CN115294037A - Digital attack counterattack sample generation method facing traffic information perception - Google Patents

Digital attack counterattack sample generation method facing traffic information perception Download PDF

Info

Publication number
CN115294037A
CN115294037A CN202210881703.4A CN202210881703A CN115294037A CN 115294037 A CN115294037 A CN 115294037A CN 202210881703 A CN202210881703 A CN 202210881703A CN 115294037 A CN115294037 A CN 115294037A
Authority
CN
China
Prior art keywords
network
target
disturbance
traffic information
sample
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
Application number
CN202210881703.4A
Other languages
Chinese (zh)
Inventor
黄世泽
张肇鑫
刘晓雯
张兵杰
秦晋哲
宋冠群
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tongji University
Original Assignee
Tongji University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tongji University filed Critical Tongji University
Priority to CN202210881703.4A priority Critical patent/CN115294037A/en
Publication of CN115294037A publication Critical patent/CN115294037A/en
Pending legal-status Critical Current

Links

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/50—Monitoring users, programs or devices to maintain the integrity of platforms, e.g. of processors, firmware or operating systems
    • G06F21/57—Certifying or maintaining trusted computer platforms, e.g. secure boots or power-downs, version controls, system software checks, secure updates or assessing vulnerabilities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computer Hardware Design (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Security & Cryptography (AREA)
  • Software Systems (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a digital attack counterattack sample generation method facing traffic information perception, which mainly aims at an example segmentation detector network. The method comprises the following steps: acquiring an original picture to be detected; inputting the picture to an instance segmentation network; calculating a loss function of the picture in the example segmentation network; calculating the countermeasure disturbance by using a gradient algorithm; utilizing a disturbance interception method to limit the counterdisturbance in a target area; adding the confrontation sample which is an original picture and the confrontation disturbance; and setting iteration times, performing loop iteration and outputting a confrontation sample. By taking the YOLACT instance segmentation network as an example, the countermeasure sample can be generated through the method, the security holes and problems existing in the YOLACT instance segmentation network are exposed, and effective security effect verification is carried out on the traffic information perception system.

Description

Digital attack counterattack sample generation method facing traffic information perception
Technical Field
The invention relates to the field of rail transit, in particular to a traffic information perception oriented method for generating a countercheck sample for digital attack on a YOLACT instance segmentation network.
Background
In recent years, a vehicle-mounted traffic information sensing system has been rapidly developed, and the purpose of the sensing system is to sense the operating environment and state of a vehicle. Generally, a camera is installed in a cab, a video image of the running environment of the vehicle is shot and acquired, the running environment is perceived and identified through an intelligent algorithm, and then an intelligent decision is made for the running environment. The running environment of the vehicle is quite complex and mainly comprises three aspects of complex road surface conditions, complex and variable weather conditions and complex illumination conditions. The traditional image processing method is difficult to solve the problem of complex operation environment perception, and a deep learning algorithm is required to be adopted to detect the operation environment state in real time.
The confrontational sample is a research focus in the computer field in recent years, and the confrontational sample poses a great threat to the reliability and the safety of the deep learning network. Taking computer visual perception as an example, an anti-aliasing sample refers to an image artificially added with tiny noise, and under the condition that a human can correctly perceive the image, a deep learning model gives an incorrect perception result which is completely different from human perception. In application scenarios requiring high reliability, such as the field of automatic driving, roads, vehicles, and pedestrians in traffic environments need to use an example segmentation deep learning network for locating a target position. If the detected image is an anti-noise sample image interfered by anti-noise, the example segmentation network cannot correctly identify the target in the traffic environment, and a traffic accident can be seriously caused. In the prior art, chinese patent application CN114359672A, "Adam-based iterative fast gradient descent anti-attack method" and chinese patent application CN108491837a "anti-attack method for Improving robustness of license plate attack" and other technologies and research papers "activating adaptive attacks on following neural networks, contextual-based gradient-based attack", "embedded fool: a method to generate adaptive samples based on example segmentation network" are all directed to the classifier network for generating anti-samples, and do not relate to the introduction of the method for generating anti-samples based on the yol example segmentation network, and the anti-samples directed to the classifier cannot successfully attack the example segmentation network.
The difficulty of generating the countermeasure sample for the example segmentation network is far greater than the difficulty for the classifier and the target detector, only one target in the classifier is needed, accurate position coordinates of the target do not need to be given, the target detector contains pictures of various different targets and gives accurate position coordinates of the corresponding target, and the example segmentation network not only needs to classify the target category and the target position, but also needs to segment the outline of the target. The existing countermeasure sample generation method mainly aims at a classifier and a target detector, a traffic information sensing system needs to realize the outline segmentation and identification of a track area and an obstacle in a running environment in each frame of picture, and the existing countermeasure sample generation method is not suitable, so that the existing research method has little threat to the traffic information sensing system. Therefore, a countercheck sample generation method capable of achieving a high attack success rate for an example segmentation network is needed to optimize an example segmentation model and enhance the defense capability of the countercheck sample.
Disclosure of Invention
The invention provides a method for generating a confrontation sample aiming at a YOLACT example segmentation network to carry out digital attack, which adopts a gradient algorithm to generate the confrontation sample to attack the YOLACT example segmentation network, exposes security holes and problems existing in the YOLACT, helps to improve or propose a more effective defense method, improves the reliability of a traffic information perception system, and is an important measure for ensuring driving safety and passenger safety.
The invention provides a traffic information perception-oriented digital attack counterattack sample generation method, which comprises the following steps:
(1) Acquiring an original picture to be detected;
(2) Inputting the picture to an instance segmentation network;
(3) Calculating a loss function of the picture in the example segmentation network;
(4) Calculating the anti-disturbance by using a gradient algorithm;
(5) Utilizing a disturbance interception method to limit the counterdisturbance in a target area;
(6) Adding the confrontation sample which is an original picture and the confrontation disturbance;
(7) And setting iteration times, performing loop iteration and outputting a confrontation sample.
Further, according to the traffic information perception-oriented digital attack countermeasure sample generation method, a perception network of a main attack is a YOLACT instance segmentation network.
Further, the method for generating the digital attack countermeasure sample for traffic information perception is characterized in that the loss function in step (3) is expressed by the following formula:
Figure BDA0003764407760000031
wherein L is box (x + η, θ, l, g) represents a loss function of localization;
L cls (x + η, θ, c) represents a loss function of the classification;
L mask (x+x η ,θ,m,m gt ) A loss function representing the segmentation;
x is the input original image;
t is a certain iteration process;
Figure BDA0003764407760000032
is the calculated opposition perturbation;
θ is a network structure parameter of the YOLACT instance split network;
l is the target location predicted by the network;
g is a target true location tag;
m is target prediction mask information;
m gt real mask information for the target;
and c, predicting the confidence of the target through the network.
Further, the method for generating the digital attack countermeasure sample facing the traffic information perception is characterized in that the countermeasure disturbance is calculated by using a gradient algorithm in the step (4), and the formula is as follows:
Figure BDA0003764407760000033
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003764407760000034
the gradient corresponding to the loss function;
alpha is the learning rate;
t is a certain iteration process;
t is the total iteration number;
y gt is the true target tag value;
x is an input image;
Figure BDA0003764407760000035
is the calculated opposition perturbation;
θ is a network structure parameter of the yolcat instance split network.
Further, the method for generating the digital attack countermeasure sample facing the traffic information perception is characterized in that the countermeasure disturbance is limited to the target area by the disturbance interception method in the step (5), and the formula is as follows:
Figure BDA0003764407760000041
Figure BDA0003764407760000042
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003764407760000043
for cutting off imagesTaking the subsequent countermeasure disturbance;
I X is the image pixel position coordinates;
b is the coordinate area of a single target;
b is the resolution size of the image;
b X a coordinate area that is a target;
box top ，box left ，box bottom ，box right are respectively b X The abscissa minimum, the ordinate minimum, the abscissa maximum, and the ordinate maximum of the coordinate region.
Further, the method for generating the digital attack countermeasure sample for traffic information perception is characterized in that the countermeasure sample in the step (6) is an original picture and is added with countermeasure disturbance, and the formula is as follows:
Figure BDA0003764407760000044
wherein the content of the first and second substances,
Figure BDA0003764407760000045
to challenge the sample;
x is an input image;
Figure BDA0003764407760000046
to combat the disturbance;
Proj ε () Limiting a function for image pixel values;
ε is the size of the limit pixel value.
The invention has the beneficial effects that:
(1) And generating a YOLACT example segmentation network for resisting sample attack by using a gradient algorithm. By dividing the disturbance in the target area, the confrontation sample is more accurate and more concealed.
(2) The potential safety hazard of the deep learning algorithm is fully exposed through analog-digital attack, safety effect verification is provided for a traffic information perception system in engineering application, and a better defense method is sought in the engineering application through an attack effect.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following descriptions are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a general flow diagram of a method for generating countersample for digital attacks against a TOLACT example split network according to an embodiment of the present invention;
fig. 2 is an original picture of a tramcar running environment of an experimental example of the present invention.
Fig. 3 shows the coordinates and the contour of the position of the rail-bound region and the obstacle obtained by the example segmentation network during the operation of the tramcar.
Fig. 4 is a loss function value obtained by a gradient algorithm according to an experimental example of the present invention.
FIG. 5 is a graph showing the results of generation of a challenge sample according to an experimental example of the present invention.
Fig. 6 is a result of identification of a challenge sample according to an experimental example of the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the accompanying drawings, and it should be understood that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention provides a method for generating countersample by using gradient algorithm to carry out digital attack on a Yolcoct instance segmentation network, fig. 1 is a flow chart of the countersample generation method for carrying out digital attack on the Yolcoct instance segmentation network according to the invention, and as shown in fig. 1, the flow comprises the following steps:
(1) Acquiring an original picture to be detected;
(2) Inputting the picture to an instance segmentation network;
(3) Calculating a loss function of the picture in the example segmentation network;
(4) Calculating the countermeasure disturbance by using a gradient algorithm;
(5) Adding the original picture and the countermeasure disturbance;
(6) And setting iteration times, performing loop iteration and outputting a confrontation sample.
Through the steps, based on the tramcar running environment picture, a countermeasure sample is generated by adopting a gradient algorithm to attack the YOLACT example segmentation network. The steps can successfully generate the confrontation sample which can more effectively carry out digital attack on the YOLACT example segmentation network, expose the potential safety hazard and the problem of the neural network and ensure the driving safety.
Examples
An alternative embodiment of steps (1) to (6) of the present invention will be described in detail with reference to fig. 2 to 6.
Fig. 2 is an original picture of the tramcar running environment according to the experimental example of the present invention, and the value range of the pixel value is [0,1].
Fig. 3 shows the coordinates and the contour of the position of the rail-bound region and the obstacle obtained by the example segmentation network during the operation of the tramcar.
Defining a loss function as described in step (3), which is formulated as follows:
Figure BDA0003764407760000061
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0003764407760000062
a loss function representing the location, the resulting L being calculated box ＝0.09；
Figure BDA0003764407760000063
A loss function representing the classification is performed,L cls ＝0.33；
Figure BDA0003764407760000064
representing a loss function of the segmentation, L mask ＝0.07；
x is the input image, as shown in FIG. 2;
Figure BDA0003764407760000065
is the calculated countermeasure disturbance with an initial value of x η ＝0；
θ is a network structure parameter of the YOLACT instance split network;
l is the target location predicted by the network;
g is a target true location tag;
t is a certain iteration process;
m is target prediction mask information;
m gt real mask information for the target;
and c, predicting the confidence of the target through the network.
Fig. 4 is a loss function value obtained by a gradient algorithm according to an experimental example of the present invention. This shows that the algorithm provided by the invention can make the loss value larger and larger, and the recognition result is invalid and invalid.
Defining the step (4) of calculating the confrontation disturbance on the whole picture by using a gradient algorithm, wherein the formula is as follows:
Figure BDA0003764407760000071
wherein the content of the first and second substances,
Figure BDA0003764407760000072
the gradient corresponding to the loss function;
alpha is the learning rate, and the value is alpha =0.2;
t is a single iteration process;
t is the total iteration number, and the value T =5;
y gt is a true target tag value;
x is the input image, as shown in FIG. 2;
Figure BDA0003764407760000073
is the calculated opposition perturbation, initially taken to be
Figure BDA0003764407760000074
θ is a network structure parameter of the YOLACT instance split network.
Defining the method for restraining the counterdisturbance in the target area by using the disturbance interception method in the step (5), wherein the formula is as follows:
Figure BDA0003764407760000075
Figure BDA0003764407760000076
wherein the content of the first and second substances,
Figure BDA0003764407760000077
the counterdisturbance after the image is intercepted is obtained;
I X is the image pixel position coordinates;
b is the coordinate area of a single target;
b is the resolution size of the image, B = (1920,1080);
b X a coordinate area that is a target;
box top ,box left ,box bottom ,box right are respectively b X Abscissa minimum, ordinate minimum, abscissa maximum, ordinate maximum, box of coordinate region top ,box left ,box bottom ,box right ＝(314,0,1080,1920)。
Defining the confrontation sample in the step (6) as the original picture and the confrontation sample to be added, and the formula is as follows:
Figure BDA0003764407760000081
wherein the content of the first and second substances,
Figure BDA0003764407760000082
to challenge the sample, as shown in fig. 5;
x is the input image, as shown in FIG. 2;
Figure BDA0003764407760000083
to combat the disturbance;
Proj ε () Limiting a function for the image pixel values;
epsilon is the size of the limiting pixel value and takes the value epsilon =0.2.
FIG. 5 is a graph showing the result of generation of a challenge sample according to an experimental example of the present invention. The algorithm provided by the invention can generate the operation environment perception picture which looks normal to human eyes and is almost the same as the original picture.
Fig. 6 is a picture of the identification result of the confrontation sample according to the experimental example of the present invention. The original track area location was erroneously identified as a human with 100% confidence. An originally human location is falsely detected as a car with 100% confidence. And in the area without the object, the example split network also erroneously detected the object (cat: 100%, person: 97%). That is to say, the yolcat network does not correctly identify the trackbound regions and obstacles in the operating environment, which indicates that the countermeasure sample generation algorithm provided by the invention is effective.
Through the processing of the steps, the YOLACT instance segmentation network cannot correctly identify the image after adding the counternoise, that is, the method successfully generates the countersample aiming at the YOLACT instance segmentation network.
As known in the art, the traffic information sensing system includes a collecting device, a target detection network (YOLACT example division network), a recognition and determination module, and a recognition and output module, and uses a camera installed at the front end of a cab to collect image data of a driving environment, input the image into the corresponding target detection network for target recognition, and return the recognition result to a vehicle. According to the countermeasure sample generated by the method, the countermeasure sample attacks the traffic information perception system with the YOLACT instance segmentation network, so that the network target detection is wrong, the potential safety hazard and the problem of a deep learning algorithm YOLACT are further exposed, and a basis is provided for selecting a proper perception algorithm and a proper target identification algorithm in engineering application. If a YOLACT example is used for dividing the network in the traffic information perception system, the confrontation sample generated by the method can cause the traffic information perception system to have recognition error and fault, and threaten traffic safety. The method finds out a mechanism and a method which are possibly adopted by an attacker and threaten the safety of the traffic information sensing system, guides researchers to research a more effective defense method and a detection method of the confrontation sample, upgrades the defense system, for example, processes an input image, increases the links of detecting the confrontation sample, and the like, and ensures the traffic system and the driving safety.
It will be appreciated by those skilled in the art that embodiments of the invention may be provided as a method. A system or a computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be understood that the above examples are only for clarity of illustration and are not intended to limit the embodiments. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. And obvious variations or modifications therefrom are within the scope of the invention.

Claims (6)

1. A digital attack countermeasure sample generation method oriented to traffic information perception is characterized by comprising the following steps:
(1) Acquiring an original picture to be detected;
(2) Inputting the picture to an instance segmentation network;
(3) Calculating a loss function of the picture in the example segmentation network;
(4) Calculating the countermeasure disturbance by using a gradient algorithm;
(5) Utilizing a disturbance interception method to limit the counterdisturbance in a target area;
(6) Adding the original picture and the confrontation disturbance;
(7) And setting iteration times, performing loop iteration and outputting a confrontation sample.
2. The method for generating digital attack countermeasure samples for traffic information awareness according to claim 1, wherein the awareness network for attacks is a YOLACT instance split network.
3. The method for generating digital attack countermeasure samples for traffic information perception according to claim 1, wherein the loss function in step (3) is formulated as follows:
Figure FDA0003764407750000011
wherein L is box (x + η, θ, l, g) represents a loss function of localization;
L cls (x + η, θ, c) represents a loss function of the classification;
L mask (x+x η ,θ,m,m gt ) A loss function representing the segmentation;
x is the input original image;
t is a certain iteration process;
Figure FDA0003764407750000012
is the calculated opposition perturbation;
theta is a network structure parameter of the YOLACT instance partitioning network;
l is the target location predicted by the network;
g is a target true location tag;
m is target prediction mask information;
m gt real mask information for the target;
and c, predicting the confidence of the target through the network.
4. The method for generating digital attack countermeasure samples for traffic information perception according to claim 1 or 3, wherein the countermeasure disturbance is calculated by the following formula:
Figure FDA0003764407750000021
wherein the content of the first and second substances,
Figure FDA0003764407750000022
the gradient corresponding to the loss function;
alpha is the learning rate;
t is a certain iteration process;
t is the total iteration number;
y gt is the true target tag value;
x is the input image;
Figure FDA0003764407750000023
is the calculated opposition perturbation;
θ is a network structure parameter of the YOLACT instance split network.
5. The method for generating digital attack countermeasure samples for traffic information perception according to claim 1, wherein the disturbance interception method used in step (5) limits the countermeasure disturbance to the target area according to the following formula:
Figure FDA0003764407750000024
Figure FDA0003764407750000025
wherein the content of the first and second substances,
Figure FDA0003764407750000026
the counterdisturbance after the image is intercepted is obtained;
I X is the image pixel position coordinates;
b is the coordinate area of a single target;
b is the resolution size of the image;
b X a coordinate area that is a target;
box top ，box left ，box bottom ，box right are respectively b X The abscissa minimum value, the ordinate minimum value, the abscissa maximum value, and the ordinate maximum value of the coordinate region.
6. The method for generating digital attack countersample facing traffic information perception according to claim 1, wherein the countersample in step (6) is an original picture and the counterdisturbance are added, and the formula is as follows:
Figure FDA0003764407750000031
wherein, the first and the second end of the pipe are connected with each other,
Figure FDA0003764407750000032
to challenge the sample;
x is an input image;
Figure FDA0003764407750000033
to combat the disturbance;
Proj ε () Limiting a function for image pixel values;
ε is the size of the limit pixel value.
CN202210881703.4A 2022-07-26 2022-07-26 Digital attack counterattack sample generation method facing traffic information perception Pending CN115294037A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210881703.4A CN115294037A (en) 2022-07-26 2022-07-26 Digital attack counterattack sample generation method facing traffic information perception

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210881703.4A CN115294037A (en) 2022-07-26 2022-07-26 Digital attack counterattack sample generation method facing traffic information perception

Publications (1)

Publication Number Publication Date
CN115294037A true CN115294037A (en) 2022-11-04

Family

ID=83824438

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210881703.4A Pending CN115294037A (en) 2022-07-26 2022-07-26 Digital attack counterattack sample generation method facing traffic information perception

Country Status (1)

Country Link
CN (1) CN115294037A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115884172A (en) * 2022-12-01 2023-03-31 同济大学 Information encryption and decryption method based on countermeasure sample technology

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115884172A (en) * 2022-12-01 2023-03-31 同济大学 Information encryption and decryption method based on countermeasure sample technology
CN115884172B (en) * 2022-12-01 2023-07-04 同济大学 Information encryption and decryption method based on countermeasure sample technology

Similar Documents

Publication Publication Date Title
CN111506980B (en) Method and device for generating traffic scene for virtual driving environment
US20180357772A1 (en) Surrounding environment recognition device for moving body
CN107991671A (en) A kind of method based on radar data and vision signal fusion recognition risk object
US10783391B2 (en) Method and system for recognizing license plate
CN113554089A (en) Image classification countermeasure sample defense method and system and data processing terminal
JP2022506905A (en) Systems and methods for assessing perceptual systems
CN111783853B (en) Interpretability-based method for detecting and recovering neural network confrontation sample
CN112541404A (en) Physical attack counterattack sample generation method facing traffic information perception
CN115294037A (en) Digital attack counterattack sample generation method facing traffic information perception
US20120155711A1 (en) Apparatus and method for analyzing video
Hau et al. Shadow-catcher: Looking into shadows to detect ghost objects in autonomous vehicle 3d sensing
Salmane et al. Improving safety of level crossings by detecting hazard situations using video based processing
CN116977967A (en) Fake point cloud object identification method for automatic driving system
KR20190064147A (en) Vehicle's behavior analyzing system using aerial photo and analyzing method using the same
JPH11142168A (en) Environment-recognizing apparatus
CN115937827B (en) Monitoring video processing method for automobile emergency active risk avoidance
KR20210002893A (en) License plate recognition method using hybrid approach and system therefore
US10719718B2 (en) Device for enabling a vehicle to automatically resume moving
KR20170138842A (en) System and Method for Tracking Vehicle on the basis on Template Matching
Cuadri et al. A bioinspired collision detection algorithm for VLSI implementation
CN115359336A (en) Traffic information perception-oriented target stealth confrontation sample generation method
Xiao et al. Exorcising" Wraith": Protecting {LiDAR-based} Object Detector in Automated Driving System from Appearing Attacks
Tsuchiya et al. Real-time vehicle detection using a single rear camera for a blind spot warning system
Han et al. A unified anomaly detection methodology for lane-following of autonomous driving systems
CN112180913A (en) Special vehicle identification 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