WO2018099259A1 - Procédé et dispositif de détection d'obstacle destiné à un véhicule aérien sans pilote - Google Patents

Procédé et dispositif de détection d'obstacle destiné à un véhicule aérien sans pilote Download PDF

Info

Publication number
WO2018099259A1
WO2018099259A1 PCT/CN2017/110622 CN2017110622W WO2018099259A1 WO 2018099259 A1 WO2018099259 A1 WO 2018099259A1 CN 2017110622 W CN2017110622 W CN 2017110622W WO 2018099259 A1 WO2018099259 A1 WO 2018099259A1
Authority
WO
WIPO (PCT)
Prior art keywords
obstacle
texture
eye image
pixel
pixels
Prior art date
Application number
PCT/CN2017/110622
Other languages
English (en)
Chinese (zh)
Inventor
胡华智
尚黎民
孙海洋
Original Assignee
亿航智能设备（广州）有限公司
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 亿航智能设备（广州）有限公司 filed Critical 亿航智能设备（广州）有限公司
Publication of WO2018099259A1 publication Critical patent/WO2018099259A1/fr

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/10—Simultaneous control of position or course in three dimensions
    • G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft
    • G05D1/106—Change initiated in response to external conditions, e.g. avoidance of elevated terrain or of no-fly zones
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/10—Simultaneous control of position or course in three dimensions
    • G05D1/101—Simultaneous control of position or course in three dimensions specially adapted for aircraft
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G06T2207/10012—Stereo images

Definitions

  • the present invention relates to the field of drone technology, and in particular, to a method and device for detecting an obstacle of an unmanned aerial vehicle.
  • a stereo vision-based obstacle detection technique requires a block matching method to calculate a target depth in an image.
  • the specific detection method is to give a block of pixels in the left image and search for matching pixel blocks in the direction of the polar line in the right image.
  • the relative position of the matched pixel block in the right picture and the corresponding pixel block in the left picture, that is, the parallax is used to calculate the three-dimensional coordinates of the target in the pixel block.
  • Existing algorithms search for the target pixel block ⁇ in the right image, usually searching for a larger or even a full-length parallax range.
  • the size of the parallax corresponds to the distance between the obstacle and the camera. However, when the parallax is zero, the resolution of the stereo camera is not sufficient to calculate the target depth.
  • a primary object of the present invention is to provide a method and apparatus for detecting an obstacle of an unmanned aerial vehicle, which can effectively detect an obstacle with limited computing power.
  • the present invention provides a UAV obstacle detection method, which is applied to a short baseline stereo camera, wherein the stereo camera includes a left eye camera, and a left eye image of a specific position is acquired by the left eye camera.
  • the method includes: performing texture detection on a left eye image; marking a texture region in the left eye image according to a result of the texture detection; determining whether an obstacle pixel exists in the texture region; and if so, in the left eye camera
  • the specific location is an obstacle area.
  • the stereo camera further includes a right-eye camera corresponding to the left-eye camera, in the basis
  • the method further includes: acquiring the texture region in a first position of the left-eye image; searching in a right-eye image acquired by the right-eye camera a second position corresponding to the first position; matching a pixel of the second position.
  • the method further includes: setting a distance threshold according to a baseline length, a focal length, and a resolution of the stereo camera.
  • determining whether there is an obstacle pixel in the texture area comprising: determining whether the obstacle pixel is within a preset distance threshold; if yes, controlling the drone to change a route , so that the drone can avoid obstacles; if not, the drone does not need to change the route.
  • the method further comprises: marking a uniform region in the left-eye image, wherein the pixels in the uniform region are suspect pixels.
  • the marking the texture region and the uniform region in the left-eye image including: marking an area of the left-eye image with uniform grayscale as the uniform region, and having the image grayscale The area of the gradient is labeled as the texture area.
  • determining whether there is an obstacle pixel in the texture region comprising: determining whether a parallax of a pixel in the texture region and a pixel corresponding to the right eye image is zero; if yes, a pixel in the texture region is a pixel of zero parallax, and a plurality of pixels of the zero parallax constitute a free pixel; if not, a pixel in the texture region is a pixel of positive parallax, and a plurality of pixels of the positive parallax
  • the obstacle pixels are composed.
  • the method further includes: generating a path to avoid the obstacle area according to the open area corresponding to the free pixel.
  • the present invention further provides an unmanned aerial vehicle obstacle detecting device, which is applied to a short-baseline stereo camera, which includes a left-eye camera, and acquires a left-eye image of a specific position by the left-eye camera.
  • the device includes: a texture detecting module, configured to perform texture detection on the left eye image; a marking module, configured to mark a texture region in the left eye image according to the result of the texture detection; and a determining module, configured to determine the texture region Whether there is an obstacle pixel in the middle; if yes, the determining module determines that the obstacle area is at the specific position of the left-eye camera.
  • the device further includes: an acquiring module, configured to acquire a first location of the texture area in the left eye image; and a searching module, configured to search and search the right eye image acquired by the right eye camera The first a second position corresponding to the position; a matching module, configured to match the pixels of the second position.
  • the device further includes: a setting module, configured to set a distance threshold according to a baseline length, a focal length, and a resolution of the stereo camera.
  • a setting module configured to set a distance threshold according to a baseline length, a focal length, and a resolution of the stereo camera.
  • the determining module is specifically configured to: determine whether the obstacle pixel is within a preset distance threshold; if yes, control the drone to change a route, so that the drone avoids Barrier; if not, the drone does not need to change the route.
  • the marking module is further configured to: mark a uniform area in the left-eye image, where pixels in the uniform area are suspect pixels.
  • the marking module is specifically configured to: mark a region where the left eye image is uniform in gradation as the uniform region, and mark a region in which the image grayscale has a gradient as the texture region.
  • the determining module is further configured to: determine whether a disparity of a pixel in the texture region and a pixel corresponding to the right eye image is zero; if yes, a pixel in the texture region a pixel having zero parallax, a plurality of pixels of the zero parallax composing free pixels; if not, pixels in the texture region are pixels of positive parallax, and a plurality of pixels of the positive parallax constitute the obstacle pixel.
  • the device further includes: an evasive path generating module, configured to generate a path to avoid the obstacle area according to the open area corresponding to the free pixel.
  • an evasive path generating module configured to generate a path to avoid the obstacle area according to the open area corresponding to the free pixel.
  • the method and device for detecting an obstacle of an unmanned aerial vehicle provided by the present invention, by performing texture detection on a left eye image, marking a texture region in the left eye image according to a result of texture detection, and determining that there is an obstacle pixel in the texture region ⁇ , it is determined that the obstacle area at a specific position of the left-eye camera can improve the accuracy and system efficiency without affecting the effectiveness of the system.
  • 1 is a schematic structural view of a short baseline stereo camera
  • FIG. 2 is a schematic flow chart of a method for detecting an obstacle of an unmanned aerial vehicle according to a first embodiment of the present invention
  • FIG. 3 is a sub-flow diagram of a method for detecting an obstacle of an unmanned aerial vehicle according to a first embodiment of the present invention
  • FIG. 4 is a schematic flow chart of a method for detecting an obstacle of an unmanned aerial vehicle according to a second embodiment of the present invention
  • 5 is a schematic block diagram of a UAV obstacle detecting device according to a third embodiment of the present invention.
  • FIG. 6 is a schematic block diagram of an unmanned aerial vehicle obstacle detecting device according to a fourth embodiment of the present invention.
  • the unmanned aerial vehicle obstacle detection method and apparatus of the present invention are applied to a short baseline stereo camera.
  • Two short-baseline stereo cameras are mounted on the drone.
  • the short baseline stereo camera includes a left-eye camera 10, a right-eye camera 20, and a texture light emitter 30 between the left-eye camera 10 and the right-eye camera 20.
  • the texture light emitter 30 is used in an environment where texture is missing.
  • FIG. 2 is a schematic flowchart of a method for detecting an obstacle of an unmanned aerial vehicle according to a preferred embodiment of the present invention.
  • the UAV obstacle detection method of this embodiment includes the following steps:
  • Step 210 Perform texture detection on the left eye image.
  • a left-eye image of a specific position having a specific distance from the stereo camera is acquired by the left-eye camera.
  • one of the Laplacian operator, the Sobel operator, and the multi-level edge detection operator may be used to perform the left-eye image. Texture detection.
  • the light spot is emitted by the texture light emitter 30 to add texture to the uniform area, and then texture detection is performed.
  • texture detection For example, a uniform colorless area such as sky or wall, after detecting the texture, the texture light emitter 30 emits a light spot, that is, texture is added to the uniform color area, and texture detection is performed.
  • Step 220 Mark a texture region in the left eye image according to a result of the texture detection.
  • whether or not a texture region exists can be detected by calculating a gradient of the left eye image gradation. More specifically, a region in which the image gradation is uniform is marked as a uniform region, and a region in which the image gradation has a gradient is marked as a texture region.
  • the pixels in the uniform region are suspected pixels, and the plurality of suspect pixels constitute the suspected region.
  • Step 230 Determine whether there is an obstacle pixel in the texture area. If yes, go to step 240, if no
  • step 250 proceeds to step 250.
  • a zero parallax algorithm is performed on each pixel in the texture region to determine whether there is an obstacle pixel in the texture region of the left eye image.
  • the specific steps are: comparing each pixel ( ⁇ ', ⁇ ') in the texture area with a corresponding pixel ( ⁇ , ⁇ ) in the right-eye image taken by the right-eye camera, if the pixel in the texture area ( ⁇ ' , ⁇ ') has a parallax corresponding to the corresponding pixel ( ⁇ , ⁇ ) in the right eye image, and the pixel in the texture region ( ⁇ ', ⁇ ') is determined as an obstacle pixel if the parallax is not equal to zero; otherwise, the texture region is judged
  • the pixels ( ⁇ ', ⁇ ') are free pixels.
  • Step 240 is an obstacle area at a specific position of the left-eye camera.
  • Step 250 is an open area at a specific location of the left eye camera.
  • the method may further include the following steps:
  • Step 310 Determine whether the obstacle pixel is within a preset distance threshold; if yes, proceed to step 320, and if no, proceed to step 330.
  • Step 320 Control the drone to change the route, so that the drone performs obstacle avoidance.
  • Step 330 The drone does not need to change the route.
  • the result of the zero parallax algorithm it is divided into: a pixel of zero parallax and a pixel of positive parallax, and combined with the result of the zero parallax algorithm to determine whether to control the Unmanned The machine is used to avoid obstacles.
  • the plurality of pixels of the positive parallax constitute the obstacle pixel
  • the plurality of pixels of the zero parallax constitute a free pixel.
  • the pixel of zero parallax represents that the distance from the point to the camera is greater than the threshold, that is, the point between the point and the camera is considered to be an empty area, so that the drone does not need to control the drone to change the route.
  • the pixel of the positive parallax represents that the distance from the point to the camera is less than the threshold, i.e., the distance between the point and the camera is considered to be an obstacle, and the drone is controlled to change the route to make the drone obstacle avoidance.
  • the UAV obstacle detection method may further include the steps of:
  • a path avoiding the obstacle region is generated according to the open area corresponding to the free pixel.
  • the path of the obstacle avoidance area is reset.
  • a suitable area is generated as a path for the drone to fly.
  • the unmanned obstacle detection method by performing texture detection on the left eye image, marking the texture region in the left eye image according to the result of the texture detection, and determining that there is an obstacle pixel in the texture region, Determining the obstacle area at a specific position of the left-eye camera improves accuracy and system efficiency without affecting the effectiveness of the system.
  • a method for detecting an obstacle of an unmanned aerial vehicle includes:
  • Step 410 Perform texture detection on the left eye image.
  • Step 420 Mark a texture region in the left eye image according to the result of the texture detection.
  • Step 430 Acquire a first position of the texture area in the left eye image.
  • Step 440 Search for a second location corresponding to the first location in the right eye image acquired by the right eye camera.
  • Step 450 Match pixels of the second location.
  • the pixel in which the texture region exists is acquired at the first position corresponding to the left eye image. After matching with the right eye image acquired by the right eye camera, it is only necessary to find the second position of the first position on the corresponding right eye image and match the pixel at the second position.
  • Step 460 Determine whether there is an obstacle pixel in the texture area. If yes, go to step 470. If no, go to step 480.
  • Step 470 where the specific position of the left-eye camera is an obstacle area.
  • Step 480 An open area at the specific location of the left-eye camera.
  • the unmanned obstacle detection method of the present embodiment further improves the accuracy and system efficiency of detecting an obstacle by matching the texture region in the left-eye image to the corresponding position in the right-eye image.
  • another preferred embodiment of the present invention provides a module schematic of an unmanned aerial vehicle obstacle detecting device.
  • the UAV obstacle detecting device can perform the method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the executing method.
  • the UAV obstacle detecting apparatus includes:
  • the texture detecting module 510 is configured to perform texture detection on the left eye image.
  • a left-eye image of a specific position having a specific distance from the stereo camera is acquired by the left-eye camera.
  • the texture detecting module 510 may perform texture detection on the left eye image by using one of a Laplacian operator, a Sobel operator, and a Canny edge detection operator.
  • the texture detecting module 510 performs texture detection. For example, a uniform colorless area such as sky or wall, after detecting the texture, the texture light emitter 30 emits a light spot, that is, texture is added to the uniform color area, and texture detection is performed.
  • the marking module 520 is configured to mark the texture region in the left eye image according to the result of the texture detection. [0082] Specifically, the marking module 520 marks the texture region and the uniform region in the left eye image based on the result of the texture detection. More specifically, a region in which the image gradation is uniform is marked as a uniform region, and a region in which the image gradation has a gradient is marked as a texture region.
  • the pixels in the uniform region are suspected pixels, and the plurality of suspect pixels constitute the suspected region.
  • the determining module 530 is configured to determine whether an obstacle pixel exists in the texture area. Specifically, a zero parallax algorithm is performed on each pixel in the texture region to determine whether or not an obstacle pixel exists in the texture region of the left eye image.
  • the determining module 530 is specifically configured to compare each pixel ( ⁇ ', ⁇ ') in the texture region with a corresponding pixel ( ⁇ , ⁇ ) in the right-eye image captured by the right-eye camera, if the pixel in the texture region ( ⁇ ', ⁇ ') has a parallax corresponding to the corresponding pixel ( ⁇ , ⁇ ) in the right eye image, and the pixel in the texture region ( ⁇ ', ⁇ ') is determined as an obstacle pixel if the parallax is not equal to zero;
  • the pixels in the area ( ⁇ ', ⁇ ') are free pixels.
  • the judging module 530 determines that there is an obstacle pixel in the texture area, it is an obstacle area at a specific position of the left-eye camera.
  • the determining module 530 determines that there is an obstacle area in the texture area, it indicates that there is an obstacle at a specific position from the specific distance of the stereo camera, that is, along the flight path of the drone, the distance drone There are obstacles at a certain distance.
  • the judging module 530 determines that there is no obstacle pixel in the texture region, it is an empty region at a specific position of the left-eye camera.
  • the determining module 530 determines that there is no obstacle area in the texture area, that is, an open area, it indicates that there is no obstacle at a specific position from the specific distance of the stereo camera, that is, along the drone.
  • the flight path there is no obstacle at a certain distance from the drone, and the drone can continue to fly along the current flight path.
  • the determining module 530 is further configured to: determine whether the obstacle pixel is within a preset distance threshold; if yes, control the drone to change the route, so that the drone performs obstacle avoidance; , the drone does not need to change the route.
  • the determining module 530 divides the pixel into a pixel of zero parallax and a pixel of positive parallax according to the result of the zero parallax algorithm, and combines the result of the zero parallax algorithm to determine whether to control the drone to avoid barrier .
  • the plurality of pixels of the positive parallax constitute the obstacle pixel
  • the plurality of pixels of the zero parallax constitute a free pixel.
  • the UAV obstacle detecting device may further include:
  • a setting module 540 configured to set a distance threshold according to a baseline length, a focal length, and a resolution of the stereo camera
  • a threshold ie, a threshold
  • the threshold is the upper limit of the detection distance.
  • a pixel of zero parallax represents that the distance from the point to the camera is greater than the threshold, ie, the point between the point and the camera is considered to be an empty area, and thus the drone does not need to control the drone to change the route.
  • the pixel of the positive parallax represents that the distance from the point to the camera is less than the threshold, that is, the distance from the point to the camera is considered to be an obstacle, and the drone is controlled to change the route so that the drone can avoid obstacles.
  • the UAV obstacle detecting device may further include:
  • the evasive path generation module 550 is configured to generate a path for avoiding the obstacle region according to the open area corresponding to the free pixel.
  • the avoidance path generation module 550 when detecting the presence of an obstacle, the avoidance path generation module 550 resets the path of the avoidance obstacle area. In the open area corresponding to the free pixel, a suitable area is generated as a path for the drone to fly.
  • the avoidance path generation module 550 regenerates the path ⁇ , and needs to comprehensively consider information such as the deviation of the open area, the posture of the drone, and the like.
  • the UAV obstacle detecting device performs texture detection on the left eye image by the texture detecting module 510, and the marking module 520 marks the texture region in the left eye image according to the result of the texture detection, when the determining module 530 determines There is an obstacle pixel in the texture area, which is an obstacle area at a specific position of the left-eye camera, which can improve the accuracy and system efficiency without affecting the effectiveness of the system.
  • FIG. 6 is a block diagram of a UAV obstacle detecting device according to another preferred embodiment of the present invention.
  • the UAV obstacle detecting device is a further improvement based on the third embodiment, except that the device further includes: [0102]
  • the obtaining module 610 is configured to acquire a first location of the texture area in the left-eye image.
  • the searching module 620 is configured to search for a second location corresponding to the first location in the right eye image acquired by the right eye camera.
  • the matching module 630 is configured to match pixels of the second location.
  • the acquiring module 610 acquires the first position corresponding to the left-eye image of the pixel in which the texture region exists. After matching with the right eye image acquired by the right eye camera, the lookup module 620 only needs to find the first position on the corresponding right eye image and match the pixel at the second position through the matching module 630.
  • the UAV obstacle detecting device of the present embodiment further improves the accuracy and system efficiency of detecting an obstacle by matching the texture region in the left-eye image to the corresponding position in the right-eye image.
  • the method and device for detecting an obstacle of an unmanned aerial vehicle provided by the present invention, by performing texture detection on a left-eye image, marking a texture region in the left-eye image according to a result of texture detection, and determining that an obstacle pixel exists in the texture region ⁇ , it is determined that the obstacle area at a specific position of the left-eye camera can improve the accuracy and system efficiency without affecting the effectiveness of the system.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Automation & Control Theory (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Analysis (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Image Processing (AREA)

Abstract

La présente invention concerne le domaine technique des véhicules aériens sans pilote. L'invention concerne un procédé et un dispositif de détection d'obstacle pour un véhicule aérien sans pilote, applicables dans une caméra stéréoscopique à ligne de base courte. La caméra stéréoscopique comprend une caméra d'œil gauche ; une image d'œil gauche d'un emplacement spécifique est acquise par l'intermédiaire de la caméra d'œil gauche. Le procédé consiste : à effectuer une détection de texture par rapport à une image d'œil gauche ; à marquer une zone texturée dans l'image d'œil gauche sur la base du résultat de la détection de texture ; à déterminer si des pixels d'obstacle sont présents dans la zone texturée ; et si tel est le cas, alors l'emplacement spécifique pour la caméra d'œil gauche est une zone d'obstacle. Le procédé et le dispositif décrits dans la présente invention pour la détection d'obstacles pour le véhicule aérien sans pilote augmentent la précision et l'efficacité du système sans affecter l'efficacité du système.
PCT/CN2017/110622 2016-12-01 2017-11-12 Procédé et dispositif de détection d'obstacle destiné à un véhicule aérien sans pilote WO2018099259A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201611088538.8 2016-12-01
CN201611088538.8A CN106682584B (zh) 2016-12-01 2016-12-01 无人机障碍物检测方法及装置

Publications (1)

Publication Number Publication Date
WO2018099259A1 true WO2018099259A1 (fr) 2018-06-07

Family

ID=58866472

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/110622 WO2018099259A1 (fr) 2016-12-01 2017-11-12 Procédé et dispositif de détection d'obstacle destiné à un véhicule aérien sans pilote

Country Status (2)

Country Link
CN (1) CN106682584B (fr)
WO (1) WO2018099259A1 (fr)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106682584B (zh) * 2016-12-01 2019-12-20 广州亿航智能技术有限公司 无人机障碍物检测方法及装置
CN108985193A (zh) * 2018-06-28 2018-12-11 电子科技大学 一种基于图像检测的无人机航拍人像对准方法
CN110110702A (zh) * 2019-05-20 2019-08-09 哈尔滨理工大学 一种基于改进ssd目标检测网络的无人机规避算法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102385237A (zh) * 2010-09-08 2012-03-21 微软公司 基于结构化光和立体视觉的深度相机
CN102435174A (zh) * 2011-11-01 2012-05-02 清华大学 基于混合式双目视觉的障碍物检测方法及装置
US20160054737A1 (en) * 2014-08-22 2016-02-25 Cape Productions Inc. Methods and Apparatus for Unmanned Aerial Vehicle Autonomous Aviation
CN106127788A (zh) * 2016-07-04 2016-11-16 触景无限科技（北京）有限公司 一种视觉避障方法和装置
CN106682584A (zh) * 2016-12-01 2017-05-17 广州亿航智能技术有限公司 无人机障碍物检测方法及装置

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104021388B (zh) * 2014-05-14 2017-08-22 西安理工大学 基于双目视觉的倒车障碍物自动检测及预警方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102385237A (zh) * 2010-09-08 2012-03-21 微软公司 基于结构化光和立体视觉的深度相机
CN102435174A (zh) * 2011-11-01 2012-05-02 清华大学 基于混合式双目视觉的障碍物检测方法及装置
US20160054737A1 (en) * 2014-08-22 2016-02-25 Cape Productions Inc. Methods and Apparatus for Unmanned Aerial Vehicle Autonomous Aviation
CN106127788A (zh) * 2016-07-04 2016-11-16 触景无限科技（北京）有限公司 一种视觉避障方法和装置
CN106682584A (zh) * 2016-12-01 2017-05-17 广州亿航智能技术有限公司 无人机障碍物检测方法及装置

Also Published As

Publication number Publication date
CN106682584A (zh) 2017-05-17
CN106682584B (zh) 2019-12-20

Similar Documents

Publication Publication Date Title
US10937186B2 (en) Techniques for precisely locating landmarks in monocular camera images with deep learning
US10129521B2 (en) Depth sensing method and system for autonomous vehicles
US11151741B2 (en) System and method for obstacle avoidance
US9330472B2 (en) System and method for distorted camera image correction
US9672635B2 (en) Method for analyzing related images, image processing system, vehicle comprising such system and computer program product
EP3508936B1 (fr) Procédé et appareil d'évitement d'obstacle, objet mobile et support d'informations lisible par ordinateur
WO2019015158A1 (fr) Procédé d'évitement d'obstacle pour un véhicule aérien sans pilote et véhicule aérien sans pilote
JP2018522348A (ja) センサーの３次元姿勢を推定する方法及びシステム
US10929986B2 (en) Techniques for using a simple neural network model and standard camera for image detection in autonomous driving
JP2006252473A (ja) 障害物検出装置、キャリブレーション装置、キャリブレーション方法およびキャリブレーションプログラム
WO2018099259A1 (fr) Procédé et dispositif de détection d'obstacle destiné à un véhicule aérien sans pilote
CN105844692A (zh) 基于双目立体视觉的三维重建装置、方法、系统及无人机
KR20200071960A (ko) 카메라 및 라이다 센서 융합을 이용한 객체 검출 방법 및 그를 위한 장치
JP2006262009A (ja) 移動体用の障害物検出装置
US20220157067A1 (en) Determining a source of danger on a roadway
JP6455164B2 (ja) 視差値導出装置、機器制御システム、移動体、ロボット、視差値導出方法、およびプログラム
US11423560B2 (en) Method for improving the interpretation of the surroundings of a vehicle
JP6577595B2 (ja) 車両用外界認識装置
CN108733076B (zh) 一种无人机抓取目标物体的方法、装置及电子设备
KR102048999B1 (ko) 자율 주행 장치 및 방법
JP2020064029A (ja) 移動体制御装置
JP6668740B2 (ja) 道路面推定装置
KR20190103524A (ko) 자율 주행 장치 및 방법
JP7169689B2 (ja) 計測システム、計測方法、及び計測プログラム
JP2017116363A (ja) 車両位置推定システム、方法およびプログラム

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17875590

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 27/09/2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17875590

Country of ref document: EP

Kind code of ref document: A1