CN115424129B - Abnormality detection method and abnormality detection system for wallboard damage - Google Patents
Abnormality detection method and abnormality detection system for wallboard damage Download PDFInfo
- Publication number
- CN115424129B CN115424129B CN202211253984.5A CN202211253984A CN115424129B CN 115424129 B CN115424129 B CN 115424129B CN 202211253984 A CN202211253984 A CN 202211253984A CN 115424129 B CN115424129 B CN 115424129B
- Authority
- CN
- China
- Prior art keywords
- wallboard
- image
- data set
- images
- damage
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The application discloses an abnormality detection method and an abnormality detection system for wallboard damage, and relates to an abnormality detection method and an abnormality detection system for wallboard damage. The application aims to solve the problems that the prior method for amplifying the original truck image data set can cause the change of the fault form and dimension in the original truck image in the training process and the model training precision is poor. The process is as follows: step one, collecting images; step two, establishing a data set; step three, obtaining a neural network model for detecting wallboard damage; judging the wallboard of the truck to be tested through the trained wallboard damage network model, if no fault exists, continuing to detect the next image, and if the fault exists, executing the fifth step; and step five, outputting the position of the fault. An abnormality detection system for wall panel breakage is used for executing an abnormality detection method for wall panel breakage. The application is used in the field of abnormality detection of wallboard damage.
Description
Technical Field
The application relates to an abnormality detection method and a detection system for wallboard damage.
Background
The freight car wallboard has the effect of protecting goods, can provide certain protection to the goods in the packing box. Once the damage phenomenon exists, the goods in the truck can be damaged and lost, so that the safety of the goods is endangered. In wallboard damage fault detection, the existing method often adopts a manual image inspection mode to carry out fault detection. In the detection process, the detection result is influenced by subjective factors of a vehicle detection person, so that the problems of missed detection, false detection and the like of faults are easy to occur, and the driving safety is influenced; the automatic detection replaces manual vehicle safety detection, so that the vehicle safety detection efficiency can be improved, the interference of human factors is eliminated, the labor cost is reduced, but the data set amplification method in the automatic detection process has the problems of changing the fault form and dimension in the original truck picture in the training process and poor model training precision.
Disclosure of Invention
The application aims to solve the problems that the prior method for amplifying an original truck image data set can cause the change of fault form and dimension in an original truck image in the training process and the model training precision is poor, and provides an abnormality detection method and an abnormality detection system for wallboard damage.
The abnormality detection method for wallboard damage comprises the following specific processes:
step one, collecting images;
step two, establishing a data set:
dividing a stored image according to the wheelbase, the vehicle type and the priori knowledge of the train bottom, intercepting a wallboard position image, obtaining an original data set, filling the original image data set, and carrying out data amplification on the filled data set to obtain a sample data set of a wallboard damage network model;
step three, obtaining a neural network model for detecting wallboard damage;
judging the wallboard of the truck to be tested through the trained wallboard damage network model, if no fault exists, continuing to detect the next image, and if the fault exists, executing the fifth step;
and step five, outputting the position of the fault.
Preferably, the image is acquired in the first step; the specific process is as follows:
linear array imaging devices are arranged on the two sides and the bottom of a rail, and a truck starts to start the imaging devices to scan the moving truck through triggering sensors; and acquiring line images after progressive scanning, and storing the line images.
Preferably, in step two, a data set is established:
dividing a stored image according to the wheelbase, the vehicle type and the priori knowledge of the train bottom, intercepting a wallboard position image, obtaining an original data set, filling the original image data set, and carrying out data amplification on the filled data set to obtain a sample data set of a wallboard damage network model;
the specific process is as follows:
firstly, counting the sizes of all the intercepted wallboard images, and designing the unified size of the image according to the maximum length size and the maximum width size of the obtained intercepted wallboard images;
step two, after the uniform size of the image is designed, filling the image of all the intercepted wallboard images in a pattern to obtain the filled intercepted wallboard images;
and step two, carrying out data amplification on the filled intercepted wallboard image to obtain an amplified data set, marking the amplified data set, and obtaining a marking file corresponding to the amplified data set one by one, wherein the amplified data set and the marking file corresponding to the amplified data set one by one are used as sample data sets of the wallboard damage network model.
Preferably, in the second step, firstly, the sizes of all the intercepted wallboard images are counted, and the unified sizes of the images are designed according to the maximum length size and the maximum width size of the obtained intercepted wallboard images; the design process is as follows:
if the maximum length dimension of the obtained captured wallboard image is A, then taking 2 not greater than A n Obtaining a value of n; taking the mixture not smaller than A-2 n 2 of (2) m Obtaining a value of m;
n and m are integers;
get 2 n +2 m As a length value in the unified size of the image;
if the maximum width dimension of the obtained captured wallboard image is B, then 2 is taken to be not more than B p Obtaining a value of p; taking out not less than B-2 p 2 of (2) q Obtaining a value of q;
p and q are integers;
get 2 p +2 q As a width value in the unified size of the image.
Preferably, after the uniform size of the image is designed in the second step, filling the image of all the intercepted wallboard images with the image to obtain the filled intercepted wallboard images; the process is as follows:
and (5) filling black to the right and downwards to amplify the intercepted wallboard image to the uniform size of the image, so as to obtain the filled intercepted wallboard image.
Preferably, in the second and third steps, the filled intercepted wallboard image is subjected to data amplification to obtain an amplified data set, the amplified data set is marked, a marking file corresponding to the amplified data set one by one is obtained, and the amplified data set and the marking file corresponding to the amplified data set one by one are used as sample data sets of the wallboard damage network model.
Preferably, in step three, a neural network model for detecting wall panel breakage is obtained; the specific process is as follows:
the wallboard damage network model comprises a ResNet50 feature extraction network, an RPN network and a RoI pooling layer;
selecting a ResNet50 characteristic extraction network as a backbone network of a wallboard damage network model, and taking a sample data set subjected to data amplification as an input of the ResNet50 characteristic extraction network to obtain an output characteristic diagram; taking an output characteristic diagram of a ResNet50 backbone network as an input of an RPN network to generate a candidate frame; the RoI mapping layer obtains a candidate frame characteristic diagram with a fixed size by utilizing a candidate frame generated by an RPN network and an output characteristic diagram obtained by a ResNet50 backbone network;
training the wall board damage network model to obtain a trained wall board damage network model.
Preferably, in the fourth step, judging the wallboard of the truck to be tested through the trained wallboard damage network model, if no fault exists, continuing to detect the next image, and if the fault exists, executing the fifth step; the specific process is as follows:
cutting out wallboard pictures from the gray level images of the whole vehicle to be detected as images to be detected, filling the images to be detected, namely filling black into the images to be detected rightwards and downwards until the images to be detected are of uniform size designed in the second step, calling a trained wallboard damage network model to detect wallboard faults of the filled images to be detected, obtaining abnormal probability values of candidate frames, screening the abnormal probability values of the candidate frames according to preset thresholds, and if the probability values of the foreign matters in the candidate frames are larger than the thresholds, considering that the abnormal conditions exist in the candidate frames, and executing the fifth step; and if the probability value of the foreign matter in the candidate frame is smaller than or equal to the threshold value, considering that no abnormality exists in the candidate frame.
Preferably, outputting the position of the fault in the fifth step; the specific process is as follows:
after fault identification, calculating the position of the fault in the full-vehicle gray level image to be detected through the mapping relation from the filled image to be detected to the image to be detected and from the image to be detected to the full-vehicle gray level image to be detected;
after the position of the fault in the gray level image of the whole vehicle to be detected is calculated, uploading fault information to an alarm platform, and displaying the fault on a display interface.
An abnormality detection system for wall panel breakage is used for executing an abnormality detection method for wall panel breakage.
The application has the beneficial effects that:
the application uses a linear array camera to collect the chassis image of a truck, obtains the position image of a car coupler by a coarse positioning method, and establishes a sample data set by a data amplification and data marking method; training a truck wallboard damage fault detection model through a sample data set; judging the wallboard damage faults of the original image through a trained truck wallboard damage fault detection model; if the original image is judged to have wallboard damage faults, the position of the faults in the original image is calculated, and an alarm is given.
(1) By using the image recognition method, the manual workload of truck detection is reduced, and the detection accuracy can be greatly improved.
(2) The deep learning algorithm is applied to automatic recognition of wallboard damage faults, so that the stability and the accuracy of the whole algorithm are improved.
(3) The fast R-CNN network frame is optimized, and the fine fault form is optimized by using an OHEM method, so that better detection precision is obtained.
(4) In the original image processing and image detection, the size of the fault structure is not changed through filling processing, so that the detection precision is improved.
Drawings
FIG. 1 is a flow chart of the present application;
FIG. 2 is a training test flow chart of the present application.
Detailed Description
It should be noted that, in particular, the various embodiments of the present disclosure may be combined with each other without conflict.
The first embodiment is as follows: referring to fig. 1, a specific procedure of the abnormality detection method for wallboard damage according to the present embodiment is as follows:
step one, collecting images;
step two, establishing a data set:
dividing a stored image according to the wheelbase, the vehicle type and the priori knowledge of the train bottom, intercepting a wallboard position image, obtaining an original data set, filling the original image data set, and carrying out data amplification on the filled data set to obtain a sample data set of a wallboard damage network model;
step three, obtaining a neural network model for detecting wallboard damage;
judging the wallboard of the truck to be tested through the trained wallboard damage network model, if no fault exists, continuing to detect the next image, and if the fault exists, executing the fifth step;
and step five, outputting the position of the fault.
The second embodiment is as follows: the first difference between the present embodiment and the specific embodiment is that the image is acquired in the first step; the specific process is as follows:
high-definition linear array imaging devices are arranged on the two sides and the bottom of a rail, and a truck starts to start the imaging devices to scan the moving truck through triggering sensors; and acquiring high-definition line images after progressive scanning, and storing the line images.
Other steps and parameters are the same as in the first embodiment.
And a third specific embodiment: the difference between this embodiment and the first or second embodiment is that in the second step, a data set is created:
dividing a stored image according to the wheelbase, the vehicle type and the priori knowledge of the train bottom, intercepting a wallboard position image, obtaining an original data set, filling the original image data set, and carrying out data amplification on the filled data set to obtain a sample data set of a wallboard damage network model;
the specific process is as follows:
firstly, counting the sizes of all the intercepted wallboard images, and designing the unified size of the image according to the maximum length size and the maximum width size of the obtained intercepted wallboard images;
step two, after the uniform size of the image is designed, filling the image of all the intercepted wallboard images in a pattern to obtain the filled intercepted wallboard images;
and step two, carrying out data amplification on the filled intercepted wallboard image to obtain an amplified data set, marking the amplified data set, and obtaining a marking file corresponding to the amplified data set one by one, wherein the amplified data set and the marking file corresponding to the amplified data set one by one are used as sample data sets of the wallboard damage network model.
Other steps and parameters are the same as in the first or second embodiment.
The specific embodiment IV is as follows: the difference between the embodiment and the first to third embodiments is that in the second step, firstly, the sizes of all the intercepted wallboard images are counted, and the unified sizes of the images are designed according to the maximum length size and the maximum width size of the obtained intercepted wallboard images; the design process is as follows:
if the maximum length dimension of the obtained captured wallboard image is A, then taking 2 not greater than A n Obtaining a value of n; taking the mixture not smaller than A-2 n 2 of (2) m Obtain the minimum value of (2)A value of m;
n and m are integers;
get 2 n +2 m As a length value in the unified size of the image;
if the maximum width dimension of the obtained captured wallboard image is B, then 2 is taken to be not more than B p Obtaining a value of p; taking out not less than B-2 p 2 of (2) q Obtaining a value of q;
p and q are integers;
get 2 p +2 q As a width value in the uniform size of the image;
for example, the maximum length and width dimensions are both 1000, not greater than the maximum length dimension to the power of 2 to the power of 9, 512, then n is 9, then 1000-512=488, and the minimum dimension not less than 488 is 512, i.e., m is 9, so the final fill dimension is 1024, m=9, n=9, and the final fill dimension is as wide as the final fill dimension.
The design of the uniform size has the advantages that: the method ensures that the proportion of fault structures in the picture is not changed due to filling when the wallboard image is intercepted, and the image is reduced from large resolution to small resolution by an exponential power of 2, so that the input is better to the exponential power of 2 or better to the exponential power close to 2.
Other steps and parameters are the same as in one to three embodiments.
Fifth embodiment: the difference between the embodiment and the first to fourth embodiments is that after the uniform size of the image is designed in the second step, all the intercepted wallboard images are subjected to graphic filling to obtain the filled intercepted wallboard images; the process is as follows:
filling black to right and downwards to amplify the intercepted wallboard image to the uniform size of the image, so as to obtain a filled intercepted wallboard image;
the original image dataset is filled, so that the fault form and size in the original image can be guaranteed not to be changed in the training process, and the model training precision is better;
other steps and parameters are the same as in one to four embodiments.
Specific embodiment six: the difference between the embodiment and the specific embodiments is that, in the second to fifth steps, the filled intercepted wallboard image is subjected to data amplification to obtain an amplified data set, the amplified data set is marked, a mark file corresponding to the amplified data set one by one is obtained, and the amplified data set and the mark file corresponding to the amplified data set one by one are used as a sample data set of the wallboard damage network model.
Other steps and parameters are the same as in one of the first to fifth embodiments.
Seventh embodiment: the difference between the present embodiment and one to six embodiments is that the neural network model for detecting the wall board breakage is obtained in the third step; the specific process is as follows:
the wallboard damage network model comprises a ResNet50 feature extraction network, an RPN network and a RoI pooling layer;
selecting a ResNet50 characteristic extraction network as a backbone network of a wallboard damage network model, and taking a sample data set subjected to data amplification as an input of the ResNet50 characteristic extraction network to obtain an output characteristic diagram; taking an output characteristic diagram of a ResNet50 backbone network as an input of an RPN network to generate a candidate frame; the RoI mapping layer obtains a candidate frame characteristic diagram with a fixed size by utilizing a candidate frame generated by an RPN network and an output characteristic diagram obtained by a ResNet50 backbone network so as to carry out subsequent target classification and positioning;
training the wall board damage network model by using an OHEM method to obtain a trained wall board damage network model.
OHEM is a difficult sample mining method, and its specific implementation is as follows:
1, firstly, carrying out forward propagation of the Faster RCNN once to obtain an independent loss value of each ROI;
2 performing non-maximum suppression NMS for each ROI;
3 ordering the ROIs passing through the non-maximum suppression NMS according to loss values, and then inputting the first 50% of the ROIs with larger loss values into a classification and regression network (rpn network);
4. repeatedly executing the steps 1, 2 and 3 until convergence to obtain a trained wallboard damage network model;
by the method, the characteristics difficult to learn can be better learned, and the accuracy of the detection model is improved.
Other steps and parameters are the same as in one of the first to sixth embodiments.
Eighth embodiment: the difference between the embodiment and the first to seventh embodiments is that in the fourth step, the wall board of the truck to be tested is judged through the trained wall board damage network model, if no fault exists, the next image is continuously detected, and if the fault exists, the fifth step is executed; the specific process is as follows:
cutting out wallboard pictures from the full-vehicle high ash removal degree image to be detected as an image to be detected, filling the image to be detected, namely filling black into the right and downward of the image to be detected to ensure that the fault size and structure in the image input into the network to be detected are not changed, calling a trained wallboard damage network model to perform wallboard fault detection on the filled image to be detected to obtain a probability value of abnormality in a candidate frame, screening the probability value of abnormality in the candidate frame according to a preset threshold value, and if the probability value of the foreign matter in the candidate frame is larger than the threshold value, considering that the abnormality exists in the candidate frame, and executing a step five; and if the probability value of the foreign matter in the candidate frame is smaller than or equal to the threshold value, considering that no abnormality exists in the candidate frame.
Other steps and parameters are the same as those of one of the first to seventh embodiments.
Detailed description nine: the present embodiment is different from one of the first to eighth embodiments in that the position of the output fault in the fifth step; the specific process is as follows:
after fault identification, calculating the position of the fault in the full-vehicle high-definition gray level image to be detected through the mapping relation from the filled image to be detected to the image to be detected and from the image to be detected to the full-vehicle high-definition gray level image to be detected;
after the position of the fault in the full-vehicle height ash removal degree image to be detected is calculated, uploading fault information to an alarm platform, and displaying the fault on a display interface.
Other steps and parameters are the same as in one to eight of the embodiments.
Detailed description ten: the abnormality detection system for wall panel breakage of the present embodiment is used to perform the abnormality detection method for wall panel breakage of one of the first to ninth embodiments.
The above examples of the present application are only for describing the calculation model and calculation flow of the present application in detail, and are not limiting of the embodiments of the present application. Other variations and modifications of the above description will be apparent to those of ordinary skill in the art, and it is not intended to be exhaustive of all embodiments, all of which are within the scope of the application.
Claims (8)
1. An anomaly detection method for wallboard damage is characterized in that: the method comprises the following specific processes:
step one, collecting images;
step two, establishing a data set:
dividing a stored image according to the wheelbase, the vehicle type and the priori knowledge of the train bottom, intercepting a wallboard position image, obtaining an original data set, filling the original image data set, and carrying out data amplification on the filled data set to obtain a sample data set of a wallboard damage network model;
step three, obtaining a neural network model for detecting wallboard damage;
judging the wallboard of the truck to be tested through the trained wallboard damage network model, if no fault exists, continuing to detect the next image, and if the fault exists, executing the fifth step;
step five, outputting the position of the fault;
acquiring an image in the first step; the specific process is as follows:
linear array imaging devices are arranged on the two sides and the bottom of a rail, and a truck starts to start the imaging devices to scan the moving truck through triggering sensors; acquiring line images after progressive scanning, and storing the line images;
establishing a data set in the second step:
dividing a stored image according to the wheelbase, the vehicle type and the priori knowledge of the train bottom, intercepting a wallboard position image, obtaining an original data set, filling the original image data set, and carrying out data amplification on the filled data set to obtain a sample data set of a wallboard damage network model;
the specific process is as follows:
firstly, counting the sizes of all the intercepted wallboard images, and designing the unified size of the image according to the maximum length size and the maximum width size of the obtained intercepted wallboard images;
step two, after the uniform size of the image is designed, filling the image of all the intercepted wallboard images in a pattern to obtain the filled intercepted wallboard images;
and step two, carrying out data amplification on the filled intercepted wallboard image to obtain an amplified data set, marking the amplified data set, and obtaining a marking file corresponding to the amplified data set one by one, wherein the amplified data set and the marking file corresponding to the amplified data set one by one are used as sample data sets of the wallboard damage network model.
2. The abnormality detection method for wall plate breakage according to claim 1, characterized in that: firstly, counting the sizes of all the intercepted wallboard images, and designing the unified size of the image according to the maximum length size and the maximum width size of the obtained intercepted wallboard images; the design process is as follows:
if the maximum length dimension of the obtained captured wallboard image is A, then taking 2 not greater than A n Obtaining a value of n; taking the mixture not smaller than A-2 n 2 of (2) m Obtaining a value of m;
n and m are integers;
get 2 n +2 m As a length value in the unified size of the image;
if the maximum width dimension of the obtained captured wallboard image is B, then 2 is taken to be not more than B p Obtaining a value of p; taking out not less than B-2 p 2 of (2) q Obtaining a value of q;
the p and q are integers;
get 2 p +2 q As a width value in the unified size of the image.
3. The abnormality detection method for wall plate breakage according to claim 2, characterized in that: after the uniform size of the image is designed in the second step, filling the image of all the intercepted wallboard images with the image to obtain the filled intercepted wallboard images; the process is as follows:
and (5) filling black to the right and downwards to amplify the intercepted wallboard image to the uniform size of the image, so as to obtain the filled intercepted wallboard image.
4. The abnormality detection method for wall plate breakage according to claim 3, characterized in that: and in the second step, the filled intercepted wallboard image is subjected to data amplification to obtain an amplified data set, the amplified data set is marked to obtain a marking file corresponding to the amplified data set one by one, and the amplified data set and the marking file corresponding to the amplified data set one by one are used as sample data sets of the wallboard damage network model.
5. The abnormality detection method for wall panel breakage according to claim 4, characterized in that: the neural network model for detecting wallboard damage is obtained in the third step; the specific process is as follows:
the wallboard damage network model comprises a ResNet50 feature extraction network, an RPN network and a Roipooling layer;
selecting a ResNet50 characteristic extraction network as a backbone network of a wallboard damage network model, and taking a sample data set subjected to data amplification as an input of the ResNet50 characteristic extraction network to obtain an output characteristic diagram; taking the output characteristic diagram of the ResNet50 backbone network as the input of the RPN network to generate a candidate frame; the Roipooling layer obtains a candidate frame characteristic diagram with a fixed size by utilizing a candidate frame generated by an RPN network and an output characteristic diagram obtained by a ResNet50 backbone network;
training the wall board damage network model to obtain a trained wall board damage network model.
6. The abnormality detection method for wall board breakage according to claim 5, characterized in that: judging the wallboard of the truck to be tested through the trained wallboard damage network model in the fourth step, if no fault exists, continuing to detect the next image, and if the fault exists, executing the fifth step; the specific process is as follows:
cutting out wallboard pictures from the gray level images of the whole vehicle to be detected as images to be detected, filling the images to be detected, namely filling black into the images to be detected rightwards and downwards until the images to be detected are of uniform size designed in the second step, calling a trained wallboard damage network model to detect wallboard faults of the filled images to be detected, obtaining abnormal probability values of candidate frames, screening the abnormal probability values of the candidate frames according to preset thresholds, and if the probability values of the foreign matters in the candidate frames are larger than the thresholds, considering that the abnormal conditions exist in the candidate frames, and executing the fifth step; and if the probability value of the foreign matter in the candidate frame is smaller than or equal to the threshold value, considering that no abnormality exists in the candidate frame.
7. The abnormality detection method for wall panel breakage according to claim 6, characterized in that: outputting the position of the fault in the fifth step; the specific process is as follows:
after fault identification, calculating the position of the fault in the full-vehicle gray level image to be detected through the mapping relation from the filled image to be detected to the image to be detected and from the image to be detected to the full-vehicle gray level image to be detected;
after the position of the fault in the gray level image of the whole vehicle to be detected is calculated, uploading fault information to an alarm platform, and displaying the fault on a display interface.
8. An anomaly detection system for wallboard damage, characterized in that: the system is used for executing an abnormality detection method for wall panel breakage according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211253984.5A CN115424129B (en) | 2022-10-13 | 2022-10-13 | Abnormality detection method and abnormality detection system for wallboard damage |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211253984.5A CN115424129B (en) | 2022-10-13 | 2022-10-13 | Abnormality detection method and abnormality detection system for wallboard damage |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115424129A CN115424129A (en) | 2022-12-02 |
CN115424129B true CN115424129B (en) | 2023-08-11 |
Family
ID=84206882
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211253984.5A Active CN115424129B (en) | 2022-10-13 | 2022-10-13 | Abnormality detection method and abnormality detection system for wallboard damage |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115424129B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110363781A (en) * | 2019-06-29 | 2019-10-22 | 南京理工大学 | Molten bath profile testing method based on deep neural network |
WO2020164270A1 (en) * | 2019-02-15 | 2020-08-20 | 平安科技(深圳)有限公司 | Deep-learning-based pedestrian detection method, system and apparatus, and storage medium |
CN111652227A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting damage fault of bottom floor of railway wagon |
CN112669301A (en) * | 2020-12-31 | 2021-04-16 | 哈尔滨市科佳通用机电股份有限公司 | High-speed rail bottom plate paint removal fault detection method |
CN112990237A (en) * | 2019-12-02 | 2021-06-18 | 上海交通大学 | Subway tunnel image leakage detection method based on deep learning |
CN114266846A (en) * | 2021-12-25 | 2022-04-01 | 福州大学 | Self-learning filling method for target detection model |
-
2022
- 2022-10-13 CN CN202211253984.5A patent/CN115424129B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020164270A1 (en) * | 2019-02-15 | 2020-08-20 | 平安科技(深圳)有限公司 | Deep-learning-based pedestrian detection method, system and apparatus, and storage medium |
CN110363781A (en) * | 2019-06-29 | 2019-10-22 | 南京理工大学 | Molten bath profile testing method based on deep neural network |
CN112990237A (en) * | 2019-12-02 | 2021-06-18 | 上海交通大学 | Subway tunnel image leakage detection method based on deep learning |
CN111652227A (en) * | 2020-05-21 | 2020-09-11 | 哈尔滨市科佳通用机电股份有限公司 | Method for detecting damage fault of bottom floor of railway wagon |
CN112669301A (en) * | 2020-12-31 | 2021-04-16 | 哈尔滨市科佳通用机电股份有限公司 | High-speed rail bottom plate paint removal fault detection method |
CN114266846A (en) * | 2021-12-25 | 2022-04-01 | 福州大学 | Self-learning filling method for target detection model |
Also Published As
Publication number | Publication date |
---|---|
CN115424129A (en) | 2022-12-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR100730051B1 (en) | Defect detection apparatus and defect detection method | |
KR100241504B1 (en) | Display screen inspection method | |
KR20090101356A (en) | Defect detecting device, and defect detecting method | |
US20200265575A1 (en) | Flaw inspection apparatus and method | |
CN103927749A (en) | Image processing method and device and automatic optical detector | |
CN112233096B (en) | Vehicle apron board fault detection method | |
JP2001264257A (en) | Image defect detecting device and method, and storage medium storing procedure for image defect detecting method | |
JP4230880B2 (en) | Defect inspection method | |
KR100250631B1 (en) | Image processing method | |
CN116433938A (en) | Abnormal detection method, system and storage medium for bolt looseness of subway gearbox | |
JP2005172559A (en) | Method and device for detecting line defect on panel | |
CN112669301A (en) | High-speed rail bottom plate paint removal fault detection method | |
CN115424129B (en) | Abnormality detection method and abnormality detection system for wallboard damage | |
KR20140121068A (en) | Method and apparatus of inspecting mura of flat display | |
CN114792369B (en) | Cigarette carton filling state detection method and system based on light projection | |
JP5190619B2 (en) | Inspection method for inspection object and inspection apparatus therefor | |
CN114998194A (en) | Product defect detection method, system and storage medium | |
JPH08145907A (en) | Inspection equipment of defect | |
JP4956077B2 (en) | Defect inspection apparatus and defect inspection method | |
CN114495084A (en) | Information acquisition method and device, electronic equipment and storage medium | |
CN113449617A (en) | Track safety detection method, system, device and storage medium | |
JP2022125593A (en) | Abnormality detecting method and abnormality detecting device | |
CN116309574B (en) | Method, system, equipment and storage medium for detecting panel leakage process defects | |
JPH0735699A (en) | Method and apparatus for detecting surface defect | |
JP4741289B2 (en) | Image processing apparatus and image processing 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 |