CN113298751A - Detection method for auxiliary door blockage - Google Patents
Detection method for auxiliary door blockage Download PDFInfo
- Publication number
- CN113298751A CN113298751A CN202011050400.5A CN202011050400A CN113298751A CN 113298751 A CN113298751 A CN 113298751A CN 202011050400 A CN202011050400 A CN 202011050400A CN 113298751 A CN113298751 A CN 113298751A
- Authority
- CN
- China
- Prior art keywords
- blocked
- neural network
- network model
- image recognition
- residual error
- 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 39
- 238000003062 neural network model Methods 0.000 claims abstract description 52
- 239000000463 material Substances 0.000 claims abstract description 46
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 238000000034 method Methods 0.000 claims description 32
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 21
- 238000012795 verification Methods 0.000 claims description 16
- 238000009826 distribution Methods 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims 1
- 238000005245 sintering Methods 0.000 abstract description 8
- 230000000903 blocking effect Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 description 13
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 6
- 239000002994 raw material Substances 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 238000001035 drying Methods 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 229910052742 iron Inorganic materials 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000001914 filtration Methods 0.000 description 2
- 239000000446 fuel Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000002156 mixing Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000005465 channeling Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000009851 ferrous metallurgy Methods 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000007711 solidification Methods 0.000 description 1
- 230000008023 solidification Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a detection method of auxiliary door putty, including: installing a camera in the auxiliary door blanking area; acquiring field images below a plurality of auxiliary doors within a period of time; preprocessing a plurality of field images; labeling the preprocessed field image to obtain a labeled image set; the labels comprise a blocked large label and a normal label; constructing a blocked block image recognition residual error neural network model, and training and verifying the blocked block image recognition residual error neural network model by using a marked image set; and inputting field images acquired according to a fixed frequency into the trained and verified blocked large block image recognition residual error neural network model, detecting whether a material blocking occurs in the auxiliary gate blanking area, and outputting a detection result. The problem that whether the auxiliary door material blocking happens or not is judged through human eyes by an operator at present, the labor cost is increased, and due to the fact that hysteresis exists in manual processing, the material layer thickness is not uniform, and the quality and the yield of the sintering ore are affected can be solved.
Description
Technical Field
The application relates to the technical field of image processing, in particular to a method for detecting material blockage of an auxiliary door.
Background
In the ferrous metallurgy industry, sintering is an important pre-process. For example, before blast furnace iron making, iron-containing powder, fuel and flux are mixed according to the process proportion, and the iron-containing powder is bonded into block sintered ore by using a small amount of molten mass generated by the combustion heat of the fuel and a solidification reaction. The sintering process generally includes batching, primary mixing and distributing, secondary mixing and distributing, ignition sintering, dust removal and air draft, crushing and cooling. The material distribution process steps play a crucial role in the quality of the sinter, and if the material layer on the material distribution trolley is smooth in surface and uniform in thickness, the whole air permeability of the material layer can be improved, and the quality of the sinter is improved.
In general, the material distribution process step adjusts the flow rate of the material distribution through a main door and an auxiliary door control system so as to control the thickness of a material layer. However, in the production process, when the size of raw material block exceeded the current aperture of assisting the door, it took place raw material block card between the supplementary door and the round roller of round roller cloth machine easily, lead to the discharge amount to reduce, and then cause the bed of material thickness on the platform truck to reduce and bed of material thickness inhomogeneous, finally influence the quality and the yield of sintering deposit.
In the present sintering production process, whether take place to assist a door putty through eyes judgement usually, again because of the human cost's consideration, usually operating personnel need compromise a plurality of operation posts, often takes place the condition that can not in time discover to assist a door putty. In addition, even if an operator uses a tool to knock the large material to discharge the large material or fully opens the auxiliary door to discharge the large material, the hysteresis of manual processing exists, and when the large material is inserted manually, the condition that the thickness of the material layer is uneven or the material surface is in a groove is generated, so that the quality and the yield of the sintering ore are still influenced.
Disclosure of Invention
The application provides a detection method of auxiliary door putty to solve whether operating personnel judge through people's eye and take place auxiliary door putty, increase the human cost, and because there is the hysteresis quality in manual handling, lead to the inhomogeneous or charge level ditching of bed of material thickness, influence the quality and the yield scheduling problem of sintering deposit.
A detection method for auxiliary door blockage comprises the following steps:
installing a camera in the auxiliary door blanking area;
acquiring field images below a plurality of auxiliary doors within a period of time;
preprocessing a plurality of the field images;
labeling the preprocessed field image to obtain a labeled image set; the labels comprise a blocked big label and a normal label;
constructing a blocked block image recognition residual error neural network model, and training and verifying the blocked block image recognition residual error neural network model by using the marked image set;
inputting the field images collected according to a fixed frequency into the trained and verified blocked large block image recognition residual neural network model, detecting whether the auxiliary gate blanking area is blocked or not, and outputting a detection result; the detection result comprises that the blanking area of the auxiliary door is blocked and the blanking area of the auxiliary door is not blocked.
The application provides a detection method of auxiliary door putty, through constructing stifled bold image recognition residual error neural network model, and install the camera in auxiliary door blanking region, with stifled bold image recognition residual error neural network model of field image input that the camera was gathered, monitor the operation condition in auxiliary door blanking region, whether can intelligent detection auxiliary door blanking region takes place the putty, thereby in time adjust the aperture of auxiliary door according to the monitoring condition, the condition of mediation putty, avoid taking place because the putty causes the uneven condition of bed of material thickness. Can replace operating personnel's people's eye to judge whether take place to assist a door putty, can practice thrift the human cost. In addition, the problem that the quality and the yield of the sintered ore are influenced due to uneven material layer thickness or material surface channeling caused by hysteresis of manual treatment can be solved.
Drawings
In order to more clearly explain the technical solution of the present application, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious to those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a partial side view of a circular roller distributor;
FIG. 2 is a side view of the blanking area of the auxiliary door without material blockage;
FIG. 3 is a side view of a blanking area of an auxiliary door with material blockage;
fig. 4 is a flowchart of a method for detecting material blockage of an auxiliary door according to an embodiment of the present disclosure;
FIG. 5 is a schematic view of a camera mounting location;
FIG. 6 is a field image of the auxiliary door blanking area without material blockage;
FIG. 7 is a field image of a blanking area of an auxiliary door with material blockage;
FIG. 8 is a block image recognition residual error neural network model architecture diagram corresponding to the detection method for auxiliary door block shown in FIG. 4;
FIG. 9 is a flow chart of a first training and verification process for the block-blocked image recognition residual neural network model shown in FIG. 4;
fig. 10 is a flow chart of training and verifying the second block image recognition residual neural network model shown in fig. 4.
Detailed Description
The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, 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 application.
Fig. 1 is a partial side view of a circular roller distributor. As shown in fig. 1, the storage bin 1 is used for containing raw materials, and the raw materials are distributed through a main door 3 and an auxiliary door 4 in sequence along with the rotation of a round roller 2. In the material distribution process step, the main door 3 and the auxiliary door 4 are controlled by a control system to adjust the flow of the material distribution so as to control the thickness of a material layer. However, in the production process, when the size of the raw material block exceeds the current opening degree of the auxiliary door 4, it is easy to occur that the raw material block is caught between the auxiliary door 4 and the round roller 2 of the round roller distributor. Fig. 2 is a side view of the blanking area of the auxiliary door without material blockage, and fig. 3 is a side view of the blanking area of the auxiliary door with material blockage. Fig. 2 and fig. 3 respectively illustrate schematic diagrams of the auxiliary door blanking area when no material blockage occurs and when material blockage occurs.
Fig. 4 is a flowchart of a method for detecting material blockage of an auxiliary door according to an embodiment of the present application. As shown in fig. 4, in the method for detecting material jam of the auxiliary door provided in this embodiment, first, step S1 is executed: and a camera is arranged in the blanking area of the auxiliary door. Fig. 5 is a schematic view of a camera mounting position. As shown in fig. 5, cameras are installed opposite to the blanking area of the auxiliary door, and the type, number and installation angle of the installed cameras can be selected according to actual needs, which is not specifically limited in the present application. After the camera is installed, the camera needs to be debugged correspondingly, which is not described herein.
After the camera debugging is completed, the process proceeds to step S2: and acquiring field images below a plurality of auxiliary doors within a period of time. The method comprises the steps of collecting field images in a time period, wherein the situation that auxiliary door blocking occurs and the situation that the auxiliary door blocking does not occur need to exist in the time period, so that all situations of an auxiliary door blanking area when a circular roller distributing machine operates can be covered by the collected field images. FIG. 6 is a field image of the auxiliary door blanking area without material blockage; fig. 7 is a field image of the blanking area of the auxiliary door with material blockage. Fig. 6 and 7 show the field images of the blanking area of the auxiliary door without blockage and with blockage.
After the field image acquisition is completed, the process continues to step S3: and preprocessing a plurality of field images.
The specific pretreatment mode can include the following modes:
the size adjustment may be to adjust the field image to a set pixel size, where the set pixel size may be 1024 pixels by 1024 pixels, or may be other pixel sizes, and is not limited herein.
And adjusting the brightness, wherein the brightness of the live image can be adjusted according to the following formula:
g(x,y)＝a*f(x,y)+b，0＜a＜3，
wherein g (x, y) is the three channel values of red, green and blue of the output target image x row and y column pixel, f (x, y) is the input field image x row, the three channel values of red, green and blue of y column pixel, a is the amplification coefficient, b is the bias coefficient, a is the positive number, the value can be between 0-3, and no specific limitation is made.
The smoothing and drying process may be a linear filtering process on the live image according to the following formula:
wherein, g (i, j) is the pixel value of the output target image, (i, j) is the coordinate value of the pixel point of the target image, f (i + k, j + l) is the pixel value of the input field image, (k, l) is the coordinate value of the pixel point of the field image, and h (k, l) is the weighting coefficient of the filter.
The linear filtering process is a smoothing and drying process, and the specific modes of the resizing, brightness adjustment and smoothing and drying processes are only illustrative, and the present application is not limited thereto.
After the field image preprocessing is completed, the process proceeds to step S4: labeling the preprocessed field image to obtain a labeled image set; the labels comprise a blocked big label and a normal label. For example, the blocked block tag may be represented by 1, the normal tag may be represented by 0, and the tag may be marked in the image attribute of the live image, which is not specifically limited in this application.
After the labeled image set is formed, the labeled image set may be divided into a training set and a verification set, and then the step S5 is continuously executed: and constructing a blocked block image recognition residual error neural network model, and training and verifying the blocked block image recognition residual error neural network model by using the marked image set. Specifically, the training set is used for training the blocked image recognition residual error neural network model, and the verification set is used for verifying the blocked image recognition residual error neural network model. Because the training of the block image recognition residual error neural network model is a process of repeatedly learning the model, the required data volume is large, and relatively speaking, the verification process of the model does not need excessive data volume, so that the number of the marked images in the training set can be larger than that of the marked images in the verification set. For example, 80% of the labeled images in the labeled image set may be used as the training set, and the remaining 20% of the labeled images may be used as the verification set.
FIG. 8 is a block image recognition residual error neural network architecture diagram of the detection method for auxiliary door block material of FIG. 4; fig. 9 is a flowchart of training and verifying the first block image recognition residual neural network model shown in fig. 4. In detail, with reference to fig. 1, 8 and 9, step S5 may include the following steps:
s51: and constructing a blocked block image recognition residual error neural network model, wherein the blocked block image recognition residual error neural network model comprises an input layer, a plurality of residual error blocks, a full connection layer, a classifier and an output layer. The residual block may include a convolutional layer and a pooling layer; the output of the convolution layer is used as the input of the pooling layer, and the output of the pooling layer is used as the output of the residual block; the convolution layer is used for carrying out standardized operation and regularization operation on the input marked image and outputting image operation data; the pooling layer is used for performing dimension reduction processing on the image operation data output by the convolution layer. The dimensions of each residual block may be different from each other, and the dimensions of each residual block are determined by the convolutional layer and the pooling layer. The dimension of the convolution layer may be set to 32 dimensions, the step size may be 1, and convolution is performed by 3 × 3, which is not particularly limited in the present application.
S52: parameters in the residual block are initialized. The equation of the convolutional layer may be a linear equation, and the parameter of the convolutional layer is a parameter of the linear equation, and specifically may include a weight value and a bias value, and the parameter may further include a batch normalization scale factor. In the subsequent model training process, the weight values and the bias values are mainly trained. Since the parameter update of the front layer training will cause the change of the input data distribution of the back layer in the model training process, the input data distribution of each layer is changed all the time, therefore, batch standardization algorithm can be set for the convolution layer and the full connection layer to keep the same input data distribution. The individual layers may also be set with batch normalization scale factors, which also serve to normalize for each input sample.
S53: inputting the marked images in the training set into a residual block through an input layer, and outputting image operation data after performing standardized operation and regularization operation; and the image operation data output by each residual block is used as the input data of the next residual block, and the image operation data is finally output after all the residual blocks are traversed.
S54: and inputting the image operation data output by the last residual block into the full-connection layer, and outputting the multi-dimensional characteristic matrix after carrying out standardized operation.
S55: and inputting the multi-dimensional feature matrix into a classifier for classification, and outputting a detection result through an output layer. The classifier can adopt a softmax classifier and is used for carrying out feature classification on the multi-dimensional feature matrix to finally obtain a detection result, and the detection result can be represented in the same form with the label of the marked image so as to facilitate subsequent comparison.
S56: and comparing the detection result with the label of the marked image to obtain a comparison result. And comparing the detection result with the label of the marked image to compare whether the detection result is the same as the label.
S57: and according to the comparison result, correcting the parameters of the residual block until all the marked images in the training set are used up, and finishing the training. If the comparison result is that the detection result is the same as the label, the parameter does not need to be corrected, and if the comparison result is that the detection result is different from the label, the parameter needs to be corrected correspondingly. When all the labeled images in the training set have undergone the steps S53-S57, the model training is finished.
S58: and inputting all marked images of the verification set into the trained blocked block image recognition residual error neural network model, and verifying the blocked block image recognition residual error neural network model.
As shown in fig. 9, step S58 may specifically include the following steps:
s581: and inputting all the marked images of the verification set into the trained blocked image identification residual error neural network model, and outputting a detection result.
S582: and inputting the detection result into a loss function to obtain the loss rate.
The formula for the loss function is as follows:
wherein L is the loss rate, ynFor the input n-th label of the label image, yn' is the corresponding detection result outputted, and N is the total number of the marker images.
S583: judging whether the loss rate is less than or equal to a target loss value; the target loss value of the loss rate may be set to 0.01.
S584: and if the loss rate is greater than the target loss value, continuing training the blocked block image recognition residual error neural network model until the loss rate is less than or equal to the target loss value, and stopping training.
And if the loss rate is less than or equal to the target loss value, the blockage block image recognition residual error neural network model is verified to obtain a final blockage block image recognition residual error neural network model.
Fig. 10 is a flow chart of training and verifying the second block image recognition residual neural network model shown in fig. 4. As shown in fig. 10, in another embodiment, step S58 may include the following steps:
s581': and inputting all the marked images of the verification set into the trained blocked image identification residual error neural network model, and outputting a detection result.
S582': and comparing the detection result with the label of the corresponding marked image to obtain a comparison result.
S583': and (5) counting the error rate of the comparison result. The error rate may be obtained by dividing the number of errors by the total input amount.
S584': and if the error rate is greater than the error threshold value, continuing training the block image identification residual error neural network model until the error rate is less than or equal to the error threshold value, and stopping training. The error threshold may be set according to actual conditions, and the application is not particularly limited.
And if the error rate is less than or equal to the error threshold value, the blocked block image recognition residual error neural network model is verified to obtain the final blocked block image recognition residual error neural network model. After the final blocked block image recognition residual error neural network model is obtained, the blocked block image recognition residual error neural network model can be packaged, and the packaged model is an executable program file.
S6: inputting field images collected according to a fixed frequency into a trained and verified blocked large block image recognition residual neural network model, detecting whether a material blocking occurs in an auxiliary gate blanking area, and outputting a detection result; the detection result comprises that the blanking area of the auxiliary door is blocked and the blanking area of the auxiliary door is not blocked. The fixed frequency may be one frame of image collected per second, or one frame of image collected per two seconds, and may be set according to actual needs, which is not specifically limited in the present application.
And when the detection result is that the blanking area of the auxiliary door is blocked, the blocking alarm is carried out. The putty warning can remind the staff in time to adjust the auxiliary opening degree, or the putty warning can be connected to the adjustment procedure of auxiliary opening degree, carries out the automatic adjustment aperture, and this application does not do not specifically limit.
According to the detection method for the auxiliary door blockage, the blockage massive image recognition residual error neural network model is constructed, the camera is installed in the auxiliary door blanking area, the field image collected by the camera is input into the blockage massive image recognition residual error neural network model, the operation condition of the auxiliary door blanking area is monitored, whether blockage occurs in the auxiliary door blanking area can be intelligently detected, the opening degree of the auxiliary door is timely adjusted according to the monitoring condition, the blockage is dredged, and the condition that the material layer thickness is uneven due to the blockage is avoided. Can replace operating personnel's people's eye to judge whether take place to assist a door putty, can practice thrift the human cost. In addition, the problem that the quality and the yield of the sintered ore are influenced due to uneven thickness of a material layer caused by hysteresis of manual treatment can be solved.
The same and similar parts in the various embodiments in this specification may be referred to each other.
Claims (10)
1. A detection method for auxiliary door blockage is characterized by comprising the following steps:
installing a camera in the auxiliary door blanking area;
acquiring field images below a plurality of auxiliary doors within a period of time;
preprocessing a plurality of the field images;
labeling the preprocessed field image to obtain a labeled image set; the labels comprise a blocked big label and a normal label;
constructing a blocked block image recognition residual error neural network model, and training and verifying the blocked block image recognition residual error neural network model by using the marked image set;
inputting the field image collected according to a fixed frequency into the trained and verified blocked block image recognition residual error neural network model, detecting whether the auxiliary gate blanking area is blocked or not, and outputting a detection result; the detection result comprises that the blanking area of the auxiliary door is blocked and the blanking area of the auxiliary door is not blocked.
2. The method for detecting door jam as claimed in claim 1, wherein the preprocessing of the preprocessing step on the plurality of field images comprises:
carrying out size adjustment on the field image to set pixel size;
adjusting the brightness of the field image;
and carrying out smooth denoising processing on the adjusted field image.
3. The method for detecting the material blockage of the auxiliary door according to claim 1, wherein a material blockage alarm is given when the detection result indicates that the material blockage occurs in the blanking area of the auxiliary door.
4. The method of detecting door blockage according to claim 1, wherein the set of labeled images is divided into a training set and a validation set;
the method comprises the following steps of constructing a blocked block image recognition residual error neural network model, and training and verifying the blocked block image recognition residual error neural network by using a marked image set, wherein the steps comprise:
constructing a blocked large image recognition residual error neural network model, wherein the blocked large image recognition residual error neural network model comprises an input layer, a plurality of residual error blocks, a full connection layer, a classifier and an output layer;
initializing parameters in the residual block;
inputting the marked images in the training set into the residual block through an input layer, and outputting image operation data after performing standardized operation and regularization operation; the image operation data output by each residual block is used as the input data of the next residual block, and all the residual blocks are traversed;
inputting the image operation data output by the last residual block into the full-connection layer, and outputting a multi-dimensional feature matrix after carrying out standardized operation;
inputting the multi-dimensional feature matrix into the classifier for classification, and outputting a detection result through the output layer;
comparing the detection result with the label of the marked image to obtain a comparison result;
correcting the parameters of the residual block according to the comparison result until all the marked images in the training set are used up, and finishing the training;
inputting all the marked images of the verification set into the trained blocked large image recognition residual error neural network model, and verifying the blocked large image recognition residual error neural network model;
and if the verification is passed, obtaining the trained and verified blocked image recognition residual error neural network model.
5. The method for detecting the auxiliary door jam as claimed in claim 4, wherein the residual block comprises a convolution layer and a pooling layer; the output of the convolutional layer is used as the input of the pooling layer, and the output of the pooling layer is used as the output of the residual block; the convolution layer is used for carrying out standardization operation and regularization operation on the input marked image and outputting the image operation data; the pooling layer is used for performing dimensionality reduction processing on the image operation data output by the convolution layer.
6. The method for detecting the auxiliary door blockage according to claim 5, wherein the step of inputting all the marked images of the verification set into the trained blocked large image recognition residual error neural network model and verifying the blocked large image recognition residual error neural network model comprises the following steps:
inputting all the marked images of the verification set into the trained blocked image recognition residual error neural network model, and outputting the detection result;
inputting the detection result into a loss function to obtain a loss rate;
judging whether the loss rate is less than or equal to a target loss value;
if the loss rate is larger than a target loss value, continuing to train the blocked block image recognition residual error neural network model;
and stopping training if the loss rate is less than or equal to the target loss value.
7. The method for detecting the auxiliary door blockage according to claim 5, wherein the step of inputting all the marked images of the verification set into the trained blocked large image recognition residual error neural network model and verifying the blocked large image recognition residual error neural network model comprises the following steps:
inputting all the marked images of the verification set into the trained blocked image recognition residual error neural network model, and outputting the detection result;
comparing the detection result with the label corresponding to the marked image to obtain a comparison result;
counting the error rate of the comparison result;
if the error rate is larger than an error threshold value, continuing to train the blocked block image recognition residual error neural network model;
stopping training if the error rate is less than or equal to the error threshold.
8. The method for detecting door jam according to claim 6 or 7, wherein the number of the labeled images in the training set is greater than the number of the labeled images in the verification set.
9. The method of claim 5, wherein the dimensions of each of the residual blocks are different from each other.
10. The method for detecting the auxiliary door jam as claimed in claim 5, wherein the convolution layer and the full connection layer are provided with a batch standardization algorithm for keeping the input data in the same distribution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011050400.5A CN113298751A (en) | 2020-09-29 | 2020-09-29 | Detection method for auxiliary door blockage |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011050400.5A CN113298751A (en) | 2020-09-29 | 2020-09-29 | Detection method for auxiliary door blockage |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113298751A true CN113298751A (en) | 2021-08-24 |
Family
ID=77318303
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011050400.5A Pending CN113298751A (en) | 2020-09-29 | 2020-09-29 | Detection method for auxiliary door blockage |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113298751A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117036362A (en) * | 2023-10-10 | 2023-11-10 | 广州国巡机器人科技有限公司 | Method, device, equipment and storage medium for detecting blocking material |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944085A (en) * | 2017-10-27 | 2018-04-20 | 中冶南方工程技术有限公司 | A kind of data processing method and module for steel rolling self learning model |
| CN108510488A (en) * | 2018-03-30 | 2018-09-07 | 安徽理工大学 | Four kinds of damage detecting methods of conveyer belt based on residual error network |
| CN109879005A (en) * | 2019-04-15 | 2019-06-14 | 天津美腾科技有限公司 | Device for detecting belt tearing and method |
| CN110349146A (en) * | 2019-07-11 | 2019-10-18 | 中原工学院 | The building method of fabric defect identifying system based on lightweight convolutional neural networks |
| US20200193279A1 (en) * | 2018-12-13 | 2020-06-18 | Sri International | Runtime-throttleable neural networks |
| US20200242357A1 (en) * | 2019-01-29 | 2020-07-30 | Accenture Global Solutions Limited | Distributed and self-validating computer vision for dense object detection in digital images |
-
2020
- 2020-09-29 CN CN202011050400.5A patent/CN113298751A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944085A (en) * | 2017-10-27 | 2018-04-20 | 中冶南方工程技术有限公司 | A kind of data processing method and module for steel rolling self learning model |
| CN108510488A (en) * | 2018-03-30 | 2018-09-07 | 安徽理工大学 | Four kinds of damage detecting methods of conveyer belt based on residual error network |
| US20200193279A1 (en) * | 2018-12-13 | 2020-06-18 | Sri International | Runtime-throttleable neural networks |
| US20200242357A1 (en) * | 2019-01-29 | 2020-07-30 | Accenture Global Solutions Limited | Distributed and self-validating computer vision for dense object detection in digital images |
| CN109879005A (en) * | 2019-04-15 | 2019-06-14 | 天津美腾科技有限公司 | Device for detecting belt tearing and method |
| CN110349146A (en) * | 2019-07-11 | 2019-10-18 | 中原工学院 | The building method of fabric defect identifying system based on lightweight convolutional neural networks |
Non-Patent Citations (2)
| Title |
|---|
| SHUKAI FAN,ET AL.: "A Structured Bidirectional LSTM Deep Learning Method For 3D Terahertz Indoor Localization", 《IEEE INFOCOM 2020》 * |
| 汪冬冬 等: "基于阴影检测的传送带烟丝堵料视觉检测系统设计", 《轻工学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117036362A (en) * | 2023-10-10 | 2023-11-10 | 广州国巡机器人科技有限公司 | Method, device, equipment and storage medium for detecting blocking material |
| CN117036362B (en) * | 2023-10-10 | 2024-02-20 | 广州国巡机器人科技有限公司 | Method, device, equipment and storage medium for detecting blocking material |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100508555C (en) | Static object detecting method and system | |
| CN108062542B (en) | Method for detecting shielded human face | |
| CN113297885A (en) | Belt conveyor surface state detection method and device based on convolutional neural network | |
| CN113298751A (en) | Detection method for auxiliary door blockage | |
| CN110544261A (en) | Blast furnace tuyere coal injection state detection method based on image processing | |
| CN111476712B (en) | Trolley grate image shooting and detecting method and system of sintering machine | |
| CN113466236A (en) | Inspection result display device and recording medium | |
| CN116229053A (en) | Dynamic adjustment method for red date machine striker plate based on neural network | |
| CN113077423B (en) | Laser selective melting pool image analysis system based on convolutional neural network | |
| CN111223094B (en) | Trolley grate spacing detection method and system for sintering machine | |
| CN111222360A (en) | Method and device for detecting melting state of silicon material and storage medium | |
| CN101794377A (en) | The image processing apparatus of the coordinate position of the characteristic portion of detection face | |
| CN111768390B (en) | Image recognition-based grate cooler red river recognition method | |
| CN104567405B (en) | A kind of feed bin auxiliary door aperture control method and device | |
| CN112255148A (en) | Ore granularity detection system and detection method | |
| CN106485652A (en) | Setting contrast method and device | |
| CN116189089A (en) | Intelligent video monitoring method and system | |
| CN113298750A (en) | Detection method for wheel falling of circular cooler | |
| CN109190577B (en) | Image signal detection method for taxi passenger seat abnormity by combining headrest and human face characteristics | |
| CN107437256B (en) | Automatic dividing method and system applied to tire X-ray image channel | |
| CN114862750A (en) | Method and equipment for acquiring clear image and refrigerator | |
| CN114863277A (en) | Machine vision-based method for rapidly detecting irregular particle group overrun particles | |
| CN111223095B (en) | Trolley grate spacing detection method and system for sintering machine | |
| CN111222510A (en) | Trolley grate bar image shooting method and system of sintering machine | |
| CN117541615B (en) | Bolt loosening detection method of automatic lapping device |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210824 |