reference to Figures 3 and 4.
  • FIG. 3 shows a schematic flowchart of a defect detection method 300 according to an embodiment of the present application.
  • the defect detection method 300 includes the following steps.
  • the image to be detected is a photographed image of the workpiece to be detected. That is, when detecting defects on the workpiece to be detected, the workpiece to be detected can first be photographed to obtain the image to be detected.
  • the poles of a power battery can be imaged. Photograph the lug or pole piece to obtain the image to be inspected of the pole lug or pole piece, so as to detect the defects of the pole lug or pole piece.
  • the tool used to capture images may be images captured by a charge coupled device (CCD) camera, or may be images captured by other cameras or video cameras, which is not limited in this application.
  • CCD charge coupled device
  • the power battery production process generally includes multiple production processes.
  • a CCD camera can be used to capture the power battery to form an image to be inspected.
  • the feature map of the image to be detected includes a feature map of spatial position coordinate information.
  • the image to be detected after obtaining the image to be detected, in order for the neural network to identify whether there is a defect in the image to be detected and the location of the defect, the image to be detected needs to be processed.
  • the image to be detected can be Feature extraction is performed on the image to obtain a feature map of the image to be detected, which feature map also includes a feature map of spatial position coordinate information, so as to facilitate subsequent defect detection based on the location of the defect.
  • the feature map can be further processed. For example, the feature map can be judged based on whether the feature map includes the characteristics of the corresponding defect. Whether there are defects and output the detection results. It is also worth noting that the above steps can all be operated by machines instead of manual operations, saving manpower and improving production efficiency.
  • the above method 300 extracts the feature map of the spatial position coordinate information during defect detection, so that the defect detection combined with the spatial position coordinate information can improve the sensitivity of the detection neural network to the spatial position, thereby improving the accuracy of defect detection.
  • the image to be detected may be a photographed image of the pole tab or pole piece, that is, defect detection of the pole tab or pole piece is performed.
  • the above method 300 can be used to detect tab folding defects. It is worth noting that for pole-lug folding defects, when similar features of folding appear at the tail of the pole-lug image, it is not a defect and should not be detected, but when it appears in the middle of the pole-lug image, it is a defect. , should be checked out.
  • the feature map of the image to be detected includes the feature map of spatial position coordinate information.
  • the feature map of the spatial position coordinate information may include at least one of an x-axis direction coordinate information feature map, a y-axis direction coordinate information feature map, and a z-axis direction coordinate information feature map.
  • the feature map of the image to be detected may also include the feature map of the image information.
  • the feature map of the image information is points, lines, surface features and/or color features in the image to be detected.
  • the method of obtaining the feature map of the image to be detected based on the image to be detected in step 320 may be: first, obtaining the feature map of the image information based on the image to be detected, for example, extracting points from the image to be detected. , line, surface features, etc.; then, according to the feature map of the image information, a feature map of the spatial position coordinate information is generated.
  • the feature map of the image information has the same dimension as the feature map of the spatial position coordinate information.
  • the above method of generating the feature map of the spatial position coordinate information based on the feature map of the image information may be: generating a linear value corresponding to the spatial position coordinate information, for example, generating a linear value from -1 to 1 corresponding to the spatial position. ; According to the linear value, generate a first coordinate network, which can be a one-dimensional coordinate network or a two-dimensional coordinate network; according to the feature map of the image information, expand the dimensions of the first coordinate network to generate spatial position coordinate information
  • the feature map for example, expands the dimension of the first coordinate network to the same dimension as the obtained point/line/area feature map, thereby forming a feature map of the spatial position coordinate information, that is, the position feature is obtained. Therefore, point/line/area features and position features can be combined later to obtain the input of the next convolutional layer.
  • the method of obtaining the feature map of the image to be detected in step 320 may be: input the image to be detected into the neural network; extract the defect feature vector and the coordinate information features corresponding to the defect feature vector through the backbone network of the neural network Extract to obtain the feature map of the image to be detected. That is, it can be understood that the feature map of the image to be detected is obtained through feature extraction of the neural network, and the feature extraction of the neural network can simultaneously extract the defect features and the coordinate features corresponding to the defects.
  • the model for defect detection of the image to be detected can be obtained by a neural network through effective defect sample training.
  • a filter model can be used to perform defect detection on the image to be detected based on the feature map of the image to be detected, where the filter model includes a filter for processing the feature map of the spatial position coordinate information. Or in other words, during the defect detection process, one or more filters corresponding to spatial position coordinate information are added.
  • the filter is the filter of the first convolutional layer. That is, a channel with spatial position coordinate information is added to the number of input channels of the first convolutional layer to match the feature map of the image to be detected.
  • the filter corresponding to the spatial position coordinate information can also be a filter of other convolutional layers, such as the i-th layer, where i is a positive integer greater than 0.
  • the output detection result may be the defective tab serial number and defect category.
  • FIG. 4 shows a schematic block diagram of the defect detection method 400 of the embodiment of the present application.
  • Figure 4 includes the same steps as those in the method 300. Please refer to the relevant introduction in the above method 300, which will not be described again here.
  • the feature extraction is performed on the image to be detected 410 to obtain the feature map 421 of the image information, the x-axis direction coordinate information feature map 422, and the y-axis direction coordinate information feature map 423, where
  • the size of the image information feature map 421 is (h, w, c)
  • the size of the x-axis direction coordinate information feature map 422 is (h, w, 1)
  • the size of the y-axis direction coordinate information feature map 423 is (h, w, 1), where h represents the height of the feature map (height), w represents the width of the feature map (width), c and 1 both represent the number of feature maps;
  • the feature map 421 of the image information and the x-axis coordinate is combined to obtain the feature map 420.
  • the size of the feature map 420 is (h, w, c+2), or in other words, c of the feature map 421 of the image information
  • the channel splicing operation is performed, and the channel map of the x-axis coordinate information and the channel map of the y-axis coordinate information are spliced to c channel maps of the image information to obtain c+2 channel maps, that is, features Figure 420;
  • defect detection is performed on the obtained feature map 420 to output the detection result 430.
  • the obtained feature map 420 can be input into a detector (detector) for defect detection to obtain the detection result output by the detection neural network. 430.
  • the feature map of the above-mentioned image information is combined with the spatial position coordinate information feature map
  • a piece of x-axis coordinate information is needed.
  • the channel diagram of the information is enough. If you only need the spatial coordinate position information in the y direction, you can splice a channel diagram of the y-axis coordinate information. If the spatial coordinate position information in the x direction and the spatial coordinate position information in the y direction are both If necessary, then splice two channel maps of x-axis coordinate information and a channel map of y-axis coordinate information.
  • the size of the feature map 420 is (h, w, c+1). After splicing, After two channel maps, the size of the feature map 420 is (h, w, c+2). Similarly, after splicing three channel maps, the size of the feature map 420 is (h, w, c+3). This There is no limit to the number of spliced channel images in the application.
  • the corresponding detector when the feature map 420 includes the spatial coordinate information feature map after adaptation and optimization, the corresponding detector also needs to modify its structure accordingly to adapt to the features after adaptation and optimization. picture. For example, when the size of the feature map after adaptation optimization changes from (h, w, c) to (h, w, c+2), the number of filters in the corresponding convolution layer also needs to be 2 more.
  • FIG. 5 shows a schematic block diagram of a defect detection device 500 according to an embodiment of the present application.
  • the device 500 can execute the defect detection method in the embodiment of the present application.
  • the device 500 can be the aforementioned execution device 110 .
  • the device includes:
  • the acquisition unit 520 is used to acquire the image to be detected.
  • the processing unit 520 is configured to obtain a feature map of the image to be detected based on the image to be detected, where the feature map of the image to be detected includes a feature map of spatial position coordinate information, and is also configured to obtain a feature map of the image to be detected based on the feature map of the image to be detected. Feature map, perform defect detection on the image to be detected.
  • FIG. 6 is a schematic diagram of the hardware structure of a defect detection device according to an embodiment of the present application.
  • the defect detection device 600 shown in FIG. 6 includes a memory 601, a processor 602, a communication interface 603 and a bus 604. Among them, the memory 601, the processor 602, and the communication interface 603 implement communication connections between each other through the bus 604.
  • the memory 601 may be a read-only memory (ROM), a static storage device, and a random access memory (RAM).
  • the memory 601 can store programs. When the program stored in the memory 601 is executed by the processor 602, the processor 602 and the communication interface 603 are used to execute various steps of the defect detection method in the embodiment of the present application.
  • the processor 602 may be a general central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), a graphics processing unit (GPU), or one or more
  • the integrated circuit is used to execute relevant programs to implement the functions required to be performed by the units in the defect detection device according to the embodiment of the present application, or to perform the defect detection method according to the embodiment of the present application.
  • the processor 602 may also be an integrated circuit chip with signal processing capabilities. During the implementation process, each step of the defect detection method in the embodiment of the present application can be completed by instructions in the form of hardware integrated logic circuits or software in the processor 602 .
  • the above-mentioned processor 602 can also be a general-purpose processor, a digital signal processor (digital signal processing, DSP), an ASIC, an off-the-shelf programmable gate array (field programmable gate array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
  • DSP digital signal processing
  • ASIC application-specific integrated circuit
  • FPGA field programmable gate array
  • Each method, step and logical block diagram disclosed in the embodiment of this application can be implemented or executed.
  • a general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
  • the steps of the method disclosed in conjunction with the embodiments of the present application can be directly implemented by a hardware processor, or executed by a combination of hardware and software modules in the processor.
  • the software module can be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other mature storage media in this field.
  • the storage medium is located in the memory 601.
  • the processor 602 reads the information in the memory 601, and combines its hardware to complete the functions required by the units included in the device for defect detection in the embodiment of the present application, or to perform the defect detection in the embodiment of the present application. Methods.
  • the communication interface 603 uses a transceiver device such as but not limited to a transceiver to implement communication between the device 600 and other devices or communication networks. For example, the traffic data of the unknown device can be obtained through the communication interface 603.
  • a transceiver device such as but not limited to a transceiver to implement communication between the device 600 and other devices or communication networks.
  • the traffic data of the unknown device can be obtained through the communication interface 603.
  • Bus 604 may include a path that carries information between various components of device 600 (eg, memory 601, processor 602, communication interface 603).
  • the device 600 may also include other devices necessary for normal operation. At the same time, based on specific needs, those skilled in the art should understand that the device 600 may also include hardware devices that implement other additional functions. In addition, those skilled in the art should understand that the device 600 may only include components necessary to implement the embodiments of the present application, and does not necessarily include all components shown in FIG. 6 .
  • Embodiments of the present application also provide a computer-readable storage medium that stores program code for device execution, and the program code includes instructions for executing steps in the above defect detection method.
  • Embodiments of the present application also provide a computer program product.
  • the computer program product includes a computer program stored on a computer-readable storage medium.
  • the computer program includes program instructions. When the program instructions are executed by a computer, The computer executes the above defect detection method.
  • the above-mentioned computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.
  • the disclosed devices and methods can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components may be combined or can be integrated into another system, or some features can be ignored, or not implemented.
  • the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical or other forms.
  • the described embodiments may be implemented by software, hardware, or a combination of software and hardware.
  • the described embodiments may also be embodied by computer-readable media having computer-readable code stored thereon, the computer-readable code including instructions executable by at least one computing device.
  • the computer-readable medium can be associated with any data storage device capable of storing data readable by a computer system. Examples of computer-readable media may include read-only memory, random access memory, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), hard disk drive (Hard Disk Drive, HDD), digital Video discs (Digital Video Disc, DVD), tapes and optical data storage devices, etc.
  • the computer-readable medium can also be distributed among computer systems coupled through a network, so that the computer-readable code can be stored and executed in a distributed manner.

Landscapes

  • Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

一种缺陷检测的方法（300）、装置和计算机可读存储介质，具体方法包括：获取待检测图像（310）；根据待检测图像，获取待检测图像的特征图，待检测图像的特征图包括空间位置坐标信息的特征图（320）；根据待检测图像的特征图，对待检测图像进行缺陷检测（330）。通过修改缺陷检测的神经网络结构，在检测的过程中同时提取空间位置坐标信息的特征图，使得用于缺陷检测的神经网络对空间位置敏感，从而能够提升检测申请网络对空间位置的敏感性，进而增强检测神经网络对于一些特定缺陷类型的检测准确性，提升缺陷检测的准确度。

Description

缺陷检测的方法、装置和计算机可读存储介质 技术领域
本申请涉及图像处理技术领域，特别是涉及一种缺陷检测的方法、装置和计算机可读存储介质。
背景技术
在现代工业生产过程中，由于工艺、设备等原因，生产出的产品可能会存在一定的缺陷，因此需要通过各种手段对产品进行检测，检出缺陷以提升产品良率。
然而，在对产品进行检测的过程中，由于缺陷的位置敏感性，需要将卷积神经网络检测与逻辑后处理的方式相结合，检测效率较低，因此，亟需提升产品缺陷检测效率。
发明内容
本申请提供了一种缺陷检测的方法、装置和计算机可读存储介质，能够提升检测神经网络对空间位置的敏感性，进而增强检测神经网络对于一些特征缺陷类型的检测准确性，提升缺陷检测的准确度。
第一方面，提供了一种缺陷检测的方法，包括：获取待检测图像；根据所述待检测图像，获取所述待检测图像的特征图，所述待检测图像的特征图包括空间位置坐标信息的特征图；根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测。
本申请的技术方案中，通过修改缺陷检测的神经网络结构，在检测的过程中同时提取空间位置坐标信息的特征图，使得用于缺陷检测的神 经网络对空间位置敏感，从而能够提升检测神经网络对空间位置的敏感性，进而增强检测神经网络对于一些特定缺陷类型的检测准确性，提升缺陷检测的准确度。
在一些可能的实施方式中，所述待检测图像的特征图还包括图像信息的特征图，所述根据所述待检测图像，获取所述待检测图像的特征图，包括：根据所述待检测图像，获取所述图像信息的特征图；根据所述图像信息的特征图，生成所述空间位置坐标信息的特征图，所述空间位置坐标信息的特征图的维度与所述图像信息的特征图的维度相同。
上述实施方式，通过先获取图像信息的特征图，再根据图像信息的特征图获取空间位置坐标信息的特征图，使得两类特征图的维度相同，从而便于后续图像处理，提升缺陷检测的准确度。
在一些可能的实施方式中，所述根据所述图像信息的特征图，生成所述空间位置坐标信息的特征图，包括：生成所述空间位置坐标信息对应的线性值；根据所述线性值，生成第一坐标网络；根据所述图像信息的特征图，扩充所述第一坐标网络的维度，以生成所述空间位置坐标信息的特征图。
上述实施方式，通过生成空间位置坐标信息对应的坐标网络来生成空间位置坐标信息的特征图，从而将位置特征图拼接至了图像信息的特征图中，两者结合，从而提升了空间信息的敏感性，有利于提升缺陷检测的准确度。
在一些可能的实施方式中，所述根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测，包括：根据所述待检测图像的特征图，使用过滤器模型对所述待检测图像进行缺陷检测，其中，所述过滤器模型包括用于对所述空间位置坐标信息的特征图进行处理的过滤器。
上述实施方式，对应于空间位置坐标信息的特征图，在后续检测 中添加了相应的过滤器，从而便于缺陷检测，提升缺陷检测的准确度。
在一些可能的实施方式中，所述过滤器为第1个卷积层的过滤器。
在一些可能的实施方式中，其特征在于，所述空间位置坐标信息的特征图包括x轴方向坐标信息特征图、y轴方向坐标信息特征图以及z轴方向坐标信息特征图中的至少一项。
上述实施方式，通过定义空间位置坐标信息的特征图为x、y、z中的至少一个方法的特征图，提供了多种空间敏感类型，便于缺陷检测，提升缺陷检测的准确度。
在一些可能的实施方式中，所述方法用于极耳和/或极片的缺陷检测。
上述实施方式，在用于动力电池中的极耳/极片的缺陷检测时，能够提升动力电池的生产效率。
在一些可能的实施方式中，当所述方法用于极耳的缺陷检测时，所述极耳的缺陷检测包括极耳翻折特征的缺陷检测。
上述实施方式，由于极耳翻折特征的位置敏感性，将该方式用于翻折缺陷检测时，能够提升检测准确度。
在一些可能的实施方式中，所述根据所述待检测图像，获取所述待检测图像的特征图，包括：将所述待检测图像输入神经网络；通过所述神经网络的骨干网络进行缺陷特征向量提取以及缺陷特征向量对应的坐标信息特征提取，以获取所述待检测图像的特征图。
上述实施方式，通过同时提取缺陷的特征向量与对应坐标特征，能够提升缺陷检测的准确度。
第二方面，提供了一种缺陷检测的装置，包括：获取单元，用于获取待检测图像；处理单元，用于根据所述待检测图像，获取所述待检测图像的特征图，所述待检测图像的特征图包括空间位置坐标信息的特征图； 所述处理单元还用于根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测。
本申请的技术方案中，通过修改缺陷检测的神经网络结构，在检测的过程中同时提取空间位置坐标信息的特征图，使得用于缺陷检测的神经网络对空间位置敏感，从而能够提升检测神经网络对空间位置的敏感性，进而增强检测神经网络对于一些特定缺陷类型的检测准确性，提升缺陷检测的准确度。
在一些可能的实施方式中，所述待检测图像的特征图还包括图像信息的特征图，所述处理单元用于：根据所述待检测图像，获取所述图像信息的特征图；根据所述图像信息的特征图，生成所述空间位置坐标信息的特征图，所述空间位置坐标信息的特征图的维度与所述图像信息的特征图的维度相同。
上述实施方式，通过先获取图像信息的特征图，再根据图像信息的特征图获取空间位置坐标信息的特征图，使得两类特征图的维度相同，从而便于后续图像处理，提升缺陷检测的准确度。
在一些可能的实施方式中，所述处理单元用于：生成所述空间位置坐标信息对应的线性值；根据所述线性值，生成第一坐标网络；根据所述图像信息的特征图，扩充所述第一坐标网络的维度，以生成所述空间位置坐标信息的特征图。
上述实施方式，通过生成空间位置坐标信息对应的坐标网络来生成空间位置坐标信息的特征图，从而将位置特征图拼接至了图像信息的特征图中，两者结合，从而提升了空间信息的敏感性，有利于提升缺陷检测的准确度。
在一些可能的实施方式中，所述处理单元用于：根据所述待检测图像的特征图，使用过滤器模型对所述待检测图像进行缺陷检测，其中， 所述过滤器模型包括用于对所述空间位置坐标信息的特征图进行处理的过滤器。
上述实施方式，对应于空间位置坐标信息的特征图，在后续检测中添加了相应的过滤器，从而便于缺陷检测，提升缺陷检测的准确度。
在一些可能的实施方式中，所述过滤器为第1个卷积层的过滤器。
在一些可能的实施方式中，所述空间位置坐标信息的特征图包括x轴方向坐标信息特征图、y轴方向坐标信息特征图以及z轴方向坐标信息特征图中的至少一项。
上述实施方式，通过定义空间位置坐标信息的特征图为x、y、z中的至少一个方法的特征图，提供了多种空间敏感类型，便于缺陷检测，提升缺陷检测的准确度。
在一些可能的实施方式中，所述方法用于极耳和/或极片的缺陷检测。
上述实施方式，在用于动力电池中的极耳/极片的缺陷检测时，能够提升动力电池的生产效率。
在一些可能的实施方式中，当所述装置用于极耳的缺陷检测时，所述极耳的缺陷检测包括极耳翻折特征的缺陷检测。
上述实施方式，由于极耳翻折特征的位置敏感性，将该方式用于翻折缺陷检测时，能够提升检测准确度。
在一些可能的实施方式中，所述处理单元用于：将所述待检测图像输入神经网络；通过所述神经网络的骨干网络进行缺陷特征向量提取以及缺陷特征向量对应的坐标信息特征提取，以获取所述待检测图像的特征图。
上述实施方式，通过同时提取缺陷的特征向量与对应坐标特征，能够提升缺陷检测的准确度。
第三方面，提供了一种缺陷检测的装置，包括处理器和存储器，所述存储器用于存储程序，所述处理器用于从所述存储器中调用并运行所述程序以执行上述第一方面或第一方面的任一可能的实施方式中的缺陷检测的方法。
第四方面，提供了一种计算机可读存储介质，包括计算机程序，当所述计算机程序在计算机上运行时，使得所述计算机执行上述第一方面或第一方面的任一可能的实施方式中的缺陷检测的方法。
第五方面，提供一种包含指令的计算机程序产品，该指令被计算机执行时使得该计算机执行上述第一方面或第一方面的任一可能的实现方式中的缺陷检测的方法。
附图说明
为了更清楚地说明本申请实施例的技术方案，下面将对本申请实施例中所需要使用的附图作简单地介绍，显而易见地，下面所描述的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据附图获得其他的附图。
图1是本申请提供的系统架构的结构示意图；
图2是一种动力电池的极耳的待检测图像；
图3是本申请一实施例公开的一种缺陷检测方法的示意性流程图；
图4是本申请一实施例公开的一种缺陷检测方法的示意性框图；
图5是本申请一实施例公开的一种缺陷检测装置的示意性结构框图；
图6是本申请一实施例公开的一种缺陷检测装置的硬件结构示意图。
具体实施方式
下面结合附图和实施例对本申请的实施方式作进一步详细描述。以下实施例的详细描述和附图用于示例性地说明本申请的原理，但不能用来限制本申请的范围，即本申请不限于所描述的实施例。
本申请实施例可适用于图像处理系统，包括但不限于基于红外成像的产品。该缺陷检测系统可以应用于具有缺陷检测装置的各种电子设备，该电子设备可以为个人计算机、计算机工作站、智能手机、平板电脑、智能摄像头、媒体消费设备、可穿戴设备、机顶盒、游戏机、增强现实(augmented reality，AR)AR/虚拟现实(virtual reality，VR)设备，车载终端等，本申请公开的实施例对此不做限定。
应理解，本文中的具体的例子只是为了帮助本领域技术人员更好地理解本申请实施例，而非限制本申请实施例的范围。
还应理解，在本申请的各种实施例中，各过程的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
还应理解，本说明书中描述的各种实施方式，既可以单独实施，也可以组合实施，本申请实施例对此并不限定。
除非另有说明，本申请实施例所使用的所有技术和科学术语与本申请的技术领域的技术人员通常理解的含义相同。本申请中所使用的术语只是为了描述具体的实施例的目的，不是旨在限制本申请的范围。本申请所使用的术语“和/或”包括一个或多个相关的所列项的任意的和所有的组合。
为了更好地理解本申请实施例的方案，下面先结合图1对本申请实施例可能的应用场景进行简单的介绍。
如图1所示，本申请实施例提供了一种系统架构100。在图1中， 数据采集设备160用于采集已知缺陷样本图像。针对本申请实施例的缺陷检测的方法来说，已知缺陷样本图像可以是包括具有一处或多处缺陷的样本图像，该样本图像中的缺陷类型已知。
在采集到已知缺陷样本图像之后，数据采集设备160将这些已知缺陷样本图像存入数据库130，训练设备120基于数据库130中维护的已知缺陷样本图像训练得到目标模型/规则101。
上述目标模型/规则101能够用于实现本申请实施例的缺陷检测的方法。本申请实施例中的目标模型/规则101具体可以为神经网络。需要说明的是，在实际的应用中，所述数据库130中维护的已知缺陷样本图像不一定都来自于数据采集设备160的采集，也有可能是从其他设备接收得到的。另外需要说明的是，训练设备120也不一定完全基于数据库130维护的已知缺陷样本图像进行目标模型/规则101的训练，也有可能从云端或其他地方获取已知缺陷样本图像进行模型训练，上述描述不应该作为对本申请实施例的限定。
根据训练设备120训练得到的目标模型/规则101可以应用于不同的系统或设备中，如应用于图1所示的执行设备110，所述执行设备110可以是终端，如手机终端，平板电脑，笔记本电脑等，还可以是服务器或者云端等。在图1中，执行设备110配置输入/输出(input/output，I/O)接口112，用于与外部设备进行数据交互，用户可以通过客户设备140向I/O接口112输入数据，所述输入数据在本申请实施例中可以包括：客户设备140输入的待检测图像。
在一些实施方式中，该客户设备140可以与上述执行设备110为同一设备，例如，客户设备140可以与上述执行设备110均为终端设备。
在另一些实施方式中，该客户设备140可以与上述执行设备110为不同设备，例如，客户设备140为终端设备，而执行设备110为云端、 服务器等设备，客户设备140可以通过任何通信机制/通信标准的通信网络与执行设备110进行交互，通信网络可以是广域网、局域网、点对点连接等方式，或它们的任意组合。
执行设备110的计算模块111用于根据I/O接口112接收到的输入数据(如待检测图像)进行处理。在执行设备110的计算模块111执行计算等相关的处理过程中，执行设备110可以调用数据存储系统150中的数据、代码等以用于相应的处理，也可以将相应处理得到的数据、指令等存入数据存储系统150中。
最后，I/O接口112将处理结果，如上述得到的缺陷分类结果返回给客户设备140，从而提供给用户。
值得说明的是，训练设备120可以针对不同的目标或称不同的任务，基于不同的训练数据生成相应的目标模型/规则101，该相应的目标模型/规则101即可以用于实现上述目标或完成上述任务，从而为用户提供所需的结果。
在图1中所示情况下，用户可以手动给定输入数据，该手动给定可以通过I/O接口112提供的界面进行操作。另一种情况下，客户设备140可以自动地向I/O接口112发送输入数据，如果要求客户设备140自动发送输入数据需要获得用户的授权，则用户可以在客户设备140中设置相应权限。用户可以在客户设备140查看执行设备110输出的结果，具体的呈现形式可以是显示、声音、动作等具体方式。客户设备140也可以作为数据采集端，采集如图所示输入I/O接口112的输入数据及输出I/O接口112的输出结果作为新的样本数据，并存入数据库130。当然，也可以不经过客户设备140进行采集，而是由I/O接口112直接将如图所示输入I/O接口112的输入数据及输出I/O接口112的输出结果，作为新的样本数据存入数据库130。
值得注意的是，图1仅是本申请实施例提供的一种系统架构的示意图，图中所示设备、器件、模块等之间的位置关系不构成任何限制，例如，在图1中，数据存储系统150相对执行设备110是外部存储器，在其它情况下，也可以将数据存储系统150置于执行设备110中。
如图1所示，根据训练设备120训练得到目标模型/规则101，该目标模型/规则101在本申请实施例中可以是神经网络，具体的，本申请实施例的神经网络可以为卷积神经网络(convolutional neuron network，CNN)、区域卷积神经网络(region CNN，RCNN)或者其它类型的神经网络等等，本申请对此不做具体限定。
在本申请的描述中，需要说明的是，除非另有说明，“多个”的含义是两个以上；术语“上”、“下”、“左”、“右”、“内”、“外”等指示的方位或位置关系仅是为了便于描述本申请和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本申请的限制。此外，术语“第一”、“第二”、“第三”等仅用于描述目的，而不能理解为指示或暗示相对重要性。“垂直”并不是严格意义上的垂直，而是在误差允许范围之内。“平行”并不是严格意义上的平行，而是在误差允许范围之内。
下述描述中出现的方位词均为图中示出的方向，并不是对本申请的具体结构进行限定。在本申请的描述中，还需要说明的是，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或一体地连接；可以是直接相连，也可以通过中间媒介间接相连。对于本领域的普通技术人员而言，可视具体情况理解上述术语在本申请中的具体含义。
动力电池不仅可以被应用于水利，风力、火力和太阳能电站等初等电源系统，还可以被广泛应用于电动自行车、电动摩托车、电动汽车等 电动交通工具，以及军事装备和航空航天等多个领域。随着动力电池应用领域不断扩大，其市场需求也在不断增加。动力电池生产的过程中，多个工序可能会出现不同的缺陷，例如阳极模切、阴极模切、卷绕和凹版等，需要应用视觉技术来自动检测，即缺陷检测。缺陷检测通常是指对物品表面缺陷的检测，表面缺陷检测检测是采用先进的机器视觉检测技术，对工件表面的斑点、凹坑、划痕、色差、缺损等缺陷进行检测。
然而，由于工件制备工艺、自身结构等原因，部分缺陷具有一定的位置敏感性，即对于工件的拍摄图片中显示的相似结构，在位于工件的不同位置时，一些位置上该相似结构属于缺陷，另一些位置上该相似结构不属于缺陷。例如，图2示出了一种动力电池的极耳的待检测图像，当需要对极耳的翻折缺陷进行检测，首先需要在图像中找到翻折缺陷的局部特征，如图2所示，包括两个基本相同的局部特征210和220，其中其中局部特征210位于待检测图像的尾部位置，是由于极耳的制备工艺引起的，属于正常形态，而局部特征220位于待检测图像的中部，线条逐渐消失，说明在极耳的中部出现了翻折缺陷，因此基本相同的两个局部特征210和220，一个属于正常形态，无需被检出，而另一个属于缺陷，需要被检出，此时，极耳的翻折缺陷是具有位置敏感性的。此外，不同位置的缺陷类型也可能是不同的。
传统的检测方法需要神经网络与逻辑后处理技术相结合来达到缺陷的位置敏感性的要求，检测效率较低。例如，由于神经网络无法学习到相同的一种局部特征，在什么位置该检出，什么位置不该检出，此时神经网络可以将其全部都检出，然后再通过人力对缺陷的形态进行标注，进行逻辑后处理，因此该方式检测效率较低。
鉴于此，本申请实施例提供了一种缺陷检测的方法，通过修改缺陷检测的神经网络结构，在检测的过程中同时提取空间位置坐标信息的特 征图，使得用于缺陷检测的神经网络对空间位置敏感，从而能够提升检测神经网络对空间位置的敏感性，进而增强检测神经网络对于一些特定缺陷类型的检测准确性，提升缺陷检测的准确度。
本申请实施例提供的缺陷检测方法和装置可以但不限于针对动力电池的缺陷检测，还可以应用于现代工业制造中的其他各类产品的缺陷检测。下面结合图3和4对本申请实施例的缺陷检测方法的主要过程进行介绍。
图3示出了本申请实施例的缺陷检测的方法300的流程示意图。该缺陷检测方法300包括以下步骤。
310，获取待检测图像。
在本申请实施例中，待检测图像为待检测工件的拍摄图像，即对该待检测工件进行缺陷检测时，首先可以对待检测工件进行拍摄，以获取待检测图像，例如可以对动力电池的极耳或极片进行拍摄，以获取极耳或极片的待检测图像，从而对极耳或极片的缺陷进行检测。其中用于拍摄图像的工具可以是电荷耦合元件(charge coupled device，CCD)相机拍摄的图片，也可以是其他相机或摄像机等拍摄的图片，本申请对此不作限定。
值得注意的是，在动力电池生产过程中，一般包括多个生产工序，当动力电池轮转到最后工序时，可以利用CCD相机拍摄动力电池以形成待检测图像。320，根据待检测图像，获取待检测图像的特征图，待检测图像的特征图包括空间位置坐标信息的特征图。
值得注意的是，在本申请实施例中，在获取待检测图像后，为了使得神经网络识别出待检测图像中是否存在缺陷以及缺陷所在位置等信息，需要对待检测图像进行处理，例如可以对待检测图像进行特征提取，以获取待检测图像的特征图，其中特征图中还包括空间位置坐标信息的特征图，从而便于后续缺陷检测时，结合缺陷所在位置进行检测。
330，根据待检测图像的特征图，对待检测图像进行缺陷检测。
值得注意的是，在本申请实施例中，在获取待检测图像的特征图后，可以对特征图进行进一步的处理，例如可以根据特征图中是否包括相应缺陷的特征，来判断待检测图像中是否存在缺陷，从而输出检测结果。还值得注意的是，上述步骤可以均由机器操作，而不需要人工操作，节省了人力，提高了生产效率。
因此，上述方法300通过在缺陷检测时，提取空间位置坐标信息的特征图，使得缺陷检测结合空间位置坐标信息，能够提升检测神经网络对空间位置的敏感性，进而提升缺陷检测的准确性。
在上述步骤310中，待检测图像可以是极耳或极片的拍摄图像，即对极耳或极片的进行缺陷检测。例如，上述方法300可以用于对极耳翻折缺陷进行检测。值得注意的是，对于极耳翻折缺陷，当翻折的相似特征出现于极耳图像的尾部时，其不属于缺陷，不应当被检出，而出现于极耳图像的中部时，属于缺陷，应当被检出。
在上述步骤320中，待检测图像的特征图包括空间位置坐标信息的特征图。可选的，空间位置坐标信息的特征图可以包括x轴方向坐标信息特征图、y轴方向坐标信息特征图以及z轴方向坐标信息特征图中的至少一项。
在上述步骤320中，待检测图像的特征图还可以包括图像信息的特征图。其中可选的，图像信息的特征图为待检测图像中的点、线、面特征和/或颜色特征等等。
在本申请实施例中，步骤320中根据待检测图像，获取待检测图像的特征图的方式可以是：首先，根据待检测图像获取其图像信息的特征图，例如，从待检测图像中提取点、线、面特征等；然后，根据图像信息的特征图，生成空间位置坐标信息的特征图，图像信息的特征图与空间位 置坐标信息的特征图的维度相同。
可选的，上述根据图像信息的特征图，生成空间位置坐标信息的特征图的方式可以是：生成空间位置坐标信息对应的线性值，例如生成对应于空间位置的从-1到1的线性值；根据该线性值，生成第一坐标网络，第一坐标网络可以为一维坐标网络或二维坐标网络；根据图像信息的特征图，扩充第一坐标网络的维度，以生成空间位置坐标信息的特征图，例如将第一坐标网络的维度扩充至与获取的点/线/面特征图相同的维度，从而形成了空间位置坐标信息的特征图，即获取到了位置特征。从而后续可以将点/线/面特征与位置特征进行联合，以获取下一个卷积层的输入。
在本申请实施例中，步骤320中获取待检测图像的特征图的方式可以是：将待检测图像输入神经网络；通过神经网络的骨干网络进行缺陷特征向量提取以及缺陷特征向量对应的坐标信息特征提取，以获取待检测图像的特征图。即，可以理解为待检测图像的特征图是通过神经网络的特征提取获得的，而神经网络的特征提取中可以同时提取缺陷特征以及缺陷对应的坐标特征。
在上述步骤330中，根据待检测图像的特征图，对待检测图像进行缺陷检测的模型可以是神经网络通过有效的缺陷样本训练获得的。
在上述步骤330中，可以根据待检测图像的特征图，使用过滤器模型对待检测图像进行缺陷检测，其中，过滤器模型包括用于对空间位置坐标信息的特征图进行处理的过滤器。或者换句话说，在缺陷检测过程中，添加一个或多个对应于空间位置坐标信息的过滤器。
可选的，过滤器为第1个卷积层的过滤器。即将第一个卷积层的输入通道数中增加了空间位置坐标信息的通道，从而与待检测图像的特征图相匹配。或者可选的，对应于空间位置坐标信息的过滤器还可以是其他卷积层的过滤器，例如第i层，i为大于0的正整数。
在本申请实施例中，当上述方法300用于极耳的缺陷检测时，输出的检测结果可以是有缺陷的极耳序号以及缺陷类别。
为了便于理解上述方法300，以下结合图4对本申请实施例提供的缺陷检测方法进行进一步介绍，即图4示出了本申请实施例的缺陷检测的方法400的示意性框图。
需要说明的是，图4包括与方法300中相同的步骤，可参考上述方法300中的相关介绍，此处不再赘述。
如图4所示，首先，获取待检测图像410后，对待检测图像410进行特征提取，获取图像信息的特征图421、x轴方向坐标信息特征图422以及y轴方向坐标信息特征图423，其中图像信息的特征图421的大小为(h，w，c)，x轴方向坐标信息特征图422的大小为(h，w，1)，y轴方向坐标信息特征图423的大小为(h，w，1)，其中h表示特征图的高度(height)，w表示特征图的宽度(width)，c和1均表示特征图的数量；然后，将图像信息的特征图421、x轴方向坐标信息特征图422以及y轴方向坐标信息特征图423进行联合，以获取特征图420，特征图420的大小为(h，w，c+2)，或者说，在图像信息的特征图421的c个通道的后面进行拼接通道的操作，将x轴方向坐标信息的通道图以及y轴方向坐标信息的通道图拼接至图像信息的c张通道图后，以获取c+2张通道图，即特征图420；最后，对获取的特征图420进行缺陷检测，以输出检测结果430，例如，可以将获取的特征图420输入至检测器(detector)中进行缺陷检测，获得检测神经网络输出的检测结果430。
值得注意的是，上述获取待检测图像410、进行特征提取获取特征图420、进行缺陷检测获取/输出检测结果430的更具体的实施方式可分别参考上述方法300中的步骤310、320、330，此处不再赘述。
可选的，在本申请实施例中，在上述图像信息的特征图与空间位 置坐标信息特征图进行联合的方式中，若仅需要x方向的空间坐标位置信息，则拼接一张x轴方向坐标信息的通道图即可，若仅需要y方向的空间坐标位置信息，则拼接一张y轴方向坐标信息的通道图即可，若x方向的空间坐标位置信息和y方向的空间坐标位置信息均需要，则拼接两张x轴方向坐标信息的通道图和y轴方向坐标信息的通道图，在拼接一张通道图后，特征图420的大小为(h，w，c+1)，在拼接两张通道图后，特征图420的大小为(h，w，c+2)，类似地，在拼接三张通道图后，特征图420的大小为(h，w，c+3)，本申请对拼接的通道图的数量不作限定。
值得注意的是，在本申请实施例中，当特征图420经过适配优化后，包括空间坐标信息特征图时，对应的检测器也需要相应的修改其结构，以适应适配优化后的特征图。例如，当适配优化后的特征图的大小由(h，w，c)变为(h，w，c+2)后，相应的卷积层中过滤器的数量也需要多2个。
上文详细地描述了本申请实施例的方法实施例，下面描述本申请实施例的装置实施例，装置实施例与方法实施例相互对应，因此未详细描述的部分可参见前面方法实施例，装置可以实现上述方法中任意可能实现的方式。
图5示出了本申请一个实施例的缺陷检测的装置500的示意性框图。该装置500可以执行上述本申请实施例的缺陷检测的方法，例如，该装置500可以为前述执行设备110。
如图5所示，该装置包括：
获取单元520，用于获取待检测图像。
处理单元520，用于根据所述待检测图像，获取所述待检测图像的特征图，所述待检测图像的特征图包括空间位置坐标信息的特征图，还用于根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测。
关于上述装置500的更详细功能，可参考上述方法实施中的相关描述，此处不再赘述。
图6是本申请实施例的缺陷检测的装置的硬件结构示意图。图6所示的缺陷检测的装置600包括存储器601、处理器602、通信接口603以及总线604。其中，存储器601、处理器602、通信接口603通过总线604实现彼此之间的通信连接。
存储器601可以是只读存储器(read-only memory，ROM)，静态存储设备和随机存取存储器(random access memory，RAM)。存储器601可以存储程序，当存储器601中存储的程序被处理器602执行时，处理器602和通信接口603用于执行本申请实施例的缺陷检测的方法的各个步骤。
处理器602可以采用通用的中央处理器(central processing unit，CPU)，微处理器，应用专用集成电路(application specific integrated circuit，ASIC)，图形处理器(graphics processing unit，GPU)或者一个或多个集成电路，用于执行相关程序，以实现本申请实施例的缺陷检测的装置中的单元所需执行的功能，或者执行本申请实施例的缺陷检测的方法。
处理器602还可以是一种集成电路芯片，具有信号的处理能力。在实现过程中，本申请实施例的缺陷检测的方法的各个步骤可以通过处理器602中的硬件的集成逻辑电路或者软件形式的指令完成。
上述处理器602还可以是通用处理器、数字信号处理器(digital signal processing，DSP)、ASIC、现成可编程门阵列(field programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为 硬件处理器执行完成，或者用处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器601，处理器602读取存储器601中的信息，结合其硬件完成本申请实施例的缺陷检测的装置中包括的单元所需执行的功能，或者执行本申请实施例的缺陷检测的方法。
通信接口603使用例如但不限于收发器一类的收发装置，来实现装置600与其他设备或通信网络之间的通信。例如，可以通过通信接口603获取未知设备的流量数据。
总线604可包括在装置600各个部件(例如，存储器601、处理器602、通信接口603)之间传送信息的通路。
应注意，尽管上述装置600仅仅示出了存储器、处理器、通信接口，但是在具体实现过程中，本领域的技术人员应当理解，装置600还可以包括实现正常运行所必须的其他器件。同时，根据具体需要，本领域的技术人员应当理解，装置600还可包括实现其他附加功能的硬件器件。此外，本领域的技术人员应当理解，装置600也可仅仅包括实现本申请实施例所必须的器件，而不必包括图6中所示的全部器件。
本申请实施例还提供了一种计算机可读存储介质，存储用于设备执行的程序代码，程序代码包括用于执行上述缺陷检测的方法中的步骤的指令。
本申请实施例还提供了一种计算机程序产品，所述计算机程序产品包括存储在计算机可读存储介质上的计算机程序，所述计算机程序包括程序指令，当所述程序指令被计算机执行时，使所述计算机执行上述缺陷检测的方法。
上述的计算机可读存储介质可以是暂态计算机可读存储介质，也 可以是非暂态计算机可读存储介质。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的装置的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
本申请中使用的用词仅用于描述实施例并且不用于限制权利要求。如在实施例以及权利要求的描述中使用的，除非上下文清楚地表明，否则单数形式的“一个”和“所述”旨在同样包括复数形式。类似地，如在本申请中所使用的术语“和/或”是指包含一个或一个以上相关联的列出的任何以及所有可能的组合。另外，当用于本申请中时，术语“包括”指陈述的特征、整体、步骤、操作、元素，和/或组件的存在，但不排除一个或一个以上其它特征、整体、步骤、操作、元素、组件和/或这些的分组的存在或添加。
所描述的实施例中的各方面、实施方式、实现或特征能够单独使用或以任意组合的方式使用。所描述的实施例中的各方面可由软件、硬件或软硬件的结合实现。所描述的实施例也可以由存储有计算机可读代码的计算机可读介质体现，该计算机可读代码包括可由至少一个计算装置执行的指令。所述计算机可读介质可与任何能够存储数据的数据存储装置相关联，该数据可由计算机系统读取。用于举例的计算机可读介质可以包括只 读存储器、随机存取存储器、紧凑型光盘只读储存器(Compact Disc Read-Only Memory，CD-ROM)、硬盘驱动器(Hard Disk Drive，HDD)、数字视频光盘(Digital Video Disc，DVD)、磁带以及光数据存储装置等。所述计算机可读介质还可以分布于通过网络联接的计算机系统中，这样计算机可读代码就可以分布式存储并执行。
上述技术描述可参照附图，这些附图形成了本申请的一部分，并且通过描述在附图中示出了依照所描述的实施例的实施方式。虽然这些实施例描述的足够详细以使本领域技术人员能够实现这些实施例，但这些实施例是非限制性的；这样就可以使用其它的实施例，并且在不脱离所描述的实施例的范围的情况下还可以做出变化。比如，流程图中所描述的操作顺序是非限制性的，因此在流程图中阐释并且根据流程图描述的两个或两个以上操作的顺序可以根据若干实施例进行改变。作为另一个例子，在若干实施例中，在流程图中阐释并且根据流程图描述的一个或一个以上操作是可选的，或是可删除的。另外，某些步骤或功能可以添加到所公开的实施例中，或两个以上的步骤顺序被置换。所有这些变化被认为包含在所公开的实施例以及权利要求中。
另外，上述技术描述中使用术语以提供所描述的实施例的透彻理解。然而，并不需要过于详细的细节以实现所描述的实施例。因此，实施例的上述描述是为了阐释和描述而呈现的。上述描述中所呈现的实施例以及根据这些实施例所公开的例子是单独提供的，以添加上下文并有助于理解所描述的实施例。上述说明书不用于做到无遗漏或将所描述的实施例限制到本申请的精确形式。根据上述教导，若干修改、选择适用以及变化是可行的。在某些情况下，没有详细描述为人所熟知的处理步骤以避免不必要地影响所描述的实施例。虽然已经参考优选实施例对本申请进行了描述，但在不脱离本申请的范围的情况下，可以对其进行各种改进并且可以用等 效物替换其中的部件。尤其是，只要不存在结构冲突，各个实施例中所提到的各项技术特征均可以任意方式组合起来。本申请并不局限于文中公开的特定实施例，而是包括落入权利要求的范围内的所有技术方案。

Claims (12)

  1. 一种缺陷检测的方法，其特征在于，包括：
    获取待检测图像；
    根据所述待检测图像，获取所述待检测图像的特征图，所述待检测图像的特征图包括空间位置坐标信息的特征图；
    根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测。
  2. 根据权利要求1所述的方法，其特征在于，所述待检测图像的特征图还包括图像信息的特征图，所述根据所述待检测图像，获取所述待检测图像的特征图，包括：
    根据所述待检测图像，获取所述图像信息的特征图；
    根据所述图像信息的特征图，生成所述空间位置坐标信息的特征图，所述空间位置坐标信息的特征图的维度与所述图像信息的特征图的维度相同。
  3. 根据权利要求2所述的方法，其特征在于，所述根据所述图像信息的特征图，生成所述空间位置坐标信息的特征图，包括：
    生成所述空间位置坐标信息对应的线性值；
    根据所述线性值，生成第一坐标网络；
    根据所述图像信息的特征图，扩充所述第一坐标网络的维度，以生成所述空间位置坐标信息的特征图。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，所述根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测，包括：
    根据所述待检测图像的特征图，使用过滤器模型对所述待检测图像进行缺陷检测，其中，所述过滤器模型包括用于对所述空间位置坐标信息的特征图进行处理的过滤器。
  5. 根据权利要求4所述的方法，其特征在于，所述过滤器为第1个卷积层的过滤器。
  6. 根据权利要求1-5中任一项所述的方法，其特征在于，所述空间位置坐标信息的特征图包括x轴方向坐标信息特征图、y轴方向坐标信息特征图以及z轴方向坐标信息特征图中的至少一项。
  7. 根据权利要求1-6中任一项所述的方法，其特征在于，所述方法用于极耳和/或极片的缺陷检测。
  8. 根据权利要求7所述的方法，其特征在于，当所述方法用于极耳的缺陷检测时，所述极耳的缺陷检测包括极耳翻折特征的缺陷检测。
  9. 根据权利要求1-8中任一项所述的方法，其特征在于，所述根据所述待检测图像，获取所述待检测图像的特征图，包括：
    将所述待检测图像输入神经网络；
    通过所述神经网络的骨干网络进行缺陷特征向量提取以及缺陷特征向量对应的坐标信息特征提取，以获取所述待检测图像的特征图。
  10. 一种缺陷检测的装置，其特征在于，包括：
    获取单元，用于获取待检测图像；
    处理单元，用于根据所述待检测图像，获取所述待检测图像的特征图，所述待检测图像的特征图包括空间位置坐标信息的特征图；
    所述处理单元还用于根据所述待检测图像的特征图，对所述待检测图像进行缺陷检测。
  11. 一种缺陷检测的装置，其特征在于，包括处理器和存储器，所述存储器用于存储程序，所述处理器用于从所述存储器中调用并运行所述程序以执行权利要求1至9中任一项所述的缺陷检测的方法。
  12. 一种计算机可读存储介质，其特征在于，包括计算机程序，当所述计算机程序在计算机上运行时，使得所述计算机执行权利要求1至9中任一项所述的缺陷检测的方法。
PCT/CN2022/115774 2022-08-30 2022-08-30 缺陷检测的方法、装置和计算机可读存储介质 WO2024044947A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
PCT/CN2022/115774 WO2024044947A1 (zh) 2022-08-30 2022-08-30 缺陷检测的方法、装置和计算机可读存储介质
CN202280006731.XA CN117957438A (zh) 2022-08-30 2022-08-30 缺陷检测的方法、装置和计算机可读存储介质
EP22871168.5A EP4357765A4 (en) 2022-08-30 2022-08-30 ERROR DETECTION METHOD AND APPARATUS AND COMPUTER-READABLE STORAGE MEDIUM
US18/356,232 US20240070840A1 (en) 2022-08-30 2023-07-21 Defect detection method and apparatus, and computer-readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/115774 WO2024044947A1 (zh) 2022-08-30 2022-08-30 缺陷检测的方法、装置和计算机可读存储介质

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/356,232 Continuation US20240070840A1 (en) 2022-08-30 2023-07-21 Defect detection method and apparatus, and computer-readable storage medium

Publications (1)

Publication Number Publication Date
WO2024044947A1 true WO2024044947A1 (zh) 2024-03-07

Family

ID=89997420

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/115774 WO2024044947A1 (zh) 2022-08-30 2022-08-30 缺陷检测的方法、装置和计算机可读存储介质

Country Status (4)

Country Link
US (1) US20240070840A1 (zh)
EP (1) EP4357765A4 (zh)
CN (1) CN117957438A (zh)
WO (1) WO2024044947A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110136116A (zh) * 2019-05-15 2019-08-16 广东工业大学 一种注塑泵缺陷检测方法、装置、设备及存储介质
CN110349133A (zh) * 2019-06-25 2019-10-18 杭州汇萃智能科技有限公司 物体表面缺陷检测方法、装置
CN113591840A (zh) * 2021-06-30 2021-11-02 北京旷视科技有限公司 一种目标检测方法、装置、设备和存储介质
CN113781434A (zh) * 2021-09-10 2021-12-10 深圳市高川自动化技术有限公司 缺陷检测方法、装置、智能终端及计算机可读存储介质
WO2022121531A1 (zh) * 2020-12-09 2022-06-16 歌尔股份有限公司 一种产品缺陷检测方法和装置
CN114648480A (zh) * 2020-12-17 2022-06-21 杭州海康威视数字技术股份有限公司 表面缺陷检测方法、装置及系统
US20220198228A1 (en) * 2020-12-22 2022-06-23 Hon Hai Precision Industry Co., Ltd. Method for detecting defects in multi-scale images and computing device utilizing method

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102586069B1 (ko) * 2015-07-03 2023-10-05 에씰로 앙터나시오날 증강 현실을 위한 방법 및 시스템
JP6358351B1 (ja) * 2017-03-21 2018-07-18 Ｊｆｅスチール株式会社 表面欠陥検査方法及び表面欠陥検査装置
CN111787323B (zh) * 2020-05-23 2021-09-03 清华大学 一种基于对抗学习的可变比特率生成式压缩方法
US20220027728A1 (en) * 2020-07-21 2022-01-27 Samsung Electronics Co., Ltd. Method and apparatus with image correspondence
US11307150B2 (en) * 2020-08-17 2022-04-19 Applied Materials Israel Ltd. Automatic optimization of an examination recipe
CN116092096A (zh) * 2021-11-05 2023-05-09 同方威视技术股份有限公司 用于检验申报信息真实性的方法、系统、设备及介质
WO2023097639A1 (zh) * 2021-12-03 2023-06-08 宁德时代新能源科技股份有限公司 用于图像分割的数据标注方法和系统以及图像分割装置
EP4220552A4 (en) * 2021-12-07 2023-12-27 Contemporary Amperex Technology Co., Limited IMAGE PROCESSING METHOD AND SYSTEM

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110136116A (zh) * 2019-05-15 2019-08-16 广东工业大学 一种注塑泵缺陷检测方法、装置、设备及存储介质
CN110349133A (zh) * 2019-06-25 2019-10-18 杭州汇萃智能科技有限公司 物体表面缺陷检测方法、装置
WO2022121531A1 (zh) * 2020-12-09 2022-06-16 歌尔股份有限公司 一种产品缺陷检测方法和装置
CN114648480A (zh) * 2020-12-17 2022-06-21 杭州海康威视数字技术股份有限公司 表面缺陷检测方法、装置及系统
US20220198228A1 (en) * 2020-12-22 2022-06-23 Hon Hai Precision Industry Co., Ltd. Method for detecting defects in multi-scale images and computing device utilizing method
CN113591840A (zh) * 2021-06-30 2021-11-02 北京旷视科技有限公司 一种目标检测方法、装置、设备和存储介质
CN113781434A (zh) * 2021-09-10 2021-12-10 深圳市高川自动化技术有限公司 缺陷检测方法、装置、智能终端及计算机可读存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4357765A4 *

Also Published As

Publication number Publication date
US20240070840A1 (en) 2024-02-29
EP4357765A4 (en) 2024-05-08
CN117957438A (zh) 2024-04-30
EP4357765A1 (en) 2024-04-24

Similar Documents

Publication Publication Date Title
US9727775B2 (en) Method and system of curved object recognition using image matching for image processing
US10645364B2 (en) Dynamic calibration of multi-camera systems using multiple multi-view image frames
CN102509304A (zh) 基于智能优化的摄像机标定方法
WO2021136386A1 (zh) 数据处理方法、终端和服务器
CN111679142A (zh) 一种便携式输变电设备红外智能诊断装置及方法
CN115797359B (zh) 基于电路板上锡膏的检测方法、设备和存储介质
WO2022002262A1 (zh) 基于计算机视觉的字符序列识别方法、装置、设备和介质
CN115829921B (zh) 检测电芯缺陷的方法、装置和计算机可读存储介质
CN104202448A (zh) 一种解决移动终端摄像头拍照亮度不均的系统及其方法
CN116661477A (zh) 一种变电站无人机巡检方法、装置、设备及存储介质
Chu et al. OA-BEV: Bringing Object Awareness to Bird's-Eye-View Representation for Multi-Camera 3D Object Detection
CN114549393A (zh) 图像标注方法、装置、设备及计算机存储介质
CN116935013B (zh) 基于三维重建的电路板点云大范围拼接方法及系统
WO2024044947A1 (zh) 缺陷检测的方法、装置和计算机可读存储介质
CN103559706A (zh) 一种高速图像识别定位信息处理机及处理方法
CN111563895A (zh) 一种图片清晰度确定方法、装置、设备及存储介质
CN115861162A (zh) 定位目标区域的方法、装置及计算机可读存储介质
CN114821482A (zh) 一种基于鱼眼探头的矢量拓扑一体化客流计算方法及系统
CN115100296A (zh) 一种光伏组件故障定位方法、装置、设备及存储介质
CN112800806B (zh) 物体位姿检测追踪方法、装置、电子设备及存储介质
CN107633498B (zh) 图像暗态增强方法、装置及电子设备
CN113810665A (zh) 视频处理方法、装置、设备、存储介质及产品
WO2024000420A1 (zh) 缺陷检测方法和装置
CN114503160A (zh) 一种对象关联方法及装置、系统、电子设备、存储介质和计算机程序
CN116182807B (zh) 一种姿态信息确定方法、装置、电子设备、系统和介质

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 202280006731.X

Country of ref document: CN

ENP Entry into the national phase

Ref document number: 2022871168

Country of ref document: EP

Effective date: 20230331