CN114323709A - Driving device of bogie bottom detection platform and detection platform - Google Patents
Driving device of bogie bottom detection platform and detection platform Download PDFInfo
- Publication number
- CN114323709A CN114323709A CN202111479897.7A CN202111479897A CN114323709A CN 114323709 A CN114323709 A CN 114323709A CN 202111479897 A CN202111479897 A CN 202111479897A CN 114323709 A CN114323709 A CN 114323709A
- Authority
- CN
- China
- Prior art keywords
- rod
- driving
- assembly
- driving assembly
- bogie
- 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 61
- 239000013589 supplement Substances 0.000 claims description 13
- 238000004140 cleaning Methods 0.000 claims description 10
- 239000007921 spray Substances 0.000 claims description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 10
- 238000007689 inspection Methods 0.000 claims description 7
- 239000010865 sewage Substances 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 4
- 238000005406 washing Methods 0.000 claims description 3
- 238000003556 assay Methods 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 25
- 238000011176 pooling Methods 0.000 description 19
- 238000000034 method Methods 0.000 description 18
- 238000012549 training Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 10
- 230000008569 process Effects 0.000 description 10
- 230000007246 mechanism Effects 0.000 description 6
- 230000004913 activation Effects 0.000 description 4
- 238000007605 air drying Methods 0.000 description 4
- 230000002776 aggregation Effects 0.000 description 3
- 238000004220 aggregation Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 230000003321 amplification Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 239000000428 dust Substances 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000003199 nucleic acid amplification method Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000011179 visual inspection Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 230000003137 locomotive effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Landscapes
- Length Measuring Devices By Optical Means (AREA)
Abstract
The embodiment of the invention provides a driving device of a bogie bottom detection platform and the detection platform, wherein the driving device drives a second driving assembly to lift or drop pressure along a path close to or far away from the bottom of a bogie through a telescopic assembly; the first driving assembly and the second driving assembly are connected through the third driving assembly, the third driving assembly drives the first driving assembly to rotate relative to the second driving assembly, the camera is driven to transversely move through the first driving assembly, and the first driving assembly is driven to longitudinally move through the second driving assembly. Because the driving device has a plurality of degrees of freedom, the position required by the optimal image acquisition at the bottom of the bogie can be reached, and the reliability of the image acquisition and the accuracy of the detection result are ensured.
Description
Technical Field
The invention relates to the technical field of vehicle detection, in particular to a driving device of a bogie bottom detection platform and the detection platform.
Background
Rail vehicles such as high-speed rails and subway trains are very important vehicles in daily life, and the bottom of the rail vehicle comprises a bogie and the like, and needs to be detected and overhauled regularly. At present, detection is mainly finished by means of visual inspection, and after a trench is dug at the bottom of a locomotive, a maintainer overhauls the trench by means of visual inspection.
Among the prior art, carry out the large tracts of land scanning to vehicle bottom and side by fixed detection camera, form preliminary three-dimensional module, possess preliminary fault diagnosis function, but this kind of technique to the shooting in-process of bogie bottom again, the phenomenon of shooting omission appears easily, can't accurately obtain the image, and through artificially adjusting camera angle and position, face personnel, place configuration lack in short supply, maintainer intensity of labour is big, the operational environment is abominable, the cycle of operation is long, easy tired hourglass is examined the scheduling problem.
Disclosure of Invention
The embodiment of the invention provides a driving device of a bogie bottom detection platform and the detection platform, which can automatically adjust the position of a camera and accurately obtain an image.
In a first aspect, an embodiment of the present invention provides a driving device for a bogie bottom detection platform, where the driving device includes:
the first driving component is used for driving the camera to move transversely;
a second drive assembly for driving the first drive assembly in longitudinal movement;
a telescoping assembly for driving the second drive assembly to lift or lower along a path closer to or further from the bottom of the bogie;
and the first driving assembly and the second driving assembly are connected through the third driving assembly, the third driving assembly drives the first driving assembly to rotate relative to the second driving assembly, and a plane formed by the rotation of the first driving assembly is parallel to a plane where the bottom of the bogie is located.
As an achievable embodiment of the present invention, the first drive assembly includes:
a first plate configured with a pair of guide rods configured with a camera mount;
the first stepping motor is connected with a first screw rod, the first screw rod is connected with the camera mounting seat, and the first stepping motor drives the first screw rod to enable the camera mounting seat to move along the pair of guide rods.
As an achievable embodiment of the present invention, the second drive assembly includes:
the second plate is provided with a guide rail, the guide rail is provided with a connecting seat, and the connecting seat is connected with the first driving assembly;
the second stepping motor is connected with a second screw rod, the second screw rod is connected with the connecting seat, and the second stepping motor drives the second screw rod to enable the connecting seat to move along the guide rail.
As an implementation of the present invention, the telescopic assembly includes:
the mounting plate is arranged between the mounting plate and the second plate, and a plurality of support rods are arranged between the mounting plate and the second plate; the lower surface of the mounting plate is provided with a lower guide rail and an upper sliding block; the upper sliding block can slide along the lower guide rail;
the telescopic frame comprises a first rod and a second rod, one ends of the first rod and the second rod are connected with the second plate through hinges, a shaft rod is connected between the other ends of the first rod and the second rod, and a sliding block is arranged on the shaft rod;
and the third stepping motor is connected with a third screw rod and is used for driving the sliding block to move along the third screw rod.
As an implementation manner of the present invention, the telescopic assembly further includes:
the third rod and the fourth rod are connected with the mounting plate and the second plate through hinges, the third rod is connected with the first rod through a rotating shaft, and the fourth rod is connected with the second rod through a rotating shaft.
In a second aspect, embodiments of the present invention provide a bogie bottom inspection platform, comprising:
and the driving device is arranged on the inner side of the track and is used for shooting an image of the bottom of the bogie.
As an implementation manner of the present invention, the detection platform further includes:
the cleaning part is used for cleaning the bottom of a bogie of a railway vehicle driving into the detection platform.
As an embodiment of the present invention, the cleaning unit includes:
the spray heads are driven by the motor to swing;
the floor drain is configured around the spray head and used for sewage to flow out;
and the water baffle is arranged between the floor drain and the driving device.
As an implementation of the present invention, a fan is disposed between the water baffle and the driving device.
As an implementation manner of the present invention, the detection platform further includes:
and the light supplement lamp assembly comprises a light supplement lamp arranged on the periphery of the driving device and a light supplement lamp arranged on the first driving assembly.
According to the driving device and the detection platform of the bogie bottom detection platform provided by the embodiment of the invention, the driving device drives the second driving assembly to lift or drop pressure along a path close to or far away from the bottom of the bogie through the telescopic assembly; the first driving assembly and the second driving assembly are connected through the third driving assembly, the third driving assembly drives the first driving assembly to rotate relative to the second driving assembly, the camera is driven to transversely move through the first driving assembly, and the first driving assembly is driven to longitudinally move through the second driving assembly; the driving device has compact structure and higher automation, accurately obtains images and greatly improves the detection efficiency. On the other hand, the detection platform can realize full automation of detection aiming at the bottom of the bogie. Because the driving device has a plurality of degrees of freedom, the position required by the optimal image acquisition at the bottom of the bogie can be reached, and the reliability of the image acquisition and the accuracy of the detection result are ensured.
It should be understood that the statements herein reciting aspects are not intended to limit the critical or essential features of any embodiment of the invention, nor are they intended to limit the scope of the invention. Other features of the present invention will become apparent from the following description.
Drawings
In order to more clearly illustrate one or more embodiments or prior art solutions of the present specification, the drawings that are needed in the description of the embodiments or prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present specification, and that other drawings can be obtained by those skilled in the art without inventive exercise.
FIG. 1 is a schematic structural diagram of a driving device of a bogie bottom inspection platform according to an embodiment of the invention;
FIG. 2 is a schematic structural diagram illustrating the working state of a driving device of the bogie bottom detection platform according to the embodiment of the invention;
FIG. 3 is a schematic structural diagram illustrating a non-operating state of a driving device of the bogie bottom detection platform according to the embodiment of the invention;
FIG. 4 is a schematic diagram of a first driving assembly of a driving device of a bogie bottom detection platform according to an embodiment of the invention;
FIG. 5 is a schematic diagram of a second driving assembly of the driving device of the bogie bottom detection platform according to the embodiment of the invention;
FIG. 6 is a schematic diagram illustrating a telescopic assembly of a driving device of a bogie bottom detection platform according to an embodiment of the invention;
FIG. 7 is a schematic structural diagram of an inspection platform according to an embodiment of the present invention;
FIG. 8 illustrates a flow chart of a method of bogie bottom detection of an embodiment of the present invention;
FIG. 9 illustrates a flow chart of a method of bogie bottom detection in accordance with another embodiment of the present invention;
FIG. 10 is a schematic diagram illustrating the structure of a DCGAN network model according to an embodiment of the present invention;
FIG. 11 is a flow chart illustrating a DCGAN data generation process according to an embodiment of the present invention;
FIG. 12 shows a schematic structural diagram of the CBAM-inserted YOLO-V4 model according to an embodiment of the present invention;
fig. 13 shows a CBAM structure diagram of an embodiment of the present invention.
In the figures, a first drive assembly 10; a second drive assembly 20; a telescoping assembly 40; a third drive assembly 30; driving the camera 60; a first plate 101; a guide rod 1011; a camera mount 1012; a first stepper motor 102; a first stepper motor 102; a first lead screw 103; a second plate 201; a guide rail 2011; a connecting seat 2012; a second stepping motor 202; a second lead screw 203; a mounting plate 401; a support rod 4011; a lower rail 406; an upper slider 407; an expansion bracket 402; a first rod 4021; a second rod 4022; a lower slide 4023; a third stepper motor 403; a third lead screw 404; a third rod 4025; a fourth bar 4024; a dust curtain 400; a cleaning section 50; a spray head 501; a floor drain 502; a water guard 503; a fan 504; a square fill light 5052; the long fill light 5053.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in one or more embodiments of the present disclosure, the technical solutions in one or more embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in one or more embodiments of the present disclosure, and it is obvious that the described embodiments are only a part of the embodiments of the present disclosure, and not all embodiments. All other embodiments that can be derived by a person skilled in the art from one or more of the embodiments described herein without making any inventive step shall fall within the scope of protection of this document.
It should be noted that, the description of the embodiment of the present invention is only for clearly illustrating the technical solutions of the embodiment of the present invention, and does not limit the technical solutions provided by the embodiment of the present invention.
Fig. 1 shows a schematic structural diagram of a driving device of a bogie bottom detection platform according to an embodiment of the invention.
Referring to fig. 1, the driving apparatus includes:
a first driving assembly 10 for driving the camera 60 to move transversely;
a second drive assembly 20 for driving the first drive assembly 20 in longitudinal movement;
a telescoping assembly 40 for driving the second drive assembly 20 to lift or lower along a path toward or away from the bottom of the truck;
and the third driving assembly 30, the first driving assembly 10 and the second driving assembly 20 are connected through the third driving assembly 30, the third driving assembly 30 drives the first driving assembly 10 to rotate relative to the second driving assembly 20, and a plane formed by the rotation of the first driving assembly 10 is parallel to a plane where the bottom of the bogie is located.
The embodiment of the invention provides a driving device for a bogie bottom detection platform, which drives a second driving assembly 20 to lift or lower along a path close to or far from the bottom of a bogie through a telescopic assembly 40; the first driving assembly 10 and the second driving assembly 20 are connected through the third driving assembly 30, the third driving assembly 30 drives the first driving assembly 10 to rotate relative to the second driving assembly 20, the camera is driven to move transversely through the first driving assembly 10, and the first driving assembly is driven to move longitudinally through the second driving assembly 20; the driving device has compact structure and higher automation, and can accurately obtain images, thereby greatly improving the detection efficiency.
FIG. 2 is a schematic structural diagram illustrating the working state of a driving device of the bogie bottom detection platform according to the embodiment of the invention; FIG. 3 is a schematic structural diagram illustrating a non-operating state of a driving device of the bogie bottom detection platform according to the embodiment of the invention; as shown in fig. 2-3, first, a third stepping motor 403 is fixedly mounted on the mounting plate 401, the third stepping motor 403 is started, the third stepping motor 403 is driven by a coupler to rotate a third lead screw 404, the third lead screw 404 drives a lower slider 4023 to move from right to left, and a lower guide rail 406 and an upper slider 407 are configured on the lower surface of the mounting plate 401; the upper slide block 407 can slide along the lower guide rail 406; when the lower sliding block 4023 moves from right to left, the upper sliding block 407 moves in the opposite direction to drive the second plate 201 to lift, when the supporting angle of the supporting rod 403 reaches a preset angle, the stepping motor 403 stops rotating, after the second plate 201 is lifted, the fourth stepping motor 406 starts rotating to drive the rotating seat 405 to rotate, the top of the rotating seat 405 is provided with a top bracket, and when the top bracket rotates to the preset angle, the stepping motor 3 stops rotating; the top bracket can be provided with a square light supplement lamp 5052, a long light supplement lamp 5053 and the like. Mounting plate 401 may be configured with mounting holes 4011, and mounting plate 401 may be secured to the ground through mounting holes 4011. The camera 60 is fixedly arranged on the camera mounting seat, at the moment, the first stepping motor and the second stepping motor start to rotate to respectively drive the first driving assembly 10, and the first driving assembly 10 drives the camera 60 to transversely move; the second driving component 20 drives the first driving component 10 to move longitudinally, so as to control the front and back movement of the camera, and thus, the camera can shoot the bottom of the whole bogie.
FIG. 4 is a schematic diagram of a first driving assembly of a driving device of a bogie bottom detection platform according to an embodiment of the invention; the first drive assembly 10 includes:
a first plate 101, the first plate 101 being configured with a pair of guide rods 1011, the pair of guide rods 1011 being configured with a camera mount 1012;
a first stepping motor 102, wherein a first screw 103 is connected to the first stepping motor 102, the first screw 103 is connected to the camera mounting seat 1012, and the first stepping motor drives the first screw 103 to move the camera mounting seat 1012 laterally along the pair of guide rods 1011, so that the camera 60 moves laterally.
FIG. 5 is a schematic diagram of a second driving assembly of the driving device of the bogie bottom detection platform according to the embodiment of the invention; as shown in fig. 5, the second driving assembly 20 includes:
a second plate 201, wherein the second plate 201 is provided with a guide rail 2011, the guide rail 2011 is provided with a connecting seat 2012, and the connecting seat 2012 is connected with the first driving assembly 10;
a second stepping motor 202, the second stepping motor 202 is connected with a second lead screw 203, the second lead screw 203 is connected with the connecting seat 2012, and the second stepping motor 202 drives the second lead screw 203 to make the connecting seat 2012 move longitudinally along the guide rail 2011, so as to make the camera 60 move longitudinally.
FIG. 6 is a schematic diagram illustrating a telescopic assembly of a driving device of a bogie bottom detection platform according to an embodiment of the invention; as shown in fig. 6, the telescopic assembly 40 includes:
a mounting plate 401, wherein a plurality of support rods 4011 are arranged between the mounting plate 401 and the second plate 201, and two ends of the support rods 4011 of the mounting plate 401 are connected with the mounting plate 401 and the second plate 201 through hinges; a lower guide rail 406 and an upper slide block 407 are arranged on the lower surface of the mounting plate 401; the upper slide block 407 can slide along the lower guide rail 406;
the telescopic frame 402 comprises a first rod 4021 and a second rod 4022, and one end of the first rod 4021 and one end of the second rod 4022 are connected with the lower guide rail 406 through an upper sliding block 407;
a shaft rod is connected between the other ends of the first rod 4021 and the second rod 4022, and a lower sliding block 4023 is arranged on the shaft rod;
and a third stepping motor 403, where the third stepping motor 403 is connected to a third lead screw 404, and the third stepping motor 403 is used to drive the slider 4023 to move along the third lead screw 404.
It should be noted that the telescopic assembly 40 may further include:
Based on the same inventive concept, an embodiment of the present invention further provides a bogie bottom detection platform, fig. 7 shows a schematic structural diagram of the detection platform according to the embodiment of the present invention, and the detection platform shown in fig. 7 includes:
the driving device is configured on the inner side of the track and used for shooting an image of the bottom of the bogie; wherein the driving device is any one of the driving devices mentioned in the above embodiments. In some embodiments, the drive may be configured with a dust curtain 400.
In some embodiments, the detection platform further comprises:
a washing part 50 for washing the bottom of the bogie of the rail vehicle entering the inspection platform.
Specifically, the cleaning portion includes:
the sprayer comprises a plurality of sprayers 501, wherein the sprayers 501 are driven by a motor to swing;
a floor drain 502, which is arranged around the spray head 501 and is used for sewage to flow out;
and the water baffle 503, wherein the water baffle 503 is arranged between the floor drain 502 and the driving device 100.
In an embodiment of the present invention, a fan 504 is disposed between the water guard 503 and the driving device 100.
In some embodiments, the detection platform further comprises:
and a light supplement lamp assembly 505 including a light supplement lamp 5051 disposed at an outer periphery of the driving apparatus and a light supplement lamp 5052 disposed around the first driving assembly.
The beneficial effects of the invention are explained by a preferred embodiment, and the embodiment provides a bogie bottom detection platform, a rail vehicle firstly slowly drives into the detection platform, when the rail vehicle reaches the cleaning part 50, the adjustable spray nozzle 501 sprays water to clean the bottom of the bogie, meanwhile, the spray nozzle motor drives the adjustable spray nozzle 501 to swing left and right, the cleaning range is expanded, sewage flows out along the floor drain 502, and in order to prevent the sewage from air drying and polluting the detection area, a water baffle 503 is arranged on the left of the floor drain 502. After the cleaning is finished, the vehicle continues to run, the air drying area is reached, the fan 504 is opened, the air drying treatment is carried out on the bottom of the steering frame, preparation is carried out for subsequent detection, the vehicle reaches the detection area after the air drying is finished, the dustproof curtain 400 is opened from right to left at the moment, the driving device is started at the moment, the mounting hole 4011 can be configured on the mounting plate 401, and the mounting plate 401 is fixed with the ground through the mounting hole 4011. Firstly, a third stepping motor 403 is fixedly installed on an installation plate 401, the third stepping motor 403 is started, the third stepping motor 403 is driven by a coupler to rotate a third screw rod 404, the third screw rod 404 drives a lower sliding block 4023 to move from right to left, and a lower guide rail 406 and an upper sliding block 407 are configured on the lower surface of the installation plate 401; the upper slide block 407 can slide along the lower guide rail 406; when the lower slider 4023 moves from right to left, the upper slider 407 moves in the opposite direction to drive the second plate 201 to lift, when the supporting angle of the supporting rod 403 reaches a preset angle, the stepping motor 403 stops rotating, after the second plate 201 is lifted, the fourth stepping motor 406 starts rotating to drive the rotating seat 405 to rotate, the top support 104 is installed at the top of the rotating seat 405, and when the top support 104 rotates to the preset angle, the stepping motor 3 stops rotating; the top bracket can be provided with a square light supplement lamp 5052, a long light supplement lamp 5053 and the like.
The camera 60 is fixedly arranged on the camera mounting seat, at the moment, the first stepping motor and the second stepping motor start to rotate to respectively drive the first driving assembly 10, and the first driving assembly 10 drives the camera 60 to transversely move; the second driving component 20 drives the first driving component 10 to move longitudinally, so as to control the front and back movement of the camera, and thus, the camera can shoot the bottom of the whole bogie.
Further, performing data amplification on the collected images by using a DCGAN algorithm, then performing uniform size processing on the data, labeling fault areas in the images by using labelme software, taking the data as a training data set, wherein the data set comprises common fault data categories and position information, main faults are loss of bolts and rivets at the bottom of the bogie, abnormal falling of iron wires of check bolts and check marks and the like, and then inputting the training data set into an improved YOLO-v4 model. When detection is needed, an image to be detected is input into a trained detection model, and the detection model feeds back and gives an alarm for fault types and fault position information.
FIG. 8 illustrates a flow chart of a method of bogie bottom detection of an embodiment of the present invention;
as shown in fig. 8, the detection method includes:
s20, acquiring a bottom image of the bogie of the railway vehicle to be detected;
and S40, inputting the bogie bottom image into a pre-trained YOLO-v4 model, and feeding back and alarming fault type and fault position information through the YOLO-v4 model.
According to the bogie bottom detection method provided by the embodiment, the bottom image of the bogie of the railway vehicle to be detected is obtained; and inputting the bogie bottom image into a pre-trained YOLO-v4 model, feeding back and alarming fault type and fault position information through the YOLO-v4 model, and when detection is required, inputting an image to be detected into the trained YOLO-v4 model, and automatically feeding back and alarming the fault type and fault position information.
In some embodiments, the method of training the YOLO-v4 model comprises:
acquiring a fault data image at the bottom of a railway vehicle bogie and preprocessing the fault data image to obtain a training data set;
inputting the training data set into a YOLO-v4 model, wherein a CBAM mechanism is added to the YOLO-v4 model path aggregation network.
FIG. 9 illustrates a flow chart of a method of bogie bottom detection in accordance with another embodiment of the present invention; the YOLO-v4 model utilizes a main network (CSPDarknet53) to extract visual features of a data set, the extracted features are subjected to feature aggregation through a path aggregation network (PANet), the aggregated features are input into a detection head, the fault type and the fault position are predicted, and a loss function is utilized to train the YOLO-v4 model.
Specifically, the step of obtaining a bottom image of a bogie of the rail vehicle and preprocessing the bottom image to obtain a training data set comprises the following steps:
acquiring bottom image data of a bogie of the railway vehicle by using detection equipment;
and performing data amplification on the bogie bottom image by using a DCGAN network, then performing uniform size processing on the data, and labeling a fault area in the image by using labelme software to obtain a training data set. For the training of DCGAN, each category adopts the corresponding fault data image as a training set. According to the generation countermeasure theory, D and G can be optimized continuously and reasonably, and finally the quality of the generated samples can reach Nash equilibrium along with the increase of the number of iterations. For example, during the training process, the learning rate is set to 0.002, the network is stopped after running 1000 epochs, 900 pieces of data are generated by using DCGAN, 100 pieces of original data are added, the total data set is set to 1000 pieces, 800 pieces of data are randomly divided as a training set, 100 pieces of data are used as a test set, and 100 pieces of data are used as a verification set.
Fig. 10 is a schematic structural diagram of a DCGAN network model according to an embodiment of the present invention, and fig. 11 is a flowchart of a DCGAN data generation process according to an embodiment of the present invention; as shown in fig. 10-11, the DCGAN network includes:
firstly, Gaussian white noise is input into the generating network G, and is input into the generating network to obtain noise data G (Z). And the discrimination network compares the obtained G (Z) with the real data, makes true and false judgment and feeds back the true and false judgment to the generation network, the discrimination network optimization process is similar to a binary problem, and finally the Nash balance is achieved through the mutual competition of the generation network and the discrimination network. At this time, the data generated in the generated network is extremely similar to the real data, and can be used as the test data. The whole process can be generalized to a binary infinitesimal maximum game, and the objective function can be defined as the following formula:
v (D, G) is the cross-entropy loss commonly used in the binary task, Pdata(x)For true data distribution, pg(Z)For noise distribution, d (x) represents that x is derived from real data, and g (z) represents a generated sample of random noise after passing through a generated network. During the training process, the discrimination network D is aimed at maximizing the discrimination accuracy, and the network G is generated with the aim of minimizing the maximum value of the discrimination accuracy of the discrimination network D, i.e. the maximum value of the discrimination accuracy of the discrimination network D is
FIG. 12 shows a schematic structural diagram of the CBAM-inserted YOLO-V4 model according to an embodiment of the present invention;
FIG. 13 shows a CBAM architectural diagram of an embodiment of the invention; as shown in fig. 12 to 13, the CBAM includes: a channel attention module to pay attention to given arbitrary intermediate feature map F ∈ RC×H×WCompressing the feature map in spatial dimension by maximum pooling and average pooling to obtain
And
two spatial background descriptions sharing one MLP network and outputting two characteristics of MLP of three-layer perceptronAdding graphs, normalizing by using Sigmoid function, and obtaining a final channel attention feature graph MC∈RC×1×1；
A spatial attention module, which focuses mainly on position information of a target on an image, and obtains a channel background description using maximum pooling and average pooling, respectively, in a channel dimension
And
then, using channel splicing, connecting two channel feature maps in series, and using 7 × 7 convolution operation to generate spatial attention feature map MS∈R1×H×W。
In particular, the channel attention module gives an arbitrary intermediate feature map F ∈ RC×H×WCompressing the feature map in spatial dimension by maximum pooling and average pooling to obtain
And
two spatial background descriptions share one MLP network, two feature graphs output by the MLP of the three-layer perceptron are added, a Sigmoid function is used for normalization, and the final channel attention feature graph is MC∈RC×1×1. Unlike channel attention, the spatial attention module focuses mainly on the position information of the target on the image, and first obtains the channel background description using maximum pooling and average pooling in the channel dimension
And
then, using channel splicing, two channel feature maps are concatenated, and a 7 × 7 convolution operation is used to generate spatial attention featuresSign graph MS∈R1×H×W。
The mechanism module function process is as follows:
wherein F ∈ RC×H×WIs given an intermediate feature level mapping, sign
Representing bit-wise multiplication, MCA feature diagram of the attention module at the channel level is shown. F' is the feature output after the channel attention mechanism, and MSIs the attention value of the broadcast channel in the spatial dimension. F "is the final refined output. So at the level of the channel: inputting a feature map F' according to the result of maximum Pooling operation (Max Pooling) and Global average Pooling operation (Global Avg Pooling) of each channel, respectively sending the result into a three-layer perceptron MLP, adding the output results, and sending the result into a ReLU activation function to obtain a feature map M of a channel attention moduleC. Obtaining an attention feature M at a channel levelCThe process is as follows:
wherein
And
the method is characterized in that two different spatial context descriptors are used for representing average pooling characteristics and maximum pooling characteristics, aggregating spatial information of a characteristic diagram and generating two spatial context mappings with different attributes, and the two characteristics are utilized in a formula instead of being used independently, so that the characteristic extraction capability of a network on a track fastener can be greatly improved. W0And W1Are parameters of the perceptual layer MLP.
After the channel attention module is executed, the feature map of the intermediate result is sent to the spatial attention module for continuous processing, and a final feature map of the spatial attention mechanism is obtained. The feature graph F' processed by the channel feature graph mechanism firstly calculates two feature graphs of global maximum pooling and global average pooling on the channel dimension, multiplies the two feature graphs by alignment, then uses a traditional convolution process once, and finally uses the ReLU activation to obtain the feature graph M used by the space attention moduleS. Obtaining attention M at the level of the channelSThe process is as follows:
wherein
And
the same holds for the x-average pooling level feature and the maximum pooling level feature in the spatial channel dimension, respectively. By connecting them, an effective feature descriptor can be generated to extract the effective salient features at the channel level. Wherein f is7×7Representing a convolution operation with a convolution kernel size of 7 x 7. Because CBAM is a lightweight module, and is integrated into the CNN network for end-to-end training, the overhead can be ignored. In the YOLO-V4 model map inserted into the CBAM, the size of the input image is 416 × 3, and taking the first CBAM module as an example, the size of the input feature map is 52 × 256, and after the global maximum pool and the global average pool, the sizes 1 × 256 and 1 × 256 of the two feature maps are obtained, and then the two feature maps are passed through a multi-layer perceptron and shared weight. The dimension is reduced to 1 x 16 and the dimensionality is reduced to 16, then the dimension is increased to 1 x 256, the operation is added to the two feature maps, a channel attention function with the size of 1 x 256 is obtained through Sigmoid activation, the size of 1 x 256 of the channel attention feature map is obtained, the input feature map is multiplied with the attention map, and a feature map output with the size of 52 x 52 256 is obtained. Then respectively pass through all based on channelsThe local max pool and the global average pool enter the spatial attention module. The two signatures 52 x 1 size and the number of two signatures of a channel are combined to obtain the size 52 x 2 signature, and then 7 x 7 convolution is used to reduce the number of channels to 1. And finally, obtaining a space attention diagram with the size of 52 x 1 by using the Sigmoid activation function, and multiplying the input of the space attention module by the space attention diagram to obtain an output feature diagram with the size of 52 x 256. The output feature map of the CBAM module is consistent with the input feature map. Thus, the CBAM module is successfully inserted into the network, and the sizes of the front feature map and the back feature map are not changed.
According to the bogie bottom detection method provided by the embodiment of the invention, a data enhancement model DCGAN model is used, a convolutional neural network is used for replacing a multilayer perceptron in the GAN, and a full connection layer is replaced by a global pooling layer so as to reduce the calculated amount, so that the problem of insufficient sample quantity can be quickly solved. Then, considering that the neural network has too high dependence on memory consumption, a model based on a YOLO-v4 algorithm is established, a CBAM mechanism is added in a backbone network in order to improve the model profit, useless information is effectively inhibited, the attention of the useful information is improved, meanwhile, the CBAM is a lightweight module, the performance of the added model is improved to a certain extent, and the added calculation amount is small. The lightweight YOLO-v4 model based on data enhancement has the main advantages that the precision is guaranteed, meanwhile, the characteristics of the picture can be effectively extracted, and the memory consumption is remarkably reduced.
The foregoing description is only exemplary of the preferred embodiments of the invention and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the invention herein disclosed is not limited to the particular combination of features described above, but also encompasses other arrangements formed by any combination of the above features or their equivalents without departing from the spirit of the invention. For example, the above features and (but not limited to) features having similar functions disclosed in the present invention are mutually replaced to form the technical solution.
Claims (10)
1. A drive arrangement for a truck bed inspection platform, the drive arrangement comprising:
the first driving component is used for driving the camera to move transversely;
a second drive assembly for driving the first drive assembly in longitudinal movement;
a telescoping assembly for driving the second drive assembly to lift or lower along a path closer to or further from the bottom of the bogie;
and the first driving assembly and the second driving assembly are connected through the third driving assembly, the third driving assembly drives the first driving assembly to rotate relative to the second driving assembly, and a plane formed by the rotation of the first driving assembly is parallel to a plane where the bottom of the bogie is located.
2. The drive of claim 1, wherein the first drive assembly comprises:
a first plate configured with a pair of guide rods configured with a camera mount;
the first stepping motor is connected with a first screw rod, the first screw rod is connected with the camera mounting seat, and the first stepping motor drives the first screw rod to enable the camera mounting seat to transversely move along the pair of guide rods.
3. The drive of claim 1, wherein the second drive assembly comprises:
the second plate is provided with a guide rail, the guide rail is provided with a connecting seat, and the connecting seat is connected with the first driving assembly;
the second stepping motor is connected with a second screw rod, the second screw rod is connected with the connecting seat, and the second stepping motor drives the second screw rod to enable the connecting seat to move longitudinally along the guide rail.
4. The drive of claim 3, wherein the telescoping assembly comprises:
the mounting plate is arranged between the mounting plate and the second plate, and a plurality of support rods are arranged between the mounting plate and the second plate; the lower surface of the mounting plate is provided with a lower guide rail and an upper sliding block, and the upper sliding block can slide along the lower guide rail;
the telescopic frame comprises a first rod and a second rod, one ends of the first rod and the second rod are connected with the second plate through hinges, a shaft rod is connected between the other ends of the first rod and the second rod, and a sliding block is arranged on the shaft rod;
and the third stepping motor is connected with a third screw rod and is used for driving the sliding block to move along the third screw rod.
5. The drive of claim 4, wherein the retraction assembly further comprises:
the third rod and the fourth rod are connected with the mounting plate and the second plate through hinges, the third rod is connected with the first rod through a rotating shaft, and the fourth rod is connected with the second rod through a rotating shaft.
6. A bogie bottom inspection platform based on the drive arrangement of any one of claims 1 to 5, wherein the inspection platform comprises:
and the driving device is arranged on the inner side of the track and is used for shooting an image of the bottom of the bogie.
7. The assay platform of claim 6, further comprising:
the cleaning part is used for cleaning the bottom of a bogie of a railway vehicle driving into the detection platform.
8. The testing platform of claim 7, wherein said washing portion comprises:
the spray heads are driven by the motor to swing;
the floor drain is configured around the spray head and used for sewage to flow out;
and the water baffle is arranged between the floor drain and the driving device.
9. The testing platform of claim 8, wherein a fan is disposed between said water deflector and said driving device.
10. The assay platform of any one of claims 6-9, further comprising:
and the light supplement lamp assembly comprises a light supplement lamp arranged on the periphery of the driving device and a light supplement lamp arranged on the first driving assembly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111479897.7A CN114323709A (en) | 2021-12-06 | 2021-12-06 | Driving device of bogie bottom detection platform and detection platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111479897.7A CN114323709A (en) | 2021-12-06 | 2021-12-06 | Driving device of bogie bottom detection platform and detection platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114323709A true CN114323709A (en) | 2022-04-12 |
Family
ID=81048219
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111479897.7A Pending CN114323709A (en) | 2021-12-06 | 2021-12-06 | Driving device of bogie bottom detection platform and detection platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114323709A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102007024066A1 (en) * | 2007-05-22 | 2008-11-27 | Knorr-Bremse Systeme für Schienenfahrzeuge GmbH | Device for fault monitoring of chassis components of rail vehicles |
| WO2014082364A1 (en) * | 2012-11-27 | 2014-06-05 | 长春轨道客车股份有限公司 | Rotation feature test system of bogie of rail vehicle |
| CN107200038A (en) * | 2017-05-17 | 2017-09-26 | 中国铁道科学研究院铁道科学技术研究发展中心 | A kind of repairing device for vehicle and repair method |
| CN109834693A (en) * | 2019-01-04 | 2019-06-04 | 国网天津市电力公司 | A kind of crusing robot system and its control method |
| CN110762336A (en) * | 2019-09-23 | 2020-02-07 | 北京建筑大学 | Intelligent pipeline inspection system |
| CN211234992U (en) * | 2019-12-27 | 2020-08-11 | 北京天亿时代科技有限公司 | Cleaning equipment for wheel set flaw detection machine |
| JP2020153745A (en) * | 2019-03-19 | 2020-09-24 | Ntn株式会社 | Abnormality detector |
| CN212056386U (en) * | 2020-01-06 | 2020-12-01 | 广州汽车集团股份有限公司 | Image pickup apparatus and intelligent image pickup system |
| CN112202114A (en) * | 2020-10-09 | 2021-01-08 | 王长南 | Wiring device with adjusting function for construction |
| CN112719586A (en) * | 2020-12-18 | 2021-04-30 | 柳州工学院 | Laser processing platform of automobile parts |
| CN214202620U (en) * | 2021-03-08 | 2021-09-14 | 太原城市职业技术学院 | Device capable of detecting whether automobile chassis has object or not |
-
2021
- 2021-12-06 CN CN202111479897.7A patent/CN114323709A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102007024066A1 (en) * | 2007-05-22 | 2008-11-27 | Knorr-Bremse Systeme für Schienenfahrzeuge GmbH | Device for fault monitoring of chassis components of rail vehicles |
| WO2014082364A1 (en) * | 2012-11-27 | 2014-06-05 | 长春轨道客车股份有限公司 | Rotation feature test system of bogie of rail vehicle |
| CN107200038A (en) * | 2017-05-17 | 2017-09-26 | 中国铁道科学研究院铁道科学技术研究发展中心 | A kind of repairing device for vehicle and repair method |
| CN109834693A (en) * | 2019-01-04 | 2019-06-04 | 国网天津市电力公司 | A kind of crusing robot system and its control method |
| JP2020153745A (en) * | 2019-03-19 | 2020-09-24 | Ntn株式会社 | Abnormality detector |
| CN110762336A (en) * | 2019-09-23 | 2020-02-07 | 北京建筑大学 | Intelligent pipeline inspection system |
| CN211234992U (en) * | 2019-12-27 | 2020-08-11 | 北京天亿时代科技有限公司 | Cleaning equipment for wheel set flaw detection machine |
| CN212056386U (en) * | 2020-01-06 | 2020-12-01 | 广州汽车集团股份有限公司 | Image pickup apparatus and intelligent image pickup system |
| CN112202114A (en) * | 2020-10-09 | 2021-01-08 | 王长南 | Wiring device with adjusting function for construction |
| CN112719586A (en) * | 2020-12-18 | 2021-04-30 | 柳州工学院 | Laser processing platform of automobile parts |
| CN214202620U (en) * | 2021-03-08 | 2021-09-14 | 太原城市职业技术学院 | Device capable of detecting whether automobile chassis has object or not |
Non-Patent Citations (9)
| Title |
|---|
| 尚远航等: "基于混合注意力机制的植物病害识别", 塔里木大学学报, vol. 33, no. 2, pages 132 - 103 * |
| 朱建国;: "横向滚珠丝杠驱动升降平台设计要点分析", 现代制造技术与装备, no. 06, pages 110 - 113 * |
| 棉花糖: "【论文笔记】CBAM：Convolutional Block Attention Module（卷积块注意力模型）", pages 2 - 3, Retrieved from the Internet <URL:https://blog.csdn.net/zuzhiang/article/details/108292973> * |
| 王立舒等: "基于改进YOLOv4-Tiny 的蓝莓成熟度识别方法", 农业工程学报, vol. 37, no. 18, pages 170 - 178 * |
| 王雨晨等: "基于改进YOLOv4 的安全帽佩戴检测算法", 现代信息科技, vol. 5, no. 22, pages 156 - 164 * |
| 罗迪等: "一种基于生成对抗网络与注意力机制的可见光和红外图像融合方法", 红外技术, vol. 43, no. 6, pages 566 - 574 * |
| 赵晨洁等: "基于注意力机制的病毒软件可视化检测方法", 软件工程, vol. 24, no. 6, pages 6 - 12 * |
| 赵玉蓉等: "融合混合域注意力的 YOLOv4 在船舶检测中的应用", 计算机与现代化, pages 75 - 82 * |
| 赵玉蓉等: "融合混合域注意力的YOLOv4在船舶检测中的应用", 计算机与现代化, no. 9, pages 75 - 81 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106919978B (en) | Method for identifying and detecting parts of high-speed rail contact net supporting device | |
| CN106767518A (en) | A kind of phone housing quality on-line detecting device and method based on line laser structured light | |
| CN107144569A (en) | The fan blade surface defect diagnostic method split based on selective search | |
| CN106504238A (en) | Railway contact line defect inspection method based on image procossing and convolutional neural networks | |
| CN206563560U (en) | A kind of phone housing quality on-line detecting device based on line laser structured light | |
| KR20210157000A (en) | System and method for underbody inspection of vehicle | |
| CN105044122A (en) | Copper part surface defect visual inspection system and inspection method based on semi-supervised learning model | |
| CN114623049B (en) | Wind turbine tower clearance monitoring method and computer program product | |
| CN111079822A (en) | Method for identifying dislocation fault image of middle rubber and upper and lower plates of axle box rubber pad | |
| CN109131895A (en) | Unmanned plane carry device, unmanned plane and the method for detection are automatically tracked for blade | |
| CN112455676A (en) | Intelligent monitoring and analyzing system and method for health state of photovoltaic panel | |
| CN111460997B (en) | Face recognition system and method based on omnibearing dynamic capture | |
| CN112129778A (en) | Rail train maintenance detection system based on computer vision | |
| CN114323709A (en) | Driving device of bogie bottom detection platform and detection platform | |
| CN116309313A (en) | Battery surface welding defect detection method | |
| CN107505037A (en) | A kind of supervising device that can finely identify vehicle | |
| CN114627074A (en) | Ground shooting fan blade real-time monitoring method based on deep learning | |
| CN113376707B (en) | Visual detection system and method for vehicle frame | |
| CN114266935A (en) | Bogie bottom detection method and device, electronic device and storage medium | |
| CN107633527A (en) | Target tracking method and device based on full convolutional neural networks | |
| CN109934172B (en) | GPS-free full-operation line fault visual detection and positioning method for high-speed train pantograph | |
| CN117145581A (en) | Intelligent identification-tracking accurate dust falling system and method for dust source in underground operation space | |
| CN116388669A (en) | Photovoltaic panel foreign matter detection and cleaning method based on Swin transducer | |
| CN114320709A (en) | Deep learning-based classification detection method for oil leakage inside power station generator | |
| CN112989883B (en) | Method for identifying obstacle in front of train |
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 |