CN108875640B - Method for testing cognitive ability of passable area in end-to-end unsupervised scene - Google Patents

Method for testing cognitive ability of passable area in end-to-end unsupervised scene Download PDF

Info

Publication number
CN108875640B
CN108875640B CN201810638850.2A CN201810638850A CN108875640B CN 108875640 B CN108875640 B CN 108875640B CN 201810638850 A CN201810638850 A CN 201810638850A CN 108875640 B CN108875640 B CN 108875640B
Authority
CN
China
Prior art keywords
scene
area
road surface
passable
testing
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.)
Active
Application number
CN201810638850.2A
Other languages
Chinese (zh)
Other versions
CN108875640A (en
Inventor
赵祥模
刘占文
沈超
高涛
樊星
董鸣
林杉
徐江
连心雨
陈婷
王润民
张凡
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.)
Changan University
Original Assignee
Changan 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 Changan University filed Critical Changan University
Priority to CN201810638850.2A priority Critical patent/CN108875640B/en
Publication of CN108875640A publication Critical patent/CN108875640A/en
Application granted granted Critical
Publication of CN108875640B publication Critical patent/CN108875640B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0463—Neocognitrons
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Traffic Control Systems (AREA)
  • Holo Graphy (AREA)

Abstract

The invention discloses a method for testing the cognitive ability of a passable area in an end-to-end unsupervised scene, which can not only reproduce real road information but also display rich road environment by displaying the road information on the ground of a test area; the road surface information is displayed on the ground of the test area, so that the test system can obtain the same detection environment as the actual drive test when in test; by adopting the holographic technique, the object on the road surface can be reproduced in a real and three-dimensional manner, so that the test environment is closer to the real environment; the trafficable region detection system adopts different environments to test to obtain different road surface scene information to detect trafficable regions in different scenes, and can comprehensively evaluate the cognitive ability of the trafficable region detection system; therefore, the cognitive ability of the passable area detection system can be comprehensively embodied; the testing method can provide an effective and low-risk testing and evaluating means for the unmanned intelligent vehicle before actual drive test.

Description

Method for testing cognitive ability of passable area in end-to-end unsupervised scene
Technical Field
The invention belongs to the technical field of visual scene processing, and particularly relates to a method for testing cognitive ability of a passable area in an end-to-end unsupervised scene.
Background
In order to promote the development of the unmanned intelligent vehicle technology, the unmanned intelligent vehicle needs to be tested under real road conditions to detect the problems in the related technology. However, even though the unmanned smart vehicle is allowed to be tested under real road conditions, the potential safety hazard and the testing cost caused by the unmanned smart vehicle are huge. Therefore, many methods adopt an off-line testing mode to test and evaluate the cognitive ability of the passable area, although the testing cost can be saved, the methods adopt a single image or short video to directly test the cognitive algorithm, the testing flexibility is limited by a data set, and the configuration of the self vision perception system of the unmanned intelligent vehicle is not comprehensively considered. Therefore, the ability of the unmanned intelligent vehicle to recognize the passable scene area cannot be comprehensively and truly reflected.
With the development of deep learning theory and technology, the detection method of the road surface travelable area based on deep learning is more and more perfect, but the road surface condition under the real scene is extremely complex, and the perfect detection method always has the condition that cannot be considered, so that a test mode which is close to the real scene and can realize the extreme environment scene is needed for testing the capacity of the cognitive algorithm.
Disclosure of Invention
The invention aims to provide a method for testing the cognitive ability of a passable area in an end-to-end unsupervised scene, so as to overcome the defects of the prior art.
In order to achieve the purpose, the invention adopts the following technical scheme:
a method for testing cognitive ability of a passable area in an end-to-end unsupervised scene comprises the following steps:
step 1), displaying the road surface information to a road plane display area, and performing equal-ratio recovery presentation;
step 2), constructing road surface holographic projection area information by adopting the holographic projection film, and acquiring a basic road surface non-interference passable area r00；
Step 3), generating scenes with different complexity by setting different scene parameters and utilizing holographic imaging;
step 4), detecting that the passable area under different scenes is r by using different road scene informationx(ii) a x is the combination of different road surface scene information;
step 5), defining the cognitive ability of the passable area as Sx
Figure BDA0001702113400000021
Where x ∈ Ω {00, 11, 22, 33, 44, 55, 12, 13, 14, 15, 23, 25, 123, 124, 125}, k is the number of test scenes under different scene parameters, i is a different scene coefficient, and α is a coefficient of the different scenesxIs the scene luminance coefficient, betaxIs the coefficient of rainfall, gammaxAs coefficient of scene fogging, σxIs the snow coefficient, mxNumber of vehicles, nxNumber of pedestrians, rROIxFor the intersection ratio of the detected passable area and the manually marked passable area,
Figure BDA0001702113400000022
wherein r isxRepresenting passable areas detected in different scenes, RxRepresenting the passable areas marked manually in different scenes;
step 6), calculating the average capacity of the trafficable region detection in each scene of the cognitive capacity of the trafficable region:
Figure BDA0001702113400000023
wherein N represents the number of omega;
step 7), calculating the stability of the cognitive ability of the scene passable area in the passable area detection under each scene:
SSD＝∑x∈Ω(SAVG-Sx)2。
further, in step 1), the road plane information is restored in an equal ratio and presented by using a graphic display device.
Further, the display part in the graphic display equipment is arranged on the ground of the test site, and the display area is not smaller than the size of the real road surface.
Further, the holographic projection area information in the step 3) includes road surface scene information, specifically including holographic models of people, vehicles, rain, fog and snow.
Further, the different scene parameters include scene brightness α, scene rain coefficient β, scene fog coefficient γ, scene snow coefficient σ, number of vehicles m in the scene, and number of pedestrians n in the scene.
Further, specifically, in step 3), holographic projection model rows of vehicles, people, rain, fog and snow are loaded on the basic road surface respectively for testing, and the detected road surface r is obtained11、r22、r33、r44、r55。
Further, holographic projection models of vehicles and pedestrians, vehicles and rain, vehicles and fog, vehicles and snow, people and rain, people and fog and people and snow are loaded on the basic road surface respectively for testing, and the detected passable area r is obtained12、r13、r14、r15、r23、r24、r25。
Further, a holographic projection model of vehicles, pedestrians and rain is loaded on the basic road surface simultaneously for testing, and the detected passable area r is obtained123(ii) a Loading holographic projection models of vehicles, pedestrians and fog on a basic road surface simultaneously for testing to obtain a passable detection area r124(ii) a Holographic projection for simultaneously loading vehicles, pedestrians and snow on basic pavementThe shadow model is tested to obtain a passable area r125。
Compared with the prior art, the invention has the following beneficial technical effects:
according to the method for testing the cognitive ability of the passable area in the end-to-end unsupervised scene, disclosed by the invention, the real pavement information can be reproduced and the abundant pavement environment can be displayed in the way by displaying the pavement information on the ground of the test area; the road surface information is displayed on the ground of the test area, so that the test system can obtain the same detection environment as the actual drive test when in test; by adopting the holographic technique, the object on the road surface can be reproduced in a real and three-dimensional manner, so that the test environment is closer to the real environment; the trafficable region detection system adopts different environments to test to obtain different road surface scene information to detect trafficable regions in different scenes, and can comprehensively evaluate the cognitive ability of the trafficable region detection system; therefore, the cognitive ability of the passable area detection system can be comprehensively embodied; the testing method can provide an effective and low-risk testing and evaluating means for the unmanned intelligent vehicle before actual drive test.
Drawings
FIG. 1 is a schematic diagram of the test protocol of the present invention.
FIG. 2 is an example of a test scenario of the present invention.
In the figure, 1 is a road plane display area, 2 is a holographic imaging surface, 3 is a holographic scene area, 4 is holographic projection, and 5 is a tested equipment area; a is an original scene graph, b is an adjusted brightness graph 1, c is an adjusted brightness graph 2, d is a rain scene graph, e is a fog scene graph, and f is a snow scene graph.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
as shown in fig. 1 and fig. 2, a method for testing cognitive ability of a vision-based end-to-end scene passable area includes the following steps:
step 1), displaying the road surface information to a road plane display area, and performing equal-ratio recovery presentation;
step (ii) of2) Constructing holographic projection area information of the pavement by adopting the holographic projection film, and acquiring an interference-free passable area r of the basic pavement00；
Step 3), generating scenes with different complexity by setting different scene parameters and utilizing holographic imaging;
step 4), detecting that the passable area under different scenes is r by using different road scene informationx(ii) a x is the combination of different road surface scene information;
step 5), defining the cognitive ability of the passable area as Sx
Figure BDA0001702113400000051
Wherein x ∈ Ω ═ {00, 11, 22, 33, 44, 55, 12, 13, 14, 15, 23, 25, 123, 124, 125}, k is the number of test scenes under different scene parameters, i is different scene coefficients, α isxIs the scene luminance coefficient, betaxIs the coefficient of rainfall, gammaxAs coefficient of scene fogging, σxIs the snow coefficient, mxNumber of vehicles, nxNumber of pedestrians, rROIxFor the intersection ratio of the detected passable area and the manually marked passable area,
Figure BDA0001702113400000052
wherein r isxRepresenting passable areas detected in different scenes, RxRepresenting the passable areas marked manually in different scenes;
step 6), calculating the average capacity of the trafficable region detection in each scene of the cognitive capacity of the trafficable region:
Figure BDA0001702113400000053
wherein N represents the number of omega;
step 7), calculating the stability of the cognitive ability of the scene passable area in the passable area detection under each scene:
SSD＝Σx∈Ω(SAVG-Sx)2。
in the step 1), carrying out equal ratio recovery on road plane information by using a graphic display device and presenting the road plane information;
the display component in the graphic display equipment is arranged on the ground of the test site, and the display area is not smaller than the size of a real road surface;
the holographic projection area information in the step 3) comprises road surface scene information, specifically comprising holographic models of people, vehicles, rain, fog and snow;
different scene parameters comprise scene brightness alpha, scene rain coefficient beta, scene atomization coefficient gamma, scene snow coefficient sigma, number of vehicles m under the scene, and number of pedestrians n under the scene;
specifically, in the step 3), holographic projection model line tests of vehicles, people, rain, fog and snow are respectively loaded on the basic road surface to obtain the r of the detected road surface11、r22、r33、r44、r55；
Loading holographic projection models of vehicles and pedestrians, vehicles and rain, vehicles and fog, vehicles and snow, people and rain, people and fog and people and snow on a basic road surface respectively for testing to obtain a passable detection area r12、r13、r14、r15、r23、r24、r25；
Simultaneously loading a vehicle, pedestrian and rain holographic projection model on a basic road surface for testing to obtain a passable detection area r123；
Loading holographic projection models of vehicles, pedestrians and fog on a basic road surface simultaneously for testing to obtain a passable detection area r124；
Loading holographic projection models of vehicles, pedestrians and snow on a basic road surface simultaneously for testing to obtain a passable detection area r125。
The invention relates to a method for testing and evaluating cognitive ability of a passable area in an end-to-end unsupervised scene, which is used for testing and evaluating the cognitive ability of the passable area in an end-to-end unsupervised sceneThe scene information of the same road surface detects that the passable area under different scenes is rxThe method can test and evaluate the cognitive ability of the scene passable area of the unmanned intelligent vehicle based on different image vision solutions, and can provide an effective and low-risk test and evaluation means for the unmanned intelligent vehicle before actual drive test.

Claims (5)

1. A method for testing cognitive ability of a passable area in an end-to-end unsupervised scene is characterized by comprising the following steps:
step 1), displaying the road surface information to a road plane display area, performing equal-ratio recovery presentation, and performing equal-ratio recovery and presentation on the road plane information by using a graphic display device; the display component in the graphic display equipment is arranged on the ground of the test site, and the display area is not smaller than the size of a real road surface;
step 2), constructing road surface holographic projection area information by adopting the holographic projection film, and acquiring a basic road surface non-interference passable area r00；
Step 3), generating scenes with different complexity by setting different scene parameters and utilizing holographic imaging; the holographic projection area information comprises road surface scene information, specifically comprises holographic models of people, vehicles, rain, fog and snow;
step 4), detecting that the passable area under different scenes is r by using different road scene informationx(ii) a x is the combination of different road surface scene information;
step 5), defining the cognitive ability of the passable area as Sx
Figure FDA0003364755910000011
Where x ∈ Ω {00, 11, 22, 33, 44, 55, 12, 13, 14, 15, 23, 25, 123, 124, 125}, k is the number of test scenes under different scene parameters, i is a different scene coefficient, and α is a coefficient of the different scenesxIs the scene luminance coefficient, betaxIs the coefficient of rainfall, gammaxAs coefficient of scene fogging, σxIs the snow coefficient, mxNumber of vehicles, nxNumber of pedestrians, rROIxFor the intersection ratio of the detected passable area and the manually marked passable area,
Figure FDA0003364755910000012
wherein r isxRepresenting passable areas detected in different scenes, RxRepresenting the passable areas marked manually in different scenes;
step 6), calculating the average capacity of the trafficable region detection in each scene of the cognitive capacity of the trafficable region:
Figure FDA0003364755910000021
wherein N represents the number of omega;
step 7), calculating the stability of the cognitive ability of the scene passable area in the passable area detection under each scene:
SSD＝∑x∈Ω(SAVG-Sx)2。
2. the method for testing the cognitive ability of the end-to-end unsupervised scene passable area according to claim 1, wherein the different scene parameters comprise scene brightness α, scene rain coefficient β, scene fog coefficient γ, scene snow coefficient σ, number of vehicles in the scene m, and number of pedestrians in the scene n.
3. The method for testing the cognitive ability of the passable area of the end-to-end unsupervised scene as claimed in claim 1, wherein in step 3), holographic projection model rows of vehicles, people, rain, fog and snow are loaded on the basic road surface respectively for testing, and the r is the detected road surface11、r22、r33、r44、r55。
4. The method for testing the cognitive ability of the passable area of the end-to-end unsupervised scene as claimed in claim 1, wherein holographic projection models of vehicles and pedestrians, vehicles and rain, vehicles and fog, vehicles and snow, people and rain, people and fog and people and snow are loaded on the basic road surface for testing respectively, and the passable area r is obtained12、r13、r14、r15、r23、r24、r25。
5. The method for testing the cognitive ability of the passable area of the end-to-end unsupervised scene as claimed in claim 1, wherein a holographic projection model of vehicles, pedestrians and rain is loaded on the basic road surface for testing to obtain a passable area r123(ii) a Loading holographic projection models of vehicles, pedestrians and fog on a basic road surface simultaneously for testing to obtain a passable detection area r124(ii) a Loading holographic projection models of vehicles, pedestrians and snow on a basic road surface simultaneously for testing to obtain a passable detection area r125。
CN201810638850.2A 2018-06-20 2018-06-20 Method for testing cognitive ability of passable area in end-to-end unsupervised scene Active CN108875640B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810638850.2A CN108875640B (en) 2018-06-20 2018-06-20 Method for testing cognitive ability of passable area in end-to-end unsupervised scene

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810638850.2A CN108875640B (en) 2018-06-20 2018-06-20 Method for testing cognitive ability of passable area in end-to-end unsupervised scene

Publications (2)

Publication Number Publication Date
CN108875640A CN108875640A (en) 2018-11-23
CN108875640B true CN108875640B (en) 2022-04-05

Family

ID=64340071

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810638850.2A Active CN108875640B (en) 2018-06-20 2018-06-20 Method for testing cognitive ability of passable area in end-to-end unsupervised scene

Country Status (1)

Country Link
CN (1) CN108875640B (en)

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102156977A (en) * 2010-12-22 2011-08-17 浙江大学 Vision-based road detection method
CN103528585A (en) * 2013-09-26 2014-01-22 中北大学 Path planning method of passable area divided at unequal distance
CN104408757A (en) * 2014-11-07 2015-03-11 吉林大学 Method and system for adding haze effect to driving scene video
CN104933708A (en) * 2015-06-07 2015-09-23 浙江大学 Barrier detection method in vegetation environment based on multispectral and 3D feature fusion
CN105266982A (en) * 2014-07-18 2016-01-27 南京专创知识产权服务有限公司 Intelligent wheel chair
CN105550665A (en) * 2016-01-15 2016-05-04 北京理工大学 Method for detecting pilotless automobile through area based on binocular vision
CN105807762A (en) * 2015-01-21 2016-07-27 福特全球技术公司 Autonomous driving refined in virtual environments
CN106657926A (en) * 2017-01-23 2017-05-10 高艳云 Information transmission system based on intelligent terminal
CN106627456A (en) * 2015-07-20 2017-05-10 国网河南省电力公司南阳供电公司 Tyre puncture prevention device
CN107284379A (en) * 2017-07-28 2017-10-24 合肥芯福传感器技术有限公司 AR optical projection systems and projecting method for vehicle safe driving
CN206619301U (en) * 2017-03-06 2017-11-07 公安部交通管理科学研究所 Strengthen the safe driving simulated training educational system of special efficacy with environment
CN107871418A (en) * 2017-12-27 2018-04-03 吉林大学 It is a kind of to be used to evaluate the man-machine experiment porch for driving reliability altogether
CN207466645U (en) * 2017-10-25 2018-06-08 李炳午 Safe driving assistant system

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102156977A (en) * 2010-12-22 2011-08-17 浙江大学 Vision-based road detection method
CN103528585A (en) * 2013-09-26 2014-01-22 中北大学 Path planning method of passable area divided at unequal distance
CN105266982A (en) * 2014-07-18 2016-01-27 南京专创知识产权服务有限公司 Intelligent wheel chair
CN104408757A (en) * 2014-11-07 2015-03-11 吉林大学 Method and system for adding haze effect to driving scene video
CN105807762A (en) * 2015-01-21 2016-07-27 福特全球技术公司 Autonomous driving refined in virtual environments
CN104933708A (en) * 2015-06-07 2015-09-23 浙江大学 Barrier detection method in vegetation environment based on multispectral and 3D feature fusion
CN106627456A (en) * 2015-07-20 2017-05-10 国网河南省电力公司南阳供电公司 Tyre puncture prevention device
CN105550665A (en) * 2016-01-15 2016-05-04 北京理工大学 Method for detecting pilotless automobile through area based on binocular vision
CN106657926A (en) * 2017-01-23 2017-05-10 高艳云 Information transmission system based on intelligent terminal
CN206619301U (en) * 2017-03-06 2017-11-07 公安部交通管理科学研究所 Strengthen the safe driving simulated training educational system of special efficacy with environment
CN107284379A (en) * 2017-07-28 2017-10-24 合肥芯福传感器技术有限公司 AR optical projection systems and projecting method for vehicle safe driving
CN207466645U (en) * 2017-10-25 2018-06-08 李炳午 Safe driving assistant system
CN107871418A (en) * 2017-12-27 2018-04-03 吉林大学 It is a kind of to be used to evaluate the man-machine experiment porch for driving reliability altogether

Also Published As

Publication number Publication date
CN108875640A (en) 2018-11-23

Similar Documents

Publication Publication Date Title
Volk et al. Towards robust CNN-based object detection through augmentation with synthetic rain variations
Akagic et al. Pothole detection: An efficient vision based method using rgb color space image segmentation
EP3211596A1 (en) Generating a virtual world to assess real-world video analysis performance
KR101926561B1 (en) Road crack detection apparatus of patch unit and method thereof, and computer program for executing the same
CN103747240B (en) The vision significance filtering method of Fusion of Color and movable information
Wu et al. Densely pyramidal residual network for UAV-based railway images dehazing
US9286525B2 (en) Method for evaluating an object recognition device of a motor vehicle
WO2020096941A1 (en) Systems and methods for evaluating perception system quality
CN110737266A (en) automatic driving control method, device, vehicle and storage medium
JP6700373B2 (en) Apparatus and method for learning object image packaging for artificial intelligence of video animation
CN112990136B (en) Target detection method and device
Li et al. Photo-realistic simulation of road scene for data-driven methods in bad weather
Hospach et al. Simulation of falling rain for robustness testing of video-based surround sensing systems
CN104634740A (en) Monitoring method and monitoring device of haze visibility
WO2024060529A1 (en) Pavement disease recognition method and system, device, and storage medium
Talwar et al. Evaluating validity of synthetic data in perception tasks for autonomous vehicles
Halmaoui et al. Contrast restoration of road images taken in foggy weather
Jeon et al. Disc: A large-scale virtual dataset for simulating disaster scenarios
WO2020199057A1 (en) Self-piloting simulation system, method and device, and storage medium
CN113989394A (en) Image processing method and system for color temperature of automatic driving simulation environment
CN108875640B (en) Method for testing cognitive ability of passable area in end-to-end unsupervised scene
CN112597996A (en) Task-driven natural scene-based traffic sign significance detection method
WO2021146905A1 (en) Deep learning-based scene simulator construction method and apparatus, and computer device
Bu et al. Carla simulated data for rare road object detection
CN116665170A (en) Training of target detection model, target detection method, device, equipment and medium

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