CN111091150A - Railway wagon cross rod cover plate fracture detection method - Google Patents
Railway wagon cross rod cover plate fracture detection method Download PDFInfo
- Publication number
- CN111091150A CN111091150A CN201911293655.1A CN201911293655A CN111091150A CN 111091150 A CN111091150 A CN 111091150A CN 201911293655 A CN201911293655 A CN 201911293655A CN 111091150 A CN111091150 A CN 111091150A
- Authority
- CN
- China
- Prior art keywords
- cover plate
- image
- fracture
- plate fracture
- fault
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 11
- 230000000903 blocking effect Effects 0.000 claims abstract description 4
- 238000010586 diagram Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 8
- 230000003321 amplification Effects 0.000 claims description 5
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 5
- 238000011176 pooling Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 description 8
- 238000013135 deep learning Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003416 augmentation Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Multimedia (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Train Traffic Observation, Control, And Security (AREA)
Abstract
A rail wagon cross rod cover plate fracture detection method belongs to the field of rail wagon safe operation. The invention aims at the problems of low efficiency and poor reliability of the existing railway wagon cross rod cover plate fracture fault detection in a manual mode. The method comprises the following steps: establishing a cover plate fracture image sample library, marking unit frames in a fracture area or a fuzzy area in each cover plate fracture image sample in a blocking mode, and configuring a category label for each unit frame; building a fast-Rcnn algorithm frame model, and training by adopting a cover plate fracture image sample; obtaining a weight coefficient corresponding to a failed label and a suspected label; inputting the image to be recognized into the updated fast-Rcnn algorithm frame model, obtaining the fault prediction area of the image to be recognized and the confidence coefficient corresponding to the fault prediction area based on the weight coefficient, determining the fault prediction area with the confidence coefficient larger than the set threshold value as the fault area, and alarming. The invention can realize the automatic identification of the fracture fault of the cross rod cover plate.
Description
Technical Field
The invention relates to a method for detecting breakage of a cross rod cover plate of a railway wagon, and belongs to the field of safe operation of railway wagons.
Background
Conventionally, the fault of the cross rod cover plate of the railway wagon is checked manually, namely, a vehicle inspection person checks an acquired cross rod cover plate image to judge whether a fracture fault occurs. Due to high cost and low efficiency of manual detection and the fact that the manual mode is easy to generate operation fatigue or inattention and the like, the situations of missed detection, false detection and the like of images are easy to cause, and the running safety of the railway wagon is influenced to a great extent. In addition, the manual method is not favorable for the development of the automation technology.
Therefore, aiming at the defects of manual fault detection, the realization of fault detection automation of the fracture of the cross rod cover plate has great significance. With the rapid development of the fields of pattern recognition, deep learning and artificial intelligence, the technologies can be adopted to automatically detect faults of the collected cross rod cover plate images so as to improve the working efficiency and the fault detection accuracy and promote the rapid development of the automation degree of railway transportation.
Disclosure of Invention
The invention provides a railway wagon cross rod cover plate fracture detection method, which aims at the problems of low efficiency and poor reliability of the existing railway wagon cross rod cover plate fracture fault detection method by adopting a manual mode.
The invention discloses a method for detecting breakage of a cross rod cover plate of a railway wagon, which comprises the following steps of:
the method comprises the following steps: establishing a cover plate fracture image sample library, marking unit frames in a fracture area or a fuzzy area in each cover plate fracture image sample in a blocking mode, and configuring a category label for each unit frame; the category labels comprise a fault label and a suspected label;
step two: building a fast-Rcnn algorithm frame model, training by adopting cover plate fracture image samples, generating an initial detection frame on a characteristic diagram obtained by each cover plate fracture image sample in a sliding window mode, and correcting the initial detection frame based on a corresponding unit frame to obtain a target frame; gradually adjusting the model parameters until the model converges in the process of obtaining the target frame, and obtaining the weight coefficients corresponding to the fault label and the suspected label;
step three: acquiring an image of a cross rod cover plate of a running truck, and preprocessing the image to be recognized; inputting the image to be recognized into the updated fast-Rcnn algorithm frame model, obtaining the fault prediction area of the image to be recognized and the confidence coefficient corresponding to the fault prediction area based on the weight coefficient, determining the fault prediction area with the confidence coefficient larger than the set threshold value as the fault area, and alarming.
According to the method for detecting the breakage of the cross rod cover plate of the railway wagon, the cover plate breakage image sample is obtained by preprocessing the collected original cover plate breakage image.
According to the method for detecting the breakage of the cross rod cover plate of the railway wagon, the mode of preprocessing the original cover plate breakage image comprises amplification, wherein the amplification comprises image rotation, mirror image, zooming and contrast enhancement.
According to the detection method for the breakage of the cross bar cover plate of the railway wagon, the processing of the cover plate breakage image sample by the Faster-Rcnn algorithm framework model comprises the following steps:
and obtaining a characteristic diagram from the cover plate fracture image sample.
According to the rail wagon cross rod cover plate fracture detection method, the processing of the cover plate fracture image sample by the Faster-Rcnn algorithm framework model further comprises the following steps:
and generating initial detection frames by adopting two parallel regional recommendation network layers for the characteristic diagram in a sliding window mode, and performing pooling processing on all the initial detection frames to obtain a target frame.
According to the rail wagon cross rod cover plate fracture detection method, the sliding window comprises three sizes, and the length-width ratios are respectively { 1: 1,1: 2,2: 1}.
The invention has the beneficial effects that: the method adopts the trained fast-Rcnn algorithm frame model to identify the fault of the image to be identified, can realize the automatic identification of the fracture fault of the cross bar cover plate and can give an alarm.
The invention replaces manual detection with an automatic image identification mode, so that the detection result is not influenced by artificial subjective factors, the operation quality is greatly improved, and the reliability of the detection result is high. Therefore, the fault detection efficiency and accuracy are greatly improved.
The improved fast-Rcnn algorithm framework model is adopted, and the deep learning method is applied to detect whether the cross rod cover plate is broken, so that the stability is better than that of machine learning or a traditional artificial image processing method.
The method provided by the invention promotes the fault detection automation process in the transportation industry and can reduce the potential safety hazard of truck running to a great extent.
Drawings
FIG. 1 is a detailed flow chart of the method for detecting the breakage of a cross bar cover plate of a railway wagon according to the present invention;
FIG. 2 is an image of a sample library of cover plate fracture images;
FIG. 3 is a basic block diagram of a prior art fast-Rcnn algorithm framework;
FIG. 4 is a diagram of the fast-Rcnn algorithm framework model improved by the present invention;
FIG. 5 is a diagram of a prior art RPN architecture;
fig. 6 is a block diagram of an improved RPN of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
The invention is further described with reference to the following drawings and specific examples, which are not intended to be limiting.
In a first embodiment, as shown in fig. 1 to 4, the invention provides a method for detecting a breakage of a cross bar cover plate of a railway wagon, which includes the following steps:
the method comprises the following steps: establishing a cover plate fracture image sample library, marking unit frames in a fracture area or a fuzzy area in each cover plate fracture image sample in a blocking mode, and configuring a category label for each unit frame; the category labels comprise a fault label and a suspected label;
step two: building a fast-Rcnn algorithm frame model, training by adopting cover plate fracture image samples, generating an initial detection frame on a characteristic diagram obtained by each cover plate fracture image sample in a sliding window mode, and correcting the initial detection frame based on a corresponding unit frame to obtain a target frame; gradually adjusting the model parameters until the model converges in the process of obtaining the target frame, and obtaining the weight coefficients corresponding to the fault label and the suspected label;
step three: acquiring an image of a cross rod cover plate of a running truck, and preprocessing the image to be recognized; inputting the image to be recognized into the updated fast-Rcnn algorithm frame model, obtaining the fault prediction area of the image to be recognized and the confidence coefficient corresponding to the fault prediction area based on the weight coefficient, determining the fault prediction area with the confidence coefficient larger than the set threshold value as the fault area, and alarming.
In the embodiment, the class labels are divided into the fault labels and the suspected labels, that is, not only the regions determined as faults in the image sample are marked, but also the regions which are not the fault regions but are easy to be considered as similar visually are marked, so that a Faster-Rcnn algorithm framework model is trained, and the method is beneficial to obtaining more accurate fault recognition results in the process of automatically recognizing the images to be recognized in the follow-up process.
The equipment for acquiring the images of the cross rod cover plate can be a high-definition camera built around a running track of the truck, and the high-definition images of the cross rod cover plate of the truck are acquired after the truck passes through the position of the camera.
The unit frame is used for marking the broken or cracked positions in the cover plate broken image samples by using a tool so as to obtain the XML annotation file of each image sample. In order to prevent interference of other similar broken or cracked forms on the image, the invention divides the image into a plurality of types and marks failure labels or suspected labels respectively, namely marking the unit frames which can be determined as failure with the failure labels, and marking the suspected labels for the unit frames which have other forms similar failures but can be determined as non-failure. Therefore, most interference can be eliminated, and the stability and the recognition rate of recognition can be ensured.
The Faster-Rcnn algorithm framework model first sets the required class labels according to the number of classes. The network structure comprises: an input layer, a pooling layer, a convolution layer, and an output layer. The specific structure is shown in fig. 3.
Further, the cover plate fracture image sample is obtained by preprocessing the collected original cover plate fracture image.
The process of obtaining the cover plate fracture image sample from the original cover plate fracture image comprises the following steps:
based on the image information, the component areas of the crossbar cover are first located and then preprocessed.
The original cover plate fracture image is obtained through camera equipment arranged around the railway wagon, and the camera equipment obtains a linear array image which is a gray image. Due to the badness of the conditions when the truck runs, the part shape of the cross rod cover plate can be influenced by natural conditions such as rainwater and muddy water, so that the diversity and complexity of the part images are ensured when the images are collected, different features are provided for identification, and the identification accuracy is increased to the maximum extent. The cross bar cover plate component is positioned at the bottom of the truck, and the influence of natural or artificial factors on the cross bar cover plate component is more obvious, so that images with different forms can be searched to the maximum extent in the process of collecting original cover plate fracture images.
As an example, the way to pre-process the original cover fracture image includes augmentation, which includes rotating, mirroring, scaling, and enhancing contrast of the image.
And establishing a deep learning training set for the cover plate fracture image sample obtained after amplification. And (4) building a neural network structure, training until the model converges, and obtaining the parameter weight. And detecting the cross rod cover plate according to the weight information, and uploading an alarm if a breaking fault is found, so that the driving safety is ensured.
The original cover plate fracture image is amplified, so that the diversity of data can be improved, and the size and the dimension of the image are uniform, such as the amplified image shown in fig. 2.
Still further, the processing of the fast-Rcnn algorithm framework model on the cover plate fracture image sample comprises:
and obtaining a characteristic diagram from the cover plate fracture image sample.
As shown in fig. 4, the processing of the cover plate fracture image sample by the fast-Rcnn algorithm framework model further includes:
and generating initial detection frames by adopting two parallel regional recommendation network layers for the characteristic diagram in a sliding window mode, and performing pooling processing on all the initial detection frames to obtain a target frame.
As shown in fig. 5 and 6, the sliding window includes three dimensions, and the length-width ratios are { 1: 1,1: 2,2: 1}.
In the embodiment, the method for acquiring the existing feature map from a single image is changed into the method for inputting two images, two regional recommendation networks are obtained, and then the recommendation frames are extracted and sent to the full connection layer.
In the existing algorithm, the RPN network convolves the feature map through a 3 × 3 sliding window to generate a candidate frame; in the embodiment, the sliding window is improved and is respectively realized by 1 × 1, 3 × 3 and 5 × 5 convolutions, and feature fusion is added, and a specific fusion method is splicing, namely, the input feature graph is stacked in a specified dimension. As shown in fig. 4, 5 and 6.
The method adopts a form of two image input networks, and provides another feasible technical scheme for the fault detection of the cross rod cover plate; a plurality of convolution kernels are adopted for convolution, so that multi-scale detection can be more robust, and the detection capability of the model is improved; the method improves the accuracy of fault identification and can better ensure the driving safety.
Because the image in the cover plate fracture image sample library is large, the convolution layer in the fast-Rcnn algorithm frame model comprises a conv layer, a relu layer and a pooling layer, the sizes of convolution kernels of the conv layer are all 3 multiplied by 3, edge expanding processing is carried out on all convolutions in the convolution layer, and the sizes of input matrixes and output matrixes are guaranteed to be unchanged. One important parameter in the construction of RPN networks is anchors, which introduces a multi-scale approach in this detection algorithm. The feature map obtained after convolution has 9 rectangles per point, 3 shapes, and the aspect ratio is { 1: 1,1: 2,2: 1 as the initial detection frame, and then determining the real detection frame position by 2 corrections. Meanwhile, the position relation between a real frame (namely the marked unit frame) and the prediction frame is obtained through a translation and scaling formula, and the target frame is further accurately output.
Calculating the weight of the sample data set:
after the basic network structure of Faster-Rcnn is obtained, the next important step is to obtain the dataset weights. Firstly, taking ImageNet data set weight as initial weight of an image sample, training annotated data according to a back propagation mode, and selecting a loss function and an optimizer before training.
The significance of the loss function is that it can calculate the error value between the model output (predicted result) and the true label of the image, and the smaller this value is, the smaller the error is, the closer the predicted value and the true value are, i.e. the higher the accuracy is. The loss function may consist of two parts:
the first part to the right of the equation is the classification penalty and the second part is the prediction box regression penalty.
Wherein:
Wherein:
where R is the smoothed L1 loss function, which is expressed as:
wherein i is an integer, piFor the probability that the ith prediction is the target,is the probability of marking the sample label, with a value of 0 or 1, tiIn order to predict the coordinates of the frame,to mark the frame coordinates, NclsAs a normalized value of the classification term, NregIs the number of normalized back boxes.
The optimizer uses a numerical method to continuously update the weights and the deviations in the continuous batch training to minimize the loss function error. Because of the advantages of high efficiency, small occupied memory, suitability for large-scale data and the like of the Adam optimization function, the model selects Adam as an optimizer. The formula is as follows:
wherein W is a weight coefficient, WiFor the last weight coefficient or initial weight coefficient, η is the learning rate.
After the basic parameters are set, training can be started on the sample data set. And selecting proper iteration times according to actual conditions, and repeating the training until the loss function is converged and the confidence coefficient reaches a stable value. The optimal weights are saved for predictive use.
In the third step, after the high-definition linear array gray image of the cross rod cover plate is obtained, the area of the cross rod cover plate component to be detected is firstly positioned, and the fast-Rcnn model weight is loaded for prediction. A confidence threshold value can be set by self, and when the model predicts that a certain area in the image is possible to be a fault, if the confidence threshold value is lower than the threshold value, no alarm is given; and if the threshold value is higher than the threshold value, the information is directly uploaded to an alarm platform. The set threshold value must be obtained through a large number of prediction experiments to be representative. The most suitable threshold is selected to ensure the highest recognition accuracy.
Although the invention herein has been described with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present invention. It is therefore to be understood that numerous modifications may be made to the illustrative embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that features described in different dependent claims and herein may be combined in ways different from those described in the original claims. It is also to be understood that features described in connection with individual embodiments may be used in other described embodiments.
Claims (6)
1. A rail wagon cross rod cover plate fracture detection method is characterized by comprising the following steps:
the method comprises the following steps: establishing a cover plate fracture image sample library, marking unit frames in a fracture area or a fuzzy area in each cover plate fracture image sample in a blocking mode, and configuring a category label for each unit frame; the category labels comprise a fault label and a suspected label;
step two: building a fast-Rcnn algorithm frame model, training by adopting cover plate fracture image samples, generating an initial detection frame on a characteristic diagram obtained by each cover plate fracture image sample in a sliding window mode, and correcting the initial detection frame based on a corresponding unit frame to obtain a target frame; gradually adjusting the model parameters until the model converges in the process of obtaining the target frame, and obtaining the weight coefficients corresponding to the fault label and the suspected label;
step three: acquiring an image of a cross rod cover plate of a running truck, and preprocessing the image to be recognized; inputting the image to be recognized into the updated fast-Rcnn algorithm frame model, obtaining the fault prediction area of the image to be recognized and the confidence coefficient corresponding to the fault prediction area based on the weight coefficient, determining the fault prediction area with the confidence coefficient larger than the set threshold value as the fault area, and alarming.
2. The rail wagon cross bar cover plate fracture detection method as claimed in claim 1, wherein the cover plate fracture image sample is obtained by preprocessing the acquired original cover plate fracture image.
3. The rail wagon cross-bar cover fracture detection method of claim 2, wherein the preprocessing of the original cover fracture image comprises amplification, the amplification comprising image rotation, mirroring, scaling and contrast enhancement.
4. The rail wagon cross-bar cover plate fracture detection method according to claim 3, wherein the processing of the cover plate fracture image sample by the Faster-Rcnn algorithm framework model comprises:
and obtaining a characteristic diagram from the cover plate fracture image sample.
5. The rail wagon cross-bar cover fracture detection method of claim 4, wherein the processing of the cover fracture image sample by the Faster-Rcnn algorithm framework model further comprises:
and generating initial detection frames by adopting two parallel regional recommendation network layers for the characteristic diagram in a sliding window mode, and performing pooling processing on all the initial detection frames to obtain a target frame.
6. The rail wagon cross bar cover plate fracture detection method of claim 5, wherein the sliding window comprises three dimensions, and the length-width ratios are respectively { 1: 1,1: 2,2: 1}.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911293655.1A CN111091150A (en) | 2019-12-12 | 2019-12-12 | Railway wagon cross rod cover plate fracture detection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911293655.1A CN111091150A (en) | 2019-12-12 | 2019-12-12 | Railway wagon cross rod cover plate fracture detection method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111091150A true CN111091150A (en) | 2020-05-01 |
Family
ID=70395092
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911293655.1A Pending CN111091150A (en) | 2019-12-12 | 2019-12-12 | Railway wagon cross rod cover plate fracture detection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111091150A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104573744A (en) * | 2015-01-19 | 2015-04-29 | 上海交通大学 | Fine granularity classification recognition method and object part location and feature extraction method thereof |
CN105354568A (en) * | 2015-08-24 | 2016-02-24 | 西安电子科技大学 | Convolutional neural network based vehicle logo identification method |
CN109946746A (en) * | 2019-03-21 | 2019-06-28 | 长安大学 | A kind of safe examination system and method based on deep neural network |
CN110097053A (en) * | 2019-04-24 | 2019-08-06 | 上海电力学院 | A kind of power equipment appearance defect inspection method based on improvement Faster-RCNN |
-
2019
- 2019-12-12 CN CN201911293655.1A patent/CN111091150A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104573744A (en) * | 2015-01-19 | 2015-04-29 | 上海交通大学 | Fine granularity classification recognition method and object part location and feature extraction method thereof |
CN105354568A (en) * | 2015-08-24 | 2016-02-24 | 西安电子科技大学 | Convolutional neural network based vehicle logo identification method |
CN109946746A (en) * | 2019-03-21 | 2019-06-28 | 长安大学 | A kind of safe examination system and method based on deep neural network |
CN110097053A (en) * | 2019-04-24 | 2019-08-06 | 上海电力学院 | A kind of power equipment appearance defect inspection method based on improvement Faster-RCNN |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111079747B (en) | Railway wagon bogie side frame fracture fault image identification method | |
CN111754498A (en) | Conveyor belt carrier roller detection method based on YOLOv3 | |
CN112381788B (en) | Part surface defect increment detection method based on double-branch matching network | |
KR20200012334A (en) | Method for detecting defects in semiconductor device | |
CN111881970A (en) | Intelligent outer broken image identification method based on deep learning | |
CN111160432A (en) | Automatic classification method and system for panel production defects | |
CN116883763B (en) | Deep learning-based automobile part defect detection method and system | |
CN112171057A (en) | Quality detection method and device based on laser welding and storage medium | |
CN117349583A (en) | Intelligent detection method and system for low-temperature liquid storage tank | |
CN117372433B (en) | Thickness parameter control method, device, equipment and storage medium | |
CN114299040A (en) | Ceramic tile flaw detection method and device and electronic equipment | |
CN113724219A (en) | Building surface disease detection method and system based on convolutional neural network | |
CN115830302B (en) | Multi-scale feature extraction fusion power distribution network equipment positioning identification method | |
CN116740728A (en) | Dynamic acquisition method and system for wafer code reader | |
CN116678888A (en) | Component defect detection method and device | |
CN111091150A (en) | Railway wagon cross rod cover plate fracture detection method | |
CN110569717A (en) | partial discharge detection method, device, system, equipment and readable storage medium | |
CN116030056A (en) | Detection method and system for steel surface cracks | |
CN115082650A (en) | Implementation method of automatic pipeline defect labeling tool based on convolutional neural network | |
CN116091416A (en) | Method and device for training assembly defect detection and change detection models of printed circuit board | |
KR20230063742A (en) | Method for detecting defect of product using hierarchical CNN in smart factory, and recording medium thereof | |
KR102629092B1 (en) | System and method for detecting defects in solar panel | |
CN118037741B (en) | Welding quality detection method and system for herringbone doors based on machine vision | |
CN117593890B (en) | Detection method and device for road spilled objects, electronic equipment and storage medium | |
KR102664206B1 (en) | Aircraft inspection automation system based on image and explainable ai |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200501 |
|
RJ01 | Rejection of invention patent application after publication |