CN109760635A - A kind of line traffic control windshield wiper control system based on GAN network - Google Patents
A kind of line traffic control windshield wiper control system based on GAN network Download PDFInfo
- Publication number
- CN109760635A CN109760635A CN201910017056.0A CN201910017056A CN109760635A CN 109760635 A CN109760635 A CN 109760635A CN 201910017056 A CN201910017056 A CN 201910017056A CN 109760635 A CN109760635 A CN 109760635A
- Authority
- CN
- China
- Prior art keywords
- image
- network
- controller
- rain
- gan
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of line traffic control windshield wiper control systems based on GAN network, to realize that the rain brush of intelligent automobile controls, the system includes sequentially connected forward sight camera, intelligence drives controller, entire car controller, windshield-wiper controller, driving motor assembly, retarder and windshield wiper, forward sight camera acquires road ahead image in real time, and image is transferred to intelligence by USB data line and drives controller, intelligence, which is driven, is equipped with the GAN network architecture inside controller, the raindrop in image to remove rainy day acquisition, and pass through image real-time detection vehicle-surroundings rainfall size, and the adjustment signal of rain brush gear is transmitted to entire car controller by CAN line, after entire car controller acquisition intelligence drives the CAN signal of controller, the working condition of adjustment windshield-wiper controller in real time, windshield-wiper controller controls the work of driving motor assembly by pulse signal, Raindrop on the outside of rain brush removal front windshield are controlled, compared with prior art, the present invention has many advantages, such as that practical, development cost is low, applicability is high.
Description
Technical field
The present invention relates to new-energy automobile fields, more particularly, to a kind of line traffic control windshield wiper control system based on GAN network.
Background technique
With drive-by-wire, the stable development of steering-by-wire and brake-by-wire technical aspect, intelligent automobile is had been able to
It is realized under special scenes unmanned.In-vehicle camera is one of main sensors of intelligent automobile, considers waterproof requirement, usually pacifies
Mounted in cockpit.The raindrop easy to form on the outside of rainy weather, front windshield, block the sight of camera.It is unmanned
Technology does not consider that the rainy day travels temporarily, so that integrated visual pattern Processing Algorithm is not had image rain removing technology, leads to intelligent vapour
Vehicle can not usually overcome the influence of raindrop.In addition, current intelligent automobile by common vehicle reequip come, Wiper system after
The manually opened mode of system is handed down to posterity in continuation of insurance, directly can not be driven controller by intelligence and be controlled, and is brought to the development of unmanned technology
It hinders.
Meanwhile application of the deep learning in terms of the environment sensing of intelligent automobile gradually increases, environment perception technology is also
Intelligent automobile core technology, occupys an important position.Image restoration technology is constantly subjected to the attention of various countries researcher, mainly
The problem of be easy to causeing image quality decrease due to outdoor bad weather (such as rain, mist, snow, haze), influences the visual effect of image,
Interfere the extraction of image feature information.Image rain removing technology gradually attracts attention in intelligent driving field, is primarily due to forward sight
Interference of the camera acquired image by rainwater, obtained image is relatively fuzzyyer, is unable to get accurate image perception letter
Breath, cannot achieve intelligent automobile usually in the rainy day unmanned, major embodiment is as follows:
The first, Sometimes When It Rains, rainwater is attached on front windshield, the area information loss for blocking raindrop, and logical
Normal focal length, which can not be located in raindrop on glass, causes the region around raindrop also to become blurred, and image may be twisted,
Greatly reduce the extraction of image, semantic information.
Secondly, there is no the relevant image of rainfall in existing deep learning network training data set, therefore neural network without
Method accurately identifies and understands vehicle-periphery in these rainy days.
Visual pattern is one of main perception information source of intelligent automobile.Intelligent automobile line traffic control windshield wiper control system is ground
Study carefully with important practical application value.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide a kind of based on GAN network
Line traffic control windshield wiper control system.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of line traffic control windshield wiper control system based on GAN network, to realize that the rain brush of intelligent automobile controls, the system packet
Include sequentially connected forward sight camera, intelligence drives controller, entire car controller, windshield-wiper controller, driving motor assembly, retarder and scrapes
Rain device;
Forward sight camera acquires road ahead image in real time, and image is transferred to intelligence by USB data line and drives controller, intelligence
It drives and is equipped with the GAN network architecture inside controller, the raindrop in image to remove rainy day acquisition, and pass through image real-time detection
Vehicle-surroundings rainfall size, and the adjustment signal of rain brush gear is transmitted to entire car controller by CAN line, entire car controller obtains
After taking intelligence to drive the CAN signal of controller, the working condition of windshield-wiper controller is adjusted in real time, and windshield-wiper controller passes through pulse signal control
Driving motor assembly work processed, control rain brush remove raindrop on the outside of front windshield.
The forward sight camera is fixed on the inside of front windshield, covering model when angular field of view works without departing from rain brush
It encloses.
The intelligence drive controller be high-performance development board, inside configuration linux system, around be equipped with USB, HDMI and
Ethernet signaling interface, to transmit signal and debugging exploitation, development board can select TX2, PX2, Xavier or be based on
Nvidia chip independent development.
The intelligence drives controller and is internally integrated CAN card, and the CAN line traffic control of entire car controller is sent or received by CAN card
Signal processed.
The intelligence drives the linux system configured in controller and is equipped with image processing algorithm, which uses
The GAN network model of deep learning realizes rainfall estimation and the removal of image raindrop, rain from the feature representation of GAN network internal
Amount estimates the frequency for adjusting rain brush, and raindrop removal is repaired for removing the raindrop region in image, and by raindrop region
Multiple, image after making rain is close to true traffic environment.
The GAN network model, including generate network, judge network and the full connection layer network of rainfall estimation, wherein it is raw
Include coded portion and decoded portion at network, generates network and generate the image after removing rain in real time, by judging Network Check institute
Whether the image of generation is consistent with true environment, and the characteristic pattern that coded portion generates connects rainfall estimation entirely by a convolutional layer
Layer network is connected, to obtain the size of vehicle-surroundings rainfall.
The data set of training GAN network model when including not rainy actual traffic acquisition without rain figure as, it is artificial rainy
Rainy image, rain when actual traffic acquisition rainy image and it is rainy when actual traffic acquisition the corresponding rain of rainy image
Amount mark collection.
The practical no rain figure picture and artificial rainy image go the pre-training of rain fractional weight for network, it is described not
It rains and artificial rainy traffic image is for the weight parameter for going to rain part in pre-training GAN network, it is practical without rain and reality
The road image of border rain is to the weight parameter for going to rain part for accurate adjustment network, the rainy image of actual traffic acquisition when raining
Collect the weight parameter for training rainfall detection part with corresponding rainfall mark.
When training GAN network model, principle is first fought according to generation, while training generates network and judges network, trains
After, the fixed parameter for generating network finally trains the network parameter of full articulamentum.
The entire car controller and windshield-wiper controller is communicated using CAN communication mode, and entire car controller sends dynamic in real time
It instructs, windshield-wiper controller Real-time Feedback operating status.
The windshield-wiper controller is internally integrated motor control algorithms, and controller accurately controls rain brush drive by pulse signal
The position and speed of dynamic motor.
The driving motor assembly, as rain brush driving motor, is cooperated limit switch to use, not taken using stepper motor
Band optical encoder, structure is simple, at low cost, the rotation angle that motor is accurately positioned can be carried out by pulse number, and do not have
There is brush, reliability is higher.
The retarder uses turbine and worm slowing-down structure, and transverse profile curve is involute, to meet revolving speed
Higher and transmission precision occasion, work have no abnormal sound, last a long time.
Compared with prior art, the invention has the following advantages that
The present invention solves the problems, such as that intelligent automobile can not control rain brush, and practicability is stronger, information data convenient sources,
Without additionally increasing sensor, development cost is low, and using the generation network of GAN network architecture training, generation removes rain figure piece more
Add and meet formal road scene, the applicability of whole system can be improved.
Detailed description of the invention
Fig. 1 is line traffic control Wiper system structure chart.
Fig. 2 is image procossing GAN network structure.
Fig. 3 is that image procossing GAN network removes rain part pre-training structure chart.
Fig. 4 is that image procossing GAN network removes rain some fine training structure figure.
Fig. 5 is image procossing GAN network rainfall detection part training structure figure.
Fig. 6 is that image procossing GAN network works normally structure chart.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.
As shown in Figure 1, the present invention provides a kind of line traffic control windshield wiper control system based on deep learning GAN network, it is different from
The detection that raindrop are completed by vision camera, and intelligence can be achieved in the Wiper system of traditional rainfall inductive sensor systems, the present invention
The unlatching and working frequency of rain brush can be directly controlled by driving controller.Intelligence is driven controller and is integrated with based on the calculation of deep learning GAN network
Method can carry out rainfall detection to the image of in-vehicle camera and remove the raindrop in image, make image clearly, improve environment sensing
Effect.
First the road conditions of vehicle front are acquired by in-vehicle camera in real time, and the image of acquisition is passed through into USB transmission
It is driven to intelligence and carries out image procossing in controller.Intelligence, which is driven, is integrated with environment sensing algorithm and decision rule algorithm in controller, each to calculate
It is communicated between method by zcm.Intelligence drive controller be internally integrated it is provided by the invention based on deep learning GAN network image
Rain removing algorithm and raindrop detection algorithm.Image after removing rain is apparent from, can be with to other image, semantic information extraction algorithms, separately
On the one hand, raindrop testing result is for controlling rain brush working condition.Rain brush control signal is converted into CAN signal by CAN card
It is sent to entire car controller.Entire car controller is forwarded to bottom CAN for CAN signal is controlled.Windshield-wiper controller receives bottom in real time
Control signal in CAN is converted into the revolving speed and corner of corresponding pulse signal control driving stepper motor.Windshield-wiper controller
Integrated circuit and resistance capacitance composition oscillating circuit and delay circuit, and by the operating of relay control Wiper motor, it is used for
It realizes multistage or continuously adjustable.
Wiper drive motor uses stepper motor, uses stepper motor and substitutes direct current generator, by defeated to stepper motor
Enter corresponding driving signal, can control direction, speed and the position of stopping etc. of stepper motor rotation.The output shaft of stepper motor
Reduction gearbox is rotated and drives, so that the Wiper arm being connected with the output shaft of reduction gearbox swings, it is not necessary to realize rain
Brush arm swing with it is synchronous and increase pull rod, structure and elevating mechanism efficiency can be simplified.
Meanwhile the present invention provides the corresponding image based on deep learning and goes rain processing and the rainfall detection network architecture, uses
The GAN network architecture realizes corresponding function, unique based on having for GAN network relative to the Local Area Network of other deep learnings
Advantage.It is as follows with advantage:
The first, training dataset is easy to obtain.GAN network is by generating network and judging the confrontation study of network, adjustment
Mutual network weight parameter, the positive negative sample without opposition.So that required training dataset is only needed comprising rain and without rain
Real roads image.
The second, the image after removing rain is more nearly actual conditions.Learnt by the confrontation of GAN network, differentiates that network can
Whether distinguish is really without rain road image, so that generating network image generated more meets real road image.
The difficult point of image rain removing technology is that the region of raindrop masking is unknown, and the big portion of information of closed area background
It point has lost completely.Depth network is capable of the raindrop texture of the rainy image of learning data concentration and the image spy without rain figure picture
Raindrop in rainy image are removed, and remove the image after rain according to one pair of latent structure of no rain figure picture by sign.When forward sight camera
It is larger to detect rainfall, when can not completely remove raindrop, intelligence, which drives controller, so that windshield-wiper controller is enhanced rain brush by CAN signal
Working frequency, to realize in-vehicle camera to the wire control technology of windshield-wiper controller.
Embodiment:
As shown in Figure 1, the present invention provides a kind of line traffic control windshield wiper control systems based on GAN network.The system is in intelligence
On the basis of capable of driving a car, Wiper system is reequiped, intelligence is enable to drive system according to actual environment situation, in real time to rain
Brush is controlled.Intelligence, which is driven, is integrated with image rain removing algorithm in system, can by the raindrop of the image obtained under rainy day environment into
Row removal operation, obtains relatively clear image, the efficiency of other image processing algorithms can be improved.This method includes following step
It is rapid:
(1) in-vehicle camera acquires the road conditions of vehicle front in real time, and the image of acquisition is driven by USB transmission to intelligence
Image procossing is carried out in controller, extracts the semantic information in image.
(2) intelligence drives and is integrated with environment sensing algorithm and decision rule algorithm in controller, is carried out between each algorithm by zcm
Communication.Intelligence drives controller and is internally integrated GAN network image rain removing algorithm and rainfall detection algorithm based on deep learning.Remove rain
Relatively clear image can be used for other perception algorithms afterwards, and raindrop testing result is for controlling rain brush working condition.Rain brush control
Signal processed is converted into CAN signal by CAN card and is sent to entire car controller.
It (3) mainly include two parts: vehicle management and motion control arithmetic inside entire car controller.Drive-by-wire chassis control and
Vehicle attachment is controlled by vehicle control unit controls.Entire car controller is equipped with two-way independence CAN: upper layer CAN and bottom CAN.On
Layer CAN is mutually communicated for driving system with intelligence, and component of the bottom CAN between drive-by-wire chassis mutually communicates.It is mentioned in invention
The windshield-wiper controller access bottom CAN arrived.Rain brush controls signal and is sent to entire car controller by upper layer CAN, then by vehicle control
Device processed is forwarded to windshield-wiper controller.
(4) windshield-wiper controller receives the control signal in bottom CAN in real time, is converted into corresponding pulse signal control and drives
The revolving speed and corner of dynamic stepper motor.Motor slows down by turbine and worm, rain brush reciprocally swinging is driven, to wipe front windshield glass off
Raindrop on the outside of glass.
In step (2), the image processing algorithm in line traffic control Wiper system uses the GAN network frame based on deep learning
Frame.As shown in Fig. 2, main comprising generating network G enerator, judging network Discriminator and the full connection of rainfall estimation
Layer network.Wherein generating network includes coded portion and decoded portion.It generates network and generates the image after removing rain in real time, by sentencing
Circuit network checks whether image generated is consistent with true environment.The characteristic pattern that coded portion generates is connected entirely by rainfall estimation
Layer network is connect, the size of vehicle periphery rainfall can be estimated.GAN network implementation approach is as follows:
1-1, production are used for the data set of GAN network training, and the data set owner needed will include: actual traffic when not raining
Acquisition without rain figure picture, artificial rainy rainy image, rain when actual traffic acquisition rainy image and it is rainy when it is practical
Four parts of the corresponding rainfall mark collection of the rainy image of traffic acquisition.GAN network reaches while optimizing by generating dual training
It generates network and judges the weight parameter of network.The loss function for generating confrontation can be expressed as:
Wherein, in generating dual training, first the fixed weight parameter for generating network, is input to generation net for rainy image
In network, obtained result and the image without rain import judge in network simultaneously, and training judges that network enables the network to know enough
Do not generate without rain figure picture and really without rain figure picture;Then the weight parameter judged in network is fixed again, will judge network
Judging result carries out backpropagation, and adjustment generates the weighting parameter in network, makes to generate image and true picture that network generates
It can be true enough.
Image removes the weighting parameter of rain part in 1-2, pre-training GAN network.Without rain and artificial have as shown in figure 3, practical
Rain figure picture is to for going the pre-training of rain fractional weight in network.The Loss of backpropagation includes that network generates image O
(output) the Perceptual loss (L between true picture T (Target)P), judge network to generate image judgement
As a result loss function (the L that backpropagation obtains is carried outGAN) two parts.The loss function for generating network output training is as follows:
LG=α LP(O,T)+βLGAN(O)
Wherein, α, β respectively represent the weight parameter of each loss function.
The weight parameter of rain part is gone in 1-3, real road image accurate adjustment network when not raining and is rainy.Such as Fig. 4 institute
Show, due to manually generated rainy image and it is practical scheme, as difference, network is avoided to intend the excessive of manually generated image
It closes, when last accurate adjustment GAN network, GAN loss function (L is used onlyGAN) adjustment network in weight parameter.
1-4, real road image and corresponding the rainfall mark when raining collect the weight for being trained rainfall detection part
Parameter.As shown in figure 5, the semantic information of rainy image, characteristic pattern generated can be extracted by generating the coded portion in network
It can include input picture most information.This feature figure is led into a convolutional network (Conv).After convolution, the spy of generation
Sign vector is directed into full connection layer network (FCs).Full connection layer network output is the testing result of rainfall.Generate network parameter
After the completion of training, need the fixed weight parameter for generating coded portion in network, real road image when by raining with it is right
The convolutional layer and full articulamentum of the rainfall mark collection training answered herein.
The application of 1-5, network.As shown in fig. 6, can directly be used after all weight parameter trainings in GAN network
Network completion image removes rain processing and rainfall detection.Real road image when input is rained, by the volume for generating network
Code part and decoded portion, the clear image after rain can be obtained.Also, the characteristic pattern of coded portion lead to convolutional layer and
Full articulamentum obtains the rainfall size of the figure.
In step 1-1, if the used data set owner of training passes through actual acquisition and artificial synthesized.Rainy true road
Road picture acquires under the weather condition of different rainfall sizes;In order to the case where the rainy day as close possible to the true road of no rain
Road image acquires in the case where being largely at the cloudy day or before and after raining.
It, can be according to raindrop iconic model formula when making artificial rain image:
I=TGT+Mraindrop
Wherein, I is the rainy picture of input;TGTFor no raindrop clearly real background image;MraindropFor the figure of raindrop
Picture.
Artificial rain image can be the artificial raindrop model and back for synthesizing by computer, generating by computer random
Scape image is combined, that is, generates image when artificial rain.Or the artificial scene manufactured when raining, it then obtains before raining
Image pair at uniform location afterwards.
The present invention is mainly directed towards intelligent automobile, and controller can be driven by intelligence and directly controls rain brush work, using only common
In-vehicle camera can complete the detection of rainfall, without additional addition rainfall detection sensor;Using deep learning GAN network implementations
Rainfall detection function, and that realizes rainy image goes rain function, makes image clearly, and meet real road scene, after being conducive to
Phase image procossing, the network provided can only acquire the training that the real road image to rain and without rainy day gas completes network, raw
At go rain figure picture to be more in line with actual traffic scene.
In short, the present invention provides a kind of this line traffic control windshield wiper control system based on deep learning GAN network, it is different from tradition
The Wiper system of rainfall inductive sensor systems, the detection that raindrop are completed by vision camera can be achieved in the present invention, and intelligence drives
Device processed can directly control the unlatching and working frequency of rain brush.Intelligence is driven controller and is integrated with based on deep learning GAN network algorithm, can
Rainfall detection is carried out to the image of in-vehicle camera and removes the raindrop in image, makes image clearly, improves the effect of environment sensing.
Claims (10)
1. a kind of line traffic control windshield wiper control system based on GAN network, to realize that the rain brush of intelligent automobile controls, feature exists
In, the system includes sequentially connected forward sight camera, that intelligence drives controller, entire car controller, windshield-wiper controller, driving motor is total
At, retarder and windshield wiper;
Forward sight camera acquires road ahead image in real time, and image is transferred to intelligence by USB data line and drives controller, and intelligence drives
The GAN network architecture is equipped with inside device processed, the raindrop in image to remove rainy day acquisition, and pass through image real-time detection vehicle
Periphery rainfall size, and the adjustment signal of rain brush gear is transmitted to entire car controller by CAN line, entire car controller obtains intelligence
After driving the CAN signal of controller, the working condition of windshield-wiper controller is adjusted in real time, and windshield-wiper controller is controlled by pulse signal and driven
Dynamic motor assembly work, control rain brush remove raindrop on the outside of front windshield.
2. a kind of line traffic control windshield wiper control system based on GAN network according to claim 1, which is characterized in that described
Forward sight camera is fixed on the inside of front windshield, coverage area when angular field of view works without departing from rain brush.
3. a kind of line traffic control windshield wiper control system based on GAN network according to claim 1, which is characterized in that described
Intelligence drive controller be high-performance development board, inside configuration linux system, around be equipped with USB, HDMI and Ethernet signal connect
Mouthful, to transmit signal and debugging exploitation.
4. a kind of line traffic control windshield wiper control system based on GAN network according to claim 1, which is characterized in that described
Intelligence drives controller and is internally integrated CAN card, and the CAN line control signal of entire car controller is sent or received by CAN card.
5. a kind of line traffic control windshield wiper control system based on GAN network according to claim 3, which is characterized in that described
Intelligence drives the linux system configured in controller and is equipped with image processing algorithm, which uses the GAN net of deep learning
Network model, realizes rainfall estimation and the removal of image raindrop from the feature representation of GAN network internal, and rainfall is estimated for adjusting rain
The frequency of brush, raindrop removal are repaired for removing the raindrop region in image, and by raindrop region, the image after making rain
Close to true traffic environment.
6. a kind of line traffic control windshield wiper control system based on GAN network according to claim 5, which is characterized in that described
GAN network model, including generate network, judge network and the full connection layer network of rainfall estimation, wherein generating network includes coding
Part and decoded portion generate network and generate the image after removing rain in real time, by whether judging Network Check image generated
It is consistent with true environment, the characteristic pattern that coded portion generates connects the full connection layer network of rainfall estimation by a convolutional layer, uses
To obtain the size of vehicle-surroundings rainfall.
7. a kind of line traffic control windshield wiper control system based on GAN network according to claim 5, which is characterized in that training GAN
When the data set of network model includes not rainy actual traffic acquire without rain figure picture, artificial rainy rainy image, rain when
The rainy image of actual traffic acquisition and it is rainy when actual traffic acquisition rainy image corresponding rainfall mark collection.
8. a kind of line traffic control windshield wiper control system based on GAN network according to claim 7, which is characterized in that described
Reality goes the pre-training of rain fractional weight without rain figure picture and artificial rainy image for network, and described is not rainy and artificial rainy
Traffic image for the weight parameter for going to rain part in pre-training GAN network, it is practical without rain and practical rainy mileage chart
Picture is to the weight parameter for going to rain part for accurate adjustment network, the rainy image with corresponding rainfall mark of actual traffic acquisition when raining
Note collects the weight parameter for training rainfall detection part.
9. a kind of line traffic control windshield wiper control system based on GAN network according to claim 8, which is characterized in that training GAN
When network model, principle is first fought according to generation, while training generates network and judges network, it is fixed to generate after training
The parameter of network finally trains the network parameter of full articulamentum.
10. a kind of line traffic control windshield wiper control system based on GAN network according to claim 1, which is characterized in that described
For driving motor assembly using stepper motor as rain brush driving motor, the retarder uses turbine and worm slowing-down structure,
Transverse profile curve is involute.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910017056.0A CN109760635B (en) | 2019-01-08 | 2019-01-08 | Drive-by-wire windshield wiper control system based on GAN network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910017056.0A CN109760635B (en) | 2019-01-08 | 2019-01-08 | Drive-by-wire windshield wiper control system based on GAN network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109760635A true CN109760635A (en) | 2019-05-17 |
CN109760635B CN109760635B (en) | 2020-11-27 |
Family
ID=66452762
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910017056.0A Active CN109760635B (en) | 2019-01-08 | 2019-01-08 | Drive-by-wire windshield wiper control system based on GAN network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109760635B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110481506A (en) * | 2019-09-27 | 2019-11-22 | 深圳市豪恩汽车电子装备股份有限公司 | Motor vehicle rain brush automatic control system and method |
CN111540193A (en) * | 2020-03-13 | 2020-08-14 | 华南理工大学 | Traffic data restoration method for generating countermeasure network based on graph convolution time sequence |
WO2021017445A1 (en) * | 2019-07-31 | 2021-02-04 | 浙江大学 | Convolutional neural network rainfall intensity classification method and quantification method aimed at rainy pictures |
WO2022193154A1 (en) * | 2021-03-16 | 2022-09-22 | 深圳市大疆创新科技有限公司 | Windshield wiper control method, automobile, and computer-readable storage medium |
CN117422689A (en) * | 2023-10-31 | 2024-01-19 | 南京邮电大学 | Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7 |
CN117422689B (en) * | 2023-10-31 | 2024-05-31 | 南京邮电大学 | Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7 |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103543638A (en) * | 2013-10-10 | 2014-01-29 | 山东神戎电子股份有限公司 | Automatic windshield wiper control method |
JP2014133425A (en) * | 2013-01-08 | 2014-07-24 | Mazda Motor Corp | Raindrop detection device |
CN104477132A (en) * | 2015-01-02 | 2015-04-01 | 江苏新瑞峰信息科技有限公司 | Automobile automatic windscreen wiper control system |
CN108230278A (en) * | 2018-02-24 | 2018-06-29 | 中山大学 | A kind of image based on generation confrontation network goes raindrop method |
CN108528395A (en) * | 2018-04-08 | 2018-09-14 | 广州大学 | A kind of Vehicular intelligent rain scrape control method and system based on image recognition |
-
2019
- 2019-01-08 CN CN201910017056.0A patent/CN109760635B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2014133425A (en) * | 2013-01-08 | 2014-07-24 | Mazda Motor Corp | Raindrop detection device |
CN103543638A (en) * | 2013-10-10 | 2014-01-29 | 山东神戎电子股份有限公司 | Automatic windshield wiper control method |
CN104477132A (en) * | 2015-01-02 | 2015-04-01 | 江苏新瑞峰信息科技有限公司 | Automobile automatic windscreen wiper control system |
CN108230278A (en) * | 2018-02-24 | 2018-06-29 | 中山大学 | A kind of image based on generation confrontation network goes raindrop method |
CN108528395A (en) * | 2018-04-08 | 2018-09-14 | 广州大学 | A kind of Vehicular intelligent rain scrape control method and system based on image recognition |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2021017445A1 (en) * | 2019-07-31 | 2021-02-04 | 浙江大学 | Convolutional neural network rainfall intensity classification method and quantification method aimed at rainy pictures |
CN110481506A (en) * | 2019-09-27 | 2019-11-22 | 深圳市豪恩汽车电子装备股份有限公司 | Motor vehicle rain brush automatic control system and method |
CN111540193A (en) * | 2020-03-13 | 2020-08-14 | 华南理工大学 | Traffic data restoration method for generating countermeasure network based on graph convolution time sequence |
CN111540193B (en) * | 2020-03-13 | 2022-07-26 | 华南理工大学 | Traffic data restoration method for generating countermeasure network based on graph convolution time sequence |
WO2022193154A1 (en) * | 2021-03-16 | 2022-09-22 | 深圳市大疆创新科技有限公司 | Windshield wiper control method, automobile, and computer-readable storage medium |
CN117422689A (en) * | 2023-10-31 | 2024-01-19 | 南京邮电大学 | Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7 |
CN117422689B (en) * | 2023-10-31 | 2024-05-31 | 南京邮电大学 | Rainy day insulator defect detection method based on improved MS-PReNet and GAM-YOLOv7 |
Also Published As
Publication number | Publication date |
---|---|
CN109760635B (en) | 2020-11-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109760635A (en) | A kind of line traffic control windshield wiper control system based on GAN network | |
US11891023B1 (en) | Using camera data to manage a vehicle parked outside in cold climates | |
CN108921013B (en) | Visual scene recognition system and method based on deep neural network | |
CN105512623B (en) | Based on multisensor travelling in fog day vision enhancement and visibility early warning system and method | |
CN113449577A (en) | Robust object detection and classification using RGB + events | |
CN110781883B (en) | Parking space size recognition system and method based on 360-degree looking-around camera | |
CN103543638B (en) | A kind of rain brush control method automatically | |
CN107862293A (en) | Radar based on confrontation generation network generates colored semantic image system and method | |
EP2539197A2 (en) | Automatic vehicle equipment monitoring, warning, and control system | |
CN111723849A (en) | Road adhesion coefficient online estimation method and system based on vehicle-mounted camera | |
CN110135302B (en) | Method, device, equipment and storage medium for training lane line recognition model | |
CN109934108A (en) | The vehicle detection and range-measurement system and implementation method of a kind of multiple target multiple types | |
CN105139347A (en) | Polarized image defogging method combined with dark channel prior principle | |
CN103714659A (en) | Fatigue driving identification system based on double-spectrum fusion | |
CN111340951A (en) | Ocean environment automatic identification method based on deep learning | |
CN111161160B (en) | Foggy weather obstacle detection method and device, electronic equipment and storage medium | |
CN102490646A (en) | Intelligent vehicle lamp system with driver eyesight following function and control method for intelligent vehicle lamp system | |
JP2023549914A (en) | Correcting surround view camera system images in rainy, backlit, and dirty conditions | |
WO2021036559A1 (en) | Automatic vehicle window cleaning method and apparatus | |
CN116829919A (en) | Automatic driving system testing method, testing system and storage medium based on aerial survey data | |
CN109963144A (en) | A kind of vehicle-mounted identifying system based on AR-HUD | |
CN110956146A (en) | Road background modeling method and device, electronic equipment and storage medium | |
CN113553998A (en) | Anti-dazzling snapshot method for license plate at night on expressway based on deep learning algorithm | |
CN111756971B (en) | Unmanned vehicle vision enhancement device and method in extreme environment | |
CN116486359A (en) | All-weather-oriented intelligent vehicle environment sensing network self-adaptive selection method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |