CN116703874A - Target detection method, device and storage medium - Google Patents
Target detection method, device and storage medium Download PDFInfo
- Publication number
- CN116703874A CN116703874A CN202310715788.3A CN202310715788A CN116703874A CN 116703874 A CN116703874 A CN 116703874A CN 202310715788 A CN202310715788 A CN 202310715788A CN 116703874 A CN116703874 A CN 116703874A
- Authority
- CN
- China
- Prior art keywords
- defect detection
- detection
- pixel point
- pixel
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 324
- 238000003860 storage Methods 0.000 title claims abstract description 11
- 230000007547 defect Effects 0.000 claims abstract description 265
- 238000013450 outlier detection Methods 0.000 claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000003709 image segmentation Methods 0.000 claims description 47
- 238000000034 method Methods 0.000 claims description 31
- 230000002950 deficient Effects 0.000 claims description 22
- 238000000605 extraction Methods 0.000 claims description 20
- 239000013598 vector Substances 0.000 claims description 19
- 230000000007 visual effect Effects 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000004512 die casting Methods 0.000 description 24
- 239000010410 layer Substances 0.000 description 21
- 238000012545 processing Methods 0.000 description 12
- 238000004891 communication Methods 0.000 description 10
- 239000011229 interlayer Substances 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 238000002372 labelling Methods 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The disclosure relates to a target detection method, a device and a storage medium, which can acquire a target image of an object to be detected; performing outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defect pixel point or not; and performing defect detection on the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
Description
Technical Field
The present disclosure relates to the field of computer vision, and in particular, to a target detection method, apparatus, and storage medium.
Background
Die casting is a piece that is die cast. The production process of the die casting is complex, and a plurality of defects (such as cold shut, shrinkage cavity, crack, dent and the like) are inevitably generated on the surface of the die casting. Clearly, the presence of defective die castings affects the quality of the products subsequently produced on the basis of the die castings, and it is therefore necessary to detect defects on the die casting surfaces.
Disclosure of Invention
In order to overcome the problems in the related art, the present disclosure provides a target detection method, apparatus, and storage medium.
According to a first aspect of an embodiment of the present disclosure, there is provided a target detection method, including:
acquiring a target image of an object to be detected;
performing outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defect pixel point or not;
and performing defect detection on the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
Optionally, the performing outlier detection on the target image to obtain a defect detection hint point includes:
extracting features of the target image to obtain a first feature map of the object to be detected;
and detecting outliers of the first feature map to obtain the defect detection prompt points.
Optionally, the performing outlier detection on the first feature map to obtain the defect detection hint point includes:
acquiring a feature vector corresponding to each first pixel point on the first feature map;
Determining an outlier score corresponding to each first pixel point respectively through a preset outlier detection model according to the feature vector corresponding to each first pixel point respectively, wherein the outlier score represents whether the corresponding first pixel point belongs to the outlier pixel point;
and determining the defect detection prompt points according to the outlier scores corresponding to the first pixel points respectively.
Optionally, the determining the defect detection hint point according to the outlier score corresponding to each first pixel point includes:
upsampling the first feature map to obtain a second feature map having the same size as the target image;
for each first pixel point, determining the discrete fraction of a second pixel point on the second feature map according to the discrete fraction of the first pixel point, wherein the second pixel point is a pixel point corresponding to the first pixel point;
and determining the defect detection prompt point according to the outlier score of each second pixel point on the second feature map.
Optionally, the defect detection prompting points comprise a defect detection positive sample prompting point and a defect detection negative sample prompting point; the determining the defect detection prompting point according to the outlier score of each second pixel point on the second feature map comprises:
Acquiring a first preset defect detection proportion;
and after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the first preset defect detection proportion in the ranking result from big to small as the defect detection positive sample prompting points, and selecting the second pixel points with the first preset defect detection proportion in the ranking result from small to big as the defect detection negative sample prompting points.
Optionally, the extracting the features of the target image to obtain a first feature map of the object to be detected includes:
inputting the target image into a feature extraction model obtained by training in advance, and then obtaining image features respectively output by at least one preset middle layer of the feature extraction model;
splicing the image features respectively output by at least one preset intermediate layer according to the feature channels to obtain splicing features;
and randomly sampling the spliced features to obtain the first feature map.
Optionally, the performing defect detection on the object to be detected according to the defect detection prompt point through a target detection model obtained by pre-training includes:
Performing defect detection on the object to be detected by executing a defect detection step;
the defect detection step includes:
acquiring target positive sample prompting points with preset number from the defect detection positive sample prompting points;
obtaining target negative sample prompting points with preset number from the defect detection negative sample prompting points;
after a preset number of target positive sample prompting points and a preset number of target negative sample prompting points are input into the target detection model, determining a target image area corresponding to the defect on the object to be detected according to an image segmentation area output by the model.
Optionally, the determining, according to the image segmentation area output by the model, the target image area corresponding to the defect on the object to be detected includes:
after the defect detection step is performed a first preset number of times, the intersection of the image division areas obtained by each detection is taken as the target image area.
Optionally, the determining, according to the image segmentation area output by the model, the target image area corresponding to the defect on the object to be detected includes:
after the defect detection step of the second preset times is executed, the image segmentation area obtained by each detection is obtained, and the second preset times are larger than the first preset times;
For each pixel point in the image segmentation area, determining a target pixel value corresponding to the pixel point according to the pixel value of the pixel point in the image segmentation area obtained by each detection;
and determining the target image area according to the target pixel value of each pixel point in the image segmentation area.
Optionally, the determining, according to the pixel value of the pixel point in the image segmentation area obtained by each detection, the target pixel value corresponding to the pixel point includes:
counting the times that the pixel value of the pixel point is a first appointed pixel value;
when the number of times is greater than or equal to a preset number of times threshold, the first appointed pixel value is used as the target pixel value corresponding to the pixel point;
and under the condition that the times are smaller than the preset times threshold, taking a second designated pixel value as the target pixel value corresponding to the pixel point, wherein the first designated pixel value represents that the corresponding pixel point is a defective pixel point, and the second designated pixel value represents that the corresponding pixel point is a non-defective pixel point.
Optionally, the determining, according to the image segmentation area output by the model, the target image area corresponding to the defect on the object to be detected includes:
Obtaining a second preset defect detection proportion, wherein the second preset defect detection proportion is larger than the first preset defect detection proportion;
after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the second preset defect detection proportion in the ranking result as mask pixel points according to the order from small to large, and taking an image area corresponding to the mask pixel points as an image mask, wherein the defects do not exist in the image area corresponding to the image mask;
and determining the target image area according to the image segmentation area output by the model and the image mask.
Optionally, the object detection model comprises a visual large model SAM.
According to a second aspect of embodiments of the present disclosure, there is provided an object detection apparatus including:
the acquisition module is configured to acquire a target image of an object to be detected;
the first detection module is configured to perform outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defective pixel point or not;
And the second detection module is configured to detect the defects of the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
Optionally, the first detection module is configured to perform feature extraction on the target image to obtain a first feature map of the object to be detected; and detecting outliers of the first feature map to obtain the defect detection prompt points.
Optionally, the first detection module is configured to obtain a feature vector corresponding to each first pixel point on the first feature map; determining an outlier score corresponding to each first pixel point respectively through a preset outlier detection model according to the feature vector corresponding to each first pixel point respectively, wherein the outlier score represents whether the corresponding first pixel point belongs to the outlier pixel point; and determining the defect detection prompt points according to the outlier scores corresponding to the first pixel points respectively.
Optionally, the first detection module is configured to upsample the first feature map to obtain a second feature map with the same size as the target image; for each first pixel point, determining the discrete fraction of a second pixel point on the second feature map according to the discrete fraction of the first pixel point, wherein the second pixel point is a pixel point corresponding to the first pixel point; and determining the defect detection prompt point according to the outlier score of each second pixel point on the second feature map.
Optionally, the defect detection prompting points comprise a defect detection positive sample prompting point and a defect detection negative sample prompting point; the first detection module is configured to obtain a first preset defect detection proportion;
and after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the first preset defect detection proportion in the ranking result from big to small as the defect detection positive sample prompting points, and selecting the second pixel points with the first preset defect detection proportion in the ranking result from small to big as the defect detection negative sample prompting points.
Optionally, the first detection module is configured to input the target image into a feature extraction model obtained by training in advance, and then obtain image features respectively output by at least one preset middle layer of the feature extraction model; splicing the image features respectively output by at least one preset intermediate layer according to the feature channels to obtain splicing features; and randomly sampling the spliced features to obtain the first feature map.
Optionally, the second detection module is configured to perform defect detection on the object to be detected by performing a defect detection step;
The defect detection step includes:
acquiring target positive sample prompting points with preset number from the defect detection positive sample prompting points;
obtaining target negative sample prompting points with preset number from the defect detection negative sample prompting points;
after a preset number of target positive sample prompting points and a preset number of target negative sample prompting points are input into the target detection model, determining a target image area corresponding to the defect on the object to be detected according to an image segmentation area output by the model.
Optionally, the second detection module is configured to take an intersection of the image segmentation areas obtained by each detection as the target image area after performing the defect detection step for a first preset number of times.
Optionally, the second detection module is configured to obtain the image segmentation area obtained by each detection after performing the defect detection step of a second preset number of times, where the second preset number of times is greater than the first preset number of times; for each pixel point in the image segmentation area, determining a target pixel value corresponding to the pixel point according to the pixel value of the pixel point in the image segmentation area obtained by each detection; and determining the target image area according to the target pixel value of each pixel point in the image segmentation area.
Optionally, the second detection module is configured to count the number of times that the pixel value of the pixel point is the first specified pixel value; when the number of times is greater than or equal to a preset number of times threshold, the first appointed pixel value is used as the target pixel value corresponding to the pixel point; and under the condition that the times are smaller than the preset times threshold, taking a second designated pixel value as the target pixel value corresponding to the pixel point, wherein the first designated pixel value represents that the corresponding pixel point is a defective pixel point, and the second designated pixel value represents that the corresponding pixel point is a non-defective pixel point.
Optionally, the second detection module is configured to obtain a second preset defect detection proportion, where the second preset defect detection proportion is greater than the first preset defect detection proportion; after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the second preset defect detection proportion in the ranking result as mask pixel points according to the order from small to large, and taking an image area corresponding to the mask pixel points as an image mask, wherein the defects do not exist in the image area corresponding to the image mask; and determining the target image area according to the image segmentation area output by the model and the image mask.
Optionally, the object detection model comprises a visual large model SAM.
According to a third aspect of the embodiments of the present disclosure, there is provided an object detection apparatus including:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
acquiring a target image of an object to be detected;
performing outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defect pixel point or not;
and performing defect detection on the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
According to a fourth aspect of embodiments of the present disclosure, there is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the steps of the object detection method provided by the first aspect of the present disclosure.
The technical scheme provided by the embodiment of the disclosure can comprise the following beneficial effects: through carrying out outlier detection on the target image, a defect detection prompt point is obtained, so that a subsequent target detection model can carry out defect detection based on the defect detection prompt point, thereby avoiding the need of manually providing defect prompt, realizing automatic defect detection on an object to be detected, and improving detection efficiency. Meanwhile, the target detection method provided by the disclosure does not need to label a large amount of defect data, so that the detection cost is saved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure.
Fig. 1 is a flow chart illustrating a method of object detection according to an exemplary embodiment.
Fig. 2 is a flow chart of another object detection method according to the embodiment shown in fig. 1.
Fig. 3 is a flow chart of a target detection method according to the embodiment shown in fig. 2.
FIG. 4 is a flow chart illustrating a method of target detection according to the embodiment shown in FIG. 3.
Fig. 5 is a block diagram illustrating an object detection apparatus according to an exemplary embodiment.
Fig. 6 is a block diagram illustrating an apparatus 800 for target detection according to an example embodiment.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
It should be noted that, all actions for acquiring signals, information or data in the present disclosure are performed under the condition of conforming to the corresponding data protection rule policy of the country of the location and obtaining the authorization given by the owner of the corresponding device.
The method and the device are mainly applied to scenes for carrying out defect detection on the object to be detected based on the computer vision technology. The object to be inspected may include, for example, a component (such as a die cast) produced on a production line. Taking a die casting as an example, when defect detection is performed on the die casting after production, an X-ray image of the die casting can be generally shot on a production line, and then a defect detection result of the die casting is obtained based on the X-ray image through a computer vision detection algorithm installed on an industrial personal computer.
Existing defect detection schemes mainly include two classes: detection methods based on traditional visual detection methods and detection methods based on visual supervision. The detection result is mainly obtained through methods such as image threshold segmentation, edge detection, outlier detection and the like based on the traditional visual detection method, but the methods are easily affected by complex background and are difficult to apply to defect detection of die castings; the detection method based on visual supervision mainly comprises the steps of target detection, image segmentation and the like, and finally realizes defect detection through image characteristics of data learning defects, however, the method generally requires a large amount of defect data labeling, and the labeling cost is higher.
In addition, in the field of computer vision, large visual models (e.g., segment Anything Model, SAM) can be made to detect and segment objects in images without sample labeling. However, the scarce semantics of such large models in nature make accurate segmentation difficult, for example, defect detection of complex die castings. The existing method can provide prompts for the visual large model SAM through a manual interaction method, for example, through appointing pixel points, boundary boxes, masks, texts and the like of targets in an image, so that accurate segmentation is realized. However, the method relies on manual interaction, and cannot be used for automatically detecting defects of the die castings, so that the detection efficiency is low, and the method is difficult to be applied to actual production links.
In order to solve the above-mentioned problems, the present disclosure provides a target detection method, apparatus and storage medium. The following detailed description of specific embodiments of the present disclosure refers to the accompanying drawings.
Fig. 1 is a flowchart illustrating a target detection method according to an exemplary embodiment, which may be applied to an industrial personal computer on a production line of an object to be detected, as shown in fig. 1, the method includes the steps of:
in step S101, a target image of an object to be detected is acquired.
Wherein the target image may be an X-ray image. In this step, an X-ray image of the object to be detected may be acquired by an X-ray image acquisition device.
After the X-ray image of the object to be detected is obtained, gray-scale processing can be performed on the X-ray image to obtain a gray-scale image, and subsequent feature extraction and defect detection can be performed based on the gray-scale image.
In step S102, outlier detection is performed on the target image to obtain a defect detection hint point, where the defect detection hint point is a pixel point on the target image that has defect detection information, and the defect detection information is used to indicate whether the pixel point is a defective pixel point.
In the present disclosure, the defect detection hint points may include a defect detection positive sample hint point and a defect detection negative sample hint point, where the defect detection positive sample hint point is a pixel point on the target image that belongs to a defect of the object to be detected, and the defect detection negative sample hint point is a pixel point on the target image that does not belong to a defect of the object to be detected.
In step S103, performing defect detection on the object to be detected according to the defect detection prompt point through a target detection model obtained by training in advance.
In this step, after the defect detection prompt point is input to the target detection model, a defect detection result of the object to be detected is output through the target detection model, where the defect detection result may be a defect area marked on the target object.
In addition, the object detection model may include a large visual model (also referred to as an "image segmentation large model" or "visual basic model"), for example, may be a SAM model, which may segment the object in the image without labeling, and through which the SAM may obtain relatively accurate defect detection results.
By adopting the method, the defect detection prompt point is obtained by carrying out outlier detection on the target image, so that the subsequent target detection model can carry out defect detection based on the defect detection prompt point, thereby avoiding the need of manually providing defect prompt, realizing automatic defect detection on the object to be detected and improving the detection efficiency. Meanwhile, the target detection method provided by the disclosure does not need to label a large amount of defect data, so that the detection cost is saved.
Fig. 2 is a flowchart of another object detection method according to the embodiment shown in fig. 1, and as shown in fig. 2, step S102 includes the following sub-steps:
In step S1021, feature extraction is performed on the target image, so as to obtain a first feature map of the object to be detected.
In this step, feature extraction of the target image can be achieved by:
inputting the target image into a feature extraction model obtained by training in advance, and then obtaining image features respectively output by at least one preset middle layer of the feature extraction model; splicing the image features respectively output by at least one preset intermediate layer according to the feature channels to obtain splicing features; and randomly sampling the spliced features to obtain the first feature map.
The feature extraction model may be, for example, a resnet50 network model pre-trained on an ImageNet dataset, and model parameters of the resnet50 network model may be initialized with parameters pre-trained on ImageNet. In addition, the model structure of the network of the reset 50 may include an input layer, an output layer, and a plurality of preset intermediate layers (for example, layer0, layer1, layer2 may be respectively recorded) connected between the input layer and the output layer, and different preset intermediate layers may be sequentially connected, so that the model structure of the network of the reset 50 may be represented as, for example, input→layer0→layer1→layer2→layer→layer n→output, the different preset intermediate layers are used to extract features of different dimensions of the input image, and after the input image is input into the network model of the reset 50 through the input layer, the feature image is subjected to layer following sampling through each preset intermediate layer, the feature image size is gradually reduced, and the feature channel number is gradually increased.
Taking defect detection of a die casting as an example, an X-ray image of the die casting acquired by an X-ray image acquisition device may be acquired and subjected to graying processing to obtain a gray-scale image. The image size of the gray scale image may then be adjusted based on the input parameters of the feature extraction model (including the size data of the input image), e.g., the image size may be adjusted to 224 x 224 images. For each image with the size adjusted to be extracted feature, the image may be input to a network of the resnet50, then the image features output by the preset interlayer layer1, the preset interlayer layer2 and the preset interlayer layer3 in the network of the resnet50 may be obtained, for example, the image feature dimension output by the preset interlayer layer1 is 256×56×56, the image feature dimension output by the preset interlayer layer2 is 512×28×28, the image feature dimension output by the preset interlayer layer3 is 1024×14×14, where the first dimension of each image feature dimension represents the number of feature channels, the second dimension and the third dimension represent the image dimensions, then the image feature output by the preset interlayer layer2 may be sampled twice to be 512×56×56, and the image feature output by the preset interlayer layer3 may be sampled four times to be 1024×56, so that after the image feature output by the three preset interlayer layers is spliced according to the feature channels, the size of the spliced feature is 1972×56, then the first dimension is not illustrated as the feature of the random feature, and the first dimension is 200, for example, and the example may be obtained.
In this way, the feature extraction model in the deep learning is adopted to perform feature extraction on the target image, after a first feature image of an object to be detected is obtained, the outlier detection is performed subsequently based on the first feature image, so that the computing resource can be saved, and meanwhile, the outlier detection efficiency is improved.
In step S1022, outlier detection is performed on the first feature map, so as to obtain the defect detection hint point.
Fig. 3 is a flowchart of a target detection method according to the embodiment shown in fig. 2, and as shown in fig. 3, step S1022 includes the following sub-steps:
in step S10221, a feature vector corresponding to each first pixel point on the first feature map is obtained.
In a possible implementation manner, for each first pixel point on the first feature map, the feature vector corresponding to each first pixel point may be acquired along the channel direction. For example, assuming that the size of the first feature map is 200×56×56 as an example, 56×56 first pixel points may be obtained along the channel direction, and the feature vector of 200 dimensions corresponding to each first pixel point is only illustrated herein, which is not limited in this disclosure.
In addition, when the outlier detection is performed, batch processing may be performed on the plurality of first feature maps, for example, in this step, the first feature maps corresponding to the N images may be superimposed to obtain feature vectors with a size of n×200×56×56, and then n×200 dimensions corresponding to each first pixel point may be obtained along the channel direction.
After the feature vector corresponding to each first pixel point in the first feature map is obtained, the feature vector can be normalized to be changed into a unit vector so as to perform subsequent outlier detection based on the unit vector.
In step S10222, according to the feature vectors corresponding to each first pixel point, an outlier score corresponding to each first pixel point is determined through a preset outlier detection model, where the outlier score characterizes whether the corresponding first pixel point belongs to an outlier pixel point.
The preset outlier detection model may be a detection model corresponding to any outlier detection algorithm, for example, the preset outlier detection model may be an outlier detection model corresponding to an LOF (Local Outlier Factor ) algorithm.
In one possible implementation manner of this step, for each first pixel, after the distance between the feature vector of the first pixel and the feature vector of each other first pixel is measured by using the LOF model, the outlier score corresponding to the first pixel is obtained, where in general, the higher the outlier score, the greater the probability that the first pixel belongs to the outlier pixel is represented.
For example, for N first feature maps of 200×56×56, after determining the outlier score corresponding to each first pixel point by a preset outlier detection model, the dimension of the obtained outlier score matrix is n×56×56.
In step S10223, the defect detection hint point is determined according to the outlier score corresponding to each first pixel point.
In this step, the first feature map may be up-sampled to obtain a second feature map having the same size as the target image; for each first pixel point, determining the discrete fraction of a second pixel point on the second feature map according to the discrete fraction of the first pixel point, wherein the second pixel point is a pixel point corresponding to the first pixel point; and determining the defect detection prompt point according to the outlier score of each second pixel point on the second feature map.
In consideration of the defect detection result to be determined in the present disclosure, a defective image area is marked on a target image, so that the first feature map needs to be up-sampled to obtain a second feature map with the same size as the target image, and second pixels on the second feature map are in one-to-one correspondence with pixels on the target image, so that after determining outlier pixels on the second feature map, defective pixels on the target image can be determined.
Taking an example of outlier detection on a first feature map, assuming that the dimension of an outlier matrix obtained after outlier detection is 56×56, and the size of an original target image is 224×224, four times of upsampling is needed to obtain a second feature map, the size of the second feature map is 224×224, it can be understood that four times of upsampling is needed to obtain the second feature map, and after each first pixel point on the first feature map, there are four second pixel points corresponding to each first pixel point on the second feature map, so that in the case that the discrete score of each first pixel point is a, the discrete score of each second pixel point corresponding to each second pixel point on the second feature map is a, so that the discrete score of each second pixel point on the second feature map is determined according to the discrete score of each first pixel point, which is merely illustrative.
In addition, in the present disclosure, the defect detection hint points include a defect detection positive sample hint point and a defect detection negative sample hint point; the defect detection positive sample prompting point is a pixel point of a defect of the object to be detected on the target image, and the defect detection negative sample prompting point is a pixel point of a defect which does not belong to the object to be detected on the target image.
Thus, in the process of determining the defect detection hint point according to the outlier score of each second pixel point on the second feature map, one possible implementation manner of the present disclosure may be: acquiring a first preset defect detection proportion; and after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the first preset defect detection proportion in the ranking result from big to small as the defect detection positive sample prompting points, and selecting the second pixel points with the first preset defect detection proportion in the ranking result from small to big as the defect detection negative sample prompting points.
Taking the defect detection of the die casting as an example, in the scene of defect detection of the die casting, the difference between the pixels of the defects (such as cracks and cold stops) on the target image and the pixels of the non-defect area on the die casting on the target image is large, and the proportion of the pixels corresponding to the defects on the die casting on the whole target image is relatively small, so that the pixels corresponding to the defects on the die casting can be regarded as outlier pixels, and therefore, which pixels belong to the defective pixels of the die casting and which pixels do not belong to the defective pixels of the die casting in the second feature map can be determined according to the outlier, and the pixels not belonging to the defective pixels of the die casting are regarded as positive sample prompting points for defect detection of the die casting.
For example, assuming that the first preset defect detection ratio is 5%, as described above, the higher the outlier score is, the greater the probability that the corresponding pixel belongs to the outlier pixel is represented, so, after the second pixels are sorted according to the outlier score, the second pixel of the first 5% (i.e., the second pixel of the first 5% of the outlier score) in the sorting result is selected as the positive sample prompting point for defect detection according to the order from large to small, and the second pixel of the second 5% (i.e., the second pixel of the second 5% of the outlier score of the second) is selected as the negative sample prompting point for defect detection according to the order from large to small, which is not limited by the disclosure.
Fig. 4 is a flowchart of a method for target detection according to the embodiment shown in fig. 3, in a possible implementation of the present disclosure, the object to be detected may be subjected to defect detection by performing a defect detection step in step S103, where the defect detection step includes the following steps, as shown in fig. 4:
in step S1031, a preset number of target positive sample prompting points are obtained from the defect detection positive sample prompting points;
in step S1032, a preset number of target negative sample prompting points are obtained from the defect detection negative sample prompting points;
In step S1033, after a preset number of the target positive sample prompting points and a preset number of the target negative sample prompting points are input into the target detection model, determining a target image area corresponding to the defect on the object to be detected according to an image segmentation area output by the model.
The target detection model may be, for example, a SAM model, and the target image region is a marked defect region.
For each target image, 3 points (i.e., a preset number) may be randomly sampled from the defect detection positive sample cue points corresponding to the target image as positive sample cue points, 3 points (i.e., a preset number) may be randomly sampled from the defect detection negative sample cue points corresponding to the target image as negative sample cue points, and then the positive sample cue points and the negative sample cue points may be input to a SAM model, through which an image segmentation region M may be output, such that the image segmentation region M may be used as the target image region, which is merely illustrated herein, and the disclosure is not limited thereto.
In another possible implementation manner of the present disclosure, in order to improve accuracy of the defect detection result, the image segmentation areas output by the multiple models may be obtained by performing the defect detection step multiple times, and then an intersection of the multiple image segmentation areas is taken as the target image area.
That is, the present disclosure may further take an intersection of the image division regions each detected as the target image region after performing the defect detection step a first preset number of times.
For example, after the defect detection step is repeatedly performed three times (i.e., the first preset number of times), three image division areas are acquired and respectively denoted as M1, M2, and M3, so that the intersection of M1, M2, and M3 may be taken as the target image area.
In still another possible implementation manner of the present disclosure, the image segmentation area obtained by each detection may be further obtained after performing the defect detection step of a second preset number of times, where the second preset number of times is greater than the first preset number of times; for each pixel point in the image segmentation area, determining a target pixel value corresponding to the pixel point according to the pixel value of the pixel point in the image segmentation area obtained by each detection; and determining the target image area according to the target pixel value of each pixel point in the image segmentation area.
The target pixel value corresponding to the pixel point can be determined according to the pixel value of the pixel point in the image segmentation area obtained by each detection in the following manner:
Counting the times that the pixel value of the pixel point is a first appointed pixel value; when the number of times is greater than or equal to a preset number of times threshold, the first appointed pixel value is used as the target pixel value corresponding to the pixel point; and under the condition that the times are smaller than the preset times threshold, taking a second designated pixel value as the target pixel value corresponding to the pixel point, wherein the first designated pixel value represents that the corresponding pixel point is a defective pixel point, and the second designated pixel value represents that the corresponding pixel point is a non-defective pixel point.
The first specified pixel value may be a pixel value indicating that the pixel belongs to a defective pixel, and the second specified pixel value may be a pixel value indicating that the pixel belongs to a non-defective pixel, for example, the first specified pixel value may be 1, and the second specified pixel value may be 0.
For example, assuming that the first specified pixel value is 1 and the second specified pixel value is 0, after repeating the above defect detection step 100 times (i.e., the second preset number of times), the pixel value of each pixel in the image division area may be counted for the image division area obtained by each detection, if the pixel is 90 times for the defect detection of 100 times, 10 times for the pixel value of 0, and more than the preset number of times threshold 80, the target pixel value corresponding to the pixel may be determined as 1, and if the number of times for the pixel value of 1 is less than the preset number of times threshold 80, the target pixel value corresponding to the pixel may be determined as 0, so that the target pixel value of each pixel in the image division area may be determined, and further the image area of the pixel with the target pixel value of 1 may be regarded as the target image area.
In addition, considering that the area range of the result of defect detection based on the visual large model may be large, in still another possible implementation of the present disclosure, the image segmentation area corresponding to the defect output by the model may be reduced by using the image mask, so as to further improve the accuracy of defect detection.
Thus, in the present disclosure, the method further comprises:
obtaining a second preset defect detection proportion, wherein the second preset defect detection proportion is larger than the first preset defect detection proportion; for example, the second preset defect detection ratio may be 30%, then after the second pixel points are sorted according to the outlier score, the second pixel points with the second preset defect detection ratio in the sorting result are selected as mask pixel points according to the order from small to large, and an image area corresponding to the mask pixel points is used as an image mask, where the image area corresponding to the image mask does not have the defects; for example, the second pixel point with the smallest outlier score of 30% in the sorting result can be selected as the mask pixel point in the order from small to large, and the image area corresponding to the mask pixel point is used as the image mask.
In this way, in determining the target image area corresponding to the defect on the object to be detected from the image division area output by the model, the target image area may be determined from the image division area output by the model and the image mask, and in general, the difference between the image division area and the image mask may be determined as the target image area having the defect.
Fig. 5 is a block diagram of an object detection apparatus according to an exemplary embodiment, as shown in fig. 5, the apparatus including:
an acquisition module 501 configured to acquire a target image of an object to be detected;
the first detection module 502 is configured to perform outlier detection on the target image to obtain a defect detection prompt point, where the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used to indicate whether the pixel point is a defective pixel point;
and the second detection module 503 is configured to detect the defect of the object to be detected according to the defect detection prompt point through a target detection model obtained through pre-training.
Optionally, the first detection module 502 is configured to perform feature extraction on the target image to obtain a first feature map of the object to be detected; and detecting outliers of the first feature map to obtain the defect detection prompt points.
Optionally, the first detection module 502 is configured to obtain a feature vector corresponding to each first pixel point on the first feature map; determining an outlier score corresponding to each first pixel point respectively through a preset outlier detection model according to the feature vector corresponding to each first pixel point respectively, wherein the outlier score represents whether the corresponding first pixel point belongs to the outlier pixel point; and determining the defect detection prompt points according to the outlier scores corresponding to the first pixel points respectively.
Optionally, the first detection module 502 is configured to upsample the first feature map to obtain a second feature map having the same size as the target image; for each first pixel point, determining the discrete fraction of a second pixel point on the second feature map according to the discrete fraction of the first pixel point, wherein the second pixel point is a pixel point corresponding to the first pixel point; and determining the defect detection prompt point according to the outlier score of each second pixel point on the second feature map.
Optionally, the defect detection prompting points comprise a defect detection positive sample prompting point and a defect detection negative sample prompting point; the first detection module 502 is configured to obtain a first preset defect detection ratio; and after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the first preset defect detection proportion in the ranking result from big to small as the defect detection positive sample prompting points, and selecting the second pixel points with the first preset defect detection proportion in the ranking result from small to big as the defect detection negative sample prompting points.
Optionally, the first detection module 502 is configured to input the target image into a feature extraction model obtained by training in advance, and then obtain image features respectively output by at least one preset middle layer of the feature extraction model; splicing the image features respectively output by at least one preset intermediate layer according to the feature channels to obtain splicing features; and randomly sampling the spliced features to obtain the first feature map.
Optionally, the second detecting module 503 is configured to perform defect detection on the object to be detected by performing a defect detecting step;
the defect detection step includes:
acquiring target positive sample prompting points with preset number from the defect detection positive sample prompting points;
obtaining target negative sample prompting points with preset number from the defect detection negative sample prompting points;
after a preset number of target positive sample prompting points and a preset number of target negative sample prompting points are input into the target detection model, determining a target image area corresponding to the defect on the object to be detected according to an image segmentation area output by the model.
Optionally, the second detecting module 503 is configured to take, as the target image area, an intersection of the image segmentation areas detected each time after performing the defect detecting step for a first preset number of times.
Optionally, the second detecting module 503 is configured to obtain the image segmentation area obtained by each detection after performing the defect detecting step of a second preset number of times, where the second preset number of times is greater than the first preset number of times;
for each pixel point in the image segmentation area, determining a target pixel value corresponding to the pixel point according to the pixel value of the pixel point in the image segmentation area obtained by each detection;
and determining the target image area according to the target pixel value of each pixel point in the image segmentation area.
Optionally, the second detecting module 503 is configured to count the number of times that the pixel value of the pixel point is the first specified pixel value;
when the number of times is greater than or equal to a preset number of times threshold, the first appointed pixel value is used as the target pixel value corresponding to the pixel point;
and under the condition that the times are smaller than the preset times threshold, taking a second designated pixel value as the target pixel value corresponding to the pixel point, wherein the first designated pixel value represents that the corresponding pixel point is a defective pixel point, and the second designated pixel value represents that the corresponding pixel point is a non-defective pixel point.
Optionally, the second detecting module 503 is configured to obtain a second preset defect detection ratio, where the second preset defect detection ratio is greater than the first preset defect detection ratio;
after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the second preset defect detection proportion in the ranking result as mask pixel points according to the order from small to large, and taking an image area corresponding to the mask pixel points as an image mask, wherein the defects do not exist in the image area corresponding to the image mask; and determining the target image area according to the image segmentation area output by the model and the image mask.
Optionally, the object detection model comprises a visual large model SAM.
The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.
The present disclosure also provides a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the steps of the object detection method provided by the present disclosure.
Fig. 6 is a block diagram illustrating an apparatus 800 for target detection according to an example embodiment. For example, apparatus 800 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, exercise device, personal digital assistant, or the like.
Referring to fig. 6, apparatus 800 may include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input/output interface 812, a sensor component 814, and a communication component 816.
The processing component 802 generally controls overall operation of the apparatus 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the target detection method described above. Further, the processing component 802 can include one or more modules that facilitate interactions between the processing component 802 and other components. For example, the processing component 802 can include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.
The memory 804 is configured to store various types of data to support operations at the apparatus 800. Examples of such data include instructions for any application or method operating on the device 800, contact data, phonebook data, messages, pictures, videos, and the like. The memory 804 may be implemented by any type or combination of volatile or nonvolatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk.
The power supply component 806 provides power to the various components of the device 800. The power components 806 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the device 800.
The multimedia component 808 includes a screen between the device 800 and the user that provides an output interface. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensor may sense not only the boundary of a touch or slide action, but also the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and/or a rear camera. The front camera and/or the rear camera may receive external multimedia data when the apparatus 800 is in an operational mode, such as a photographing mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have focal length and optical zoom capabilities.
The audio component 810 is configured to output and/or input audio signals. For example, the audio component 810 includes a Microphone (MIC) configured to receive external audio signals when the device 800 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, audio component 810 further includes a speaker for outputting audio signals.
Input/output interface 812 provides an interface between processing component 802 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: homepage button, volume button, start button, and lock button.
The sensor assembly 814 includes one or more sensors for providing status assessment of various aspects of the apparatus 800. For example, the sensor assembly 814 may detect an on/off state of the device 800, a relative positioning of the components, such as a display and keypad of the device 800, the sensor assembly 814 may also detect a change in position of the device 800 or a component of the device 800, the presence or absence of user contact with the device 800, an orientation or acceleration/deceleration of the device 800, and a change in temperature of the device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscopic sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
The communication component 816 is configured to facilitate communication between the apparatus 800 and other devices, either in a wired or wireless manner. The device 800 may access a wireless network based on a communication standard, such as WiFi,2G or 3G, or a combination thereof. In one exemplary embodiment, the communication component 816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, ultra Wideband (UWB) technology, bluetooth (BT) technology, and other technologies.
In an exemplary embodiment, the apparatus 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), digital Signal Processors (DSPs), digital Signal Processing Devices (DSPDs), programmable Logic Devices (PLDs), field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic elements for performing the above-described target detection methods.
In an exemplary embodiment, a non-transitory computer readable storage medium is also provided, such as memory 804 including instructions executable by processor 820 of apparatus 800 to perform the above-described target detection method. For example, the non-transitory computer readable storage medium may be ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.
In another exemplary embodiment, a computer program product is also provided, comprising a computer program executable by a programmable apparatus, the computer program having code portions for performing the above-described object detection method when executed by the programmable apparatus.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure. This disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (15)
1. A method of detecting an object, comprising:
acquiring a target image of an object to be detected;
Performing outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defect pixel point or not;
and performing defect detection on the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
2. The method of claim 1, wherein performing outlier detection on the target image to obtain a defect detection hint point comprises:
extracting features of the target image to obtain a first feature map of the object to be detected;
and detecting outliers of the first feature map to obtain the defect detection prompt points.
3. The method of claim 2, wherein performing outlier detection on the first feature map to obtain the defect detection hint point comprises:
acquiring a feature vector corresponding to each first pixel point on the first feature map;
determining an outlier score corresponding to each first pixel point respectively through a preset outlier detection model according to the feature vector corresponding to each first pixel point respectively, wherein the outlier score represents whether the corresponding first pixel point belongs to the outlier pixel point;
And determining the defect detection prompt points according to the outlier scores corresponding to the first pixel points respectively.
4. A method according to claim 3, wherein determining the defect detection hint point according to the outlier score corresponding to each first pixel point comprises:
upsampling the first feature map to obtain a second feature map having the same size as the target image;
for each first pixel point, determining the discrete fraction of a second pixel point on the second feature map according to the discrete fraction of the first pixel point, wherein the second pixel point is a pixel point corresponding to the first pixel point;
and determining the defect detection prompt point according to the outlier score of each second pixel point on the second feature map.
5. The method of claim 4, wherein the defect detection hint points include a defect detection positive sample hint point and a defect detection negative sample hint point; the determining the defect detection prompting point according to the outlier score of each second pixel point on the second feature map comprises:
acquiring a first preset defect detection proportion;
and after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the first preset defect detection proportion in the ranking result from big to small as the defect detection positive sample prompting points, and selecting the second pixel points with the first preset defect detection proportion in the ranking result from small to big as the defect detection negative sample prompting points.
6. The method according to claim 2, wherein the performing feature extraction on the target image to obtain the first feature map of the object to be detected includes:
inputting the target image into a feature extraction model obtained by training in advance, and then obtaining image features respectively output by at least one preset middle layer of the feature extraction model;
splicing the image features respectively output by at least one preset intermediate layer according to the feature channels to obtain splicing features;
and randomly sampling the spliced features to obtain the first feature map.
7. The method according to claim 5, wherein performing defect detection on the object to be detected by using a target detection model obtained by training in advance according to the defect detection prompt point comprises:
performing defect detection on the object to be detected by executing a defect detection step;
the defect detection step includes:
acquiring target positive sample prompting points with preset number from the defect detection positive sample prompting points;
obtaining target negative sample prompting points with preset number from the defect detection negative sample prompting points;
after a preset number of target positive sample prompting points and a preset number of target negative sample prompting points are input into the target detection model, determining a target image area corresponding to the defect on the object to be detected according to an image segmentation area output by the model.
8. The method of claim 7, wherein determining a target image region corresponding to a defect on the object to be detected from the image segmentation region output by the model comprises:
after the defect detection step is performed a first preset number of times, the intersection of the image division areas obtained by each detection is taken as the target image area.
9. The method of claim 7, wherein determining a target image region corresponding to a defect on the object to be detected from the image segmentation region output by the model comprises:
after the defect detection step of the second preset times is executed, the image segmentation area obtained by each detection is obtained, and the second preset times are larger than the first preset times;
for each pixel point in the image segmentation area, determining a target pixel value corresponding to the pixel point according to the pixel value of the pixel point in the image segmentation area obtained by each detection;
and determining the target image area according to the target pixel value of each pixel point in the image segmentation area.
10. The method according to claim 9, wherein determining the target pixel value corresponding to the pixel point according to the pixel value of the pixel point in the image segmentation area detected each time includes:
Counting the times that the pixel value of the pixel point is a first appointed pixel value;
when the number of times is greater than or equal to a preset number of times threshold, the first appointed pixel value is used as the target pixel value corresponding to the pixel point;
and under the condition that the times are smaller than the preset times threshold, taking a second designated pixel value as the target pixel value corresponding to the pixel point, wherein the first designated pixel value represents that the corresponding pixel point is a defective pixel point, and the second designated pixel value represents that the corresponding pixel point is a non-defective pixel point.
11. The method of claim 7, wherein determining a target image region corresponding to a defect on the object to be detected from the image segmentation region output by the model comprises:
obtaining a second preset defect detection proportion, wherein the second preset defect detection proportion is larger than the first preset defect detection proportion;
after the second pixel points are ranked according to the outlier score, selecting the second pixel points with the second preset defect detection proportion in the ranking result as mask pixel points according to the order from small to large, and taking an image area corresponding to the mask pixel points as an image mask, wherein the defects do not exist in the image area corresponding to the image mask;
And determining the target image area according to the image segmentation area output by the model and the image mask.
12. The method of any one of claims 1-11, wherein the object detection model comprises a visual large model SAM.
13. An object detection apparatus, comprising:
the acquisition module is configured to acquire a target image of an object to be detected;
the first detection module is configured to perform outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defective pixel point or not;
and the second detection module is configured to detect the defects of the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
14. An object detection apparatus, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
acquiring a target image of an object to be detected;
performing outlier detection on the target image to obtain a defect detection prompt point, wherein the defect detection prompt point is a pixel point with defect detection information on the target image, and the defect detection information is used for indicating whether the pixel point is a defect pixel point or not;
And performing defect detection on the object to be detected through a target detection model obtained through pre-training according to the defect detection prompt points.
15. A computer readable storage medium having stored thereon computer program instructions, which when executed by a processor, implement the steps of the method of any of claims 1 to 12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310715788.3A CN116703874B (en) | 2023-06-15 | 2023-06-15 | Target detection method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310715788.3A CN116703874B (en) | 2023-06-15 | 2023-06-15 | Target detection method, device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116703874A true CN116703874A (en) | 2023-09-05 |
| CN116703874B CN116703874B (en) | 2024-07-30 |
Family
ID=87833683
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310715788.3A Active CN116703874B (en) | 2023-06-15 | 2023-06-15 | Target detection method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116703874B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111640089A (en) * | 2020-05-09 | 2020-09-08 | 武汉精立电子技术有限公司 | Defect detection method and device based on feature map center point |
| CN111814867A (en) * | 2020-07-03 | 2020-10-23 | 浙江大华技术股份有限公司 | Defect detection model training method, defect detection method and related device |
| CN112508939A (en) * | 2020-12-22 | 2021-03-16 | 郑州金惠计算机系统工程有限公司 | Flange surface defect detection method, system and equipment |
| CN114494260A (en) * | 2022-04-18 | 2022-05-13 | 深圳思谋信息科技有限公司 | Object defect detection method and device, computer equipment and storage medium |
| WO2022127919A1 (en) * | 2020-12-17 | 2022-06-23 | 杭州海康威视数字技术股份有限公司 | Surface defect detection method, apparatus, system, storage medium, and program product |
| CN114663391A (en) * | 2022-03-22 | 2022-06-24 | 聚时科技(上海)有限公司 | Industrial image defect detection method based on unsupervised deep learning algorithm |
| CN115965563A (en) * | 2021-10-08 | 2023-04-14 | Tcl科技集团股份有限公司 | Defect detection method, device, equipment and computer readable storage medium |
| CN115994900A (en) * | 2023-01-18 | 2023-04-21 | 深圳市华汉伟业科技有限公司 | Unsupervised defect detection method and system based on transfer learning and storage medium |
-
2023
- 2023-06-15 CN CN202310715788.3A patent/CN116703874B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111640089A (en) * | 2020-05-09 | 2020-09-08 | 武汉精立电子技术有限公司 | Defect detection method and device based on feature map center point |
| CN111814867A (en) * | 2020-07-03 | 2020-10-23 | 浙江大华技术股份有限公司 | Defect detection model training method, defect detection method and related device |
| WO2022127919A1 (en) * | 2020-12-17 | 2022-06-23 | 杭州海康威视数字技术股份有限公司 | Surface defect detection method, apparatus, system, storage medium, and program product |
| CN112508939A (en) * | 2020-12-22 | 2021-03-16 | 郑州金惠计算机系统工程有限公司 | Flange surface defect detection method, system and equipment |
| CN115965563A (en) * | 2021-10-08 | 2023-04-14 | Tcl科技集团股份有限公司 | Defect detection method, device, equipment and computer readable storage medium |
| CN114663391A (en) * | 2022-03-22 | 2022-06-24 | 聚时科技(上海)有限公司 | Industrial image defect detection method based on unsupervised deep learning algorithm |
| CN114494260A (en) * | 2022-04-18 | 2022-05-13 | 深圳思谋信息科技有限公司 | Object defect detection method and device, computer equipment and storage medium |
| CN115994900A (en) * | 2023-01-18 | 2023-04-21 | 深圳市华汉伟业科技有限公司 | Unsupervised defect detection method and system based on transfer learning and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116703874B (en) | 2024-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111310616B (en) | Image processing method and device, electronic equipment and storage medium | |
| CN106651955B (en) | Method and device for positioning target object in picture | |
| CN108629354B (en) | Target detection method and device | |
| WO2023087741A1 (en) | Defect detection method and apparatus, and electronic device, storage medium and computer program product | |
| CN110674719B (en) | Target object matching method and device, electronic equipment and storage medium | |
| RU2577188C1 (en) | Method, apparatus and device for image segmentation | |
| CN109859096A (en) | Image Style Transfer method, apparatus, electronic equipment and storage medium | |
| CN106557759B (en) | Signpost information acquisition method and device | |
| CN110619350B (en) | Image detection method, device and storage medium | |
| CN106228556B (en) | image quality analysis method and device | |
| CN108062547B (en) | Character detection method and device | |
| CN107133354B (en) | Method and device for acquiring image description information | |
| CN108717542B (en) | Method and device for recognizing character area and computer readable storage medium | |
| CN112508974B (en) | Training method and device for image segmentation model, electronic equipment and storage medium | |
| CN114120034A (en) | Image classification method and device, electronic equipment and storage medium | |
| CN104077597A (en) | Image classifying method and device | |
| CN114332149A (en) | Image segmentation method and device, electronic equipment and storage medium | |
| CN111062401A (en) | Stacked object identification method and device, electronic device and storage medium | |
| CN112200040A (en) | Occlusion image detection method, device and medium | |
| CN110619325A (en) | Text recognition method and device | |
| CN106372663B (en) | Construct the method and device of disaggregated model | |
| CN113920293A (en) | Information identification method and device, electronic equipment and storage medium | |
| CN111428806B (en) | Image tag determining method and device, electronic equipment and storage medium | |
| CN112784701A (en) | Video semantic segmentation method and device and storage medium | |
| CN116310633A (en) | Key point detection model training method and key point detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |