CN110136186A - A kind of detection target matching method for mobile robot object ranging - Google Patents
A kind of detection target matching method for mobile robot object ranging Download PDFInfo
- Publication number
- CN110136186A CN110136186A CN201910389574.5A CN201910389574A CN110136186A CN 110136186 A CN110136186 A CN 110136186A CN 201910389574 A CN201910389574 A CN 201910389574A CN 110136186 A CN110136186 A CN 110136186A
- Authority
- CN
- China
- Prior art keywords
- detection target
- target
- detection
- deep learning
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C3/00—Measuring distances in line of sight; Optical rangefinders
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The present invention relates to a kind of detection target matching methods for mobile robot object ranging, this method first uses binocular camera to acquire image, and with trained complete deep learning training pattern to a left side, right image is detected, a detection target is randomly choosed in left image, and category identification is carried out with the detection target in right image, if recognizing, then determination is same object in real space, and the depth distance of the object is calculated with parallax method, then judge whether the detection target is the last one detection target, if not, next detection target is taken to repeat the above steps, if unidentified, then jump directly to judgment step.The present invention can not only realize that binocular camera obtains image and carries out the Auto-matching for detecting target in left image and right image after object detection process, and it is also equipped with the rapidity of deep learning target detection, it can be realized Auto-matching detection target, detect target depth convenient for processing equipment automatic measurement.
Description
Technical field
The present invention relates to robot field, specially a kind of detection object matching side for mobile robot object ranging
Method.
Background technique
In recent years, with the rapid development of science and technology, mobile robot have been applied to we produce, life it is each
Field becomes one of most active and most potential technology.Wherein, the detection of barrier is mobile robot research
One hot issue.In order to realize navigation feature, robot needs to detect before avoidance and hinders the object travelled in environment
Relative distance between body.Currently, common detection method has: ultrasonic distance measurement, laser pulse ranging, infrared distance measurement, optics
Ranging and stereoscopy passive ranging etc..The devices such as ultrasound, laser, infrared pass through the time between the transmitting of measurement emission source and return
Difference calculates the distance between testee and sensor, referred to as active method.Active method ranging is more convenient, rapid, calculates
It is relatively simple, therefore be widely used in real-time control.But transmitting and receiving apparatus is expensive, cost compared with
Height, and the environmental problems such as reflection, noise, intersection are difficult to avoid, and active method ranging does not have general applicability.In contrast,
Since visual sensor has many advantages, such as abundant information, detection range is wide, barrier is especially realized in Mobile Robotics Navigation
Hinder in analyte detection using more and more extensive.
For a long time simultaneously, based on the target detection of deep learning in industrial production, autonomous driving, video monitoring, figure
The picture multiple fields such as retrieval, human-computer interaction suffer from extensive use, and the target detection technique based on deep learning can be efficient
Carry out realtime graphic target detection.And binocular vision can obtain the depth of match point in left images by parallax method.But
It is when mobile robot detects barrier in actual engineer application, not need to measure any point on barrier
Depth distance information, either which kind of distance measuring method can not all measure the depth information at any point on barrier comprehensively, because
This needs to measure the depth distance information at representative position on barrier, so that it may be the position of mobile robot disturbance in judgement object
Offer reference information is provided.Therefore it can be by the entire obstacle of conduct for the detection target identification frame central point that deep learning algorithm detects
The representative position of object, and then the identification frame by matching same detection target in left and right image, can be calculated using parallax method
Detect target identification frame depth to center.How to be matched to the identification frame of same detection target in left and right image is exactly one
A critical problem, therefore, the invention discloses a kind of detection target matching methods for mobile robot object ranging.
Patent [CN109544633A;201710867746.6] a kind of target ranging method is disclosed, this method utilizes list
Mesh camera shoots to obtain traffic target, internal reference and outer ginseng by reading monocular camera, and reads traffic from preset standard
Then the size of target calculates traffic target at a distance from datum target.The patented method does not refer to how traffic target selects
It takes and determines, if new image, object ranging under multiple images target conditions is simultaneously not suitable for for new scene.
Patent [CN109212540A;201811062793.4] disclose a kind of ranging side based on laser radar system
Method, the ranging data that multiple lidar measurements by receiving laser radar system obtain, establishes three-dimensional coordinate model, determines
The distance between the unmanned equipment of target and each barrier, compared with the existing technology in pass through single laser radar carry out vision
, it can be achieved that the bigger ranging calibration tasks of distance range for the mode of the ranging calibration of image algorithm, but can not be to single
Barrier carries out category identification, can not also be distinguished each barrier, be only able to detect barrier.
Patent [CN109029363A;201810562681.9] disclose a kind of object ranging side based on deep learning
Method establishes the target database under different distance, builds object ranging model, the loss function of design object ranging model, if
The training method for counting object ranging model, tests trained object ranging model.Recurrence is converted by object ranging problem to ask
Topic incorporates algorithm of target detection model, to realize target detection and object ranging in an algorithm model.But it is special at this
It is mentioned in sharp claim 3 and the single-frame images of acquisition is manually demarcated using calibration tool, calibration content includes: detected
The bounding box center point coordinate (x, y) of target, the bounding box width w and height h that are detected target, the bounding box for being detected target
The range information L of classification information c and detected target range video camera.The content needs of above-mentioned calibration are accomplished manually, for institute
If state target database data volume it is very big when, artificial calibration be it is unpractical, ImageNet is to the mark of image data set
Task amount is big to be just enough to illustrate, the artificial foundation of database is to consume very much manpower, object in terms of deep learning and artificial intelligence
Power, financial resources;Meanwhile the patent of invention obtains image progress depth information calculating using common monocular-camera and is inherently not so good as
It is more accurate that binocular camera calculates depth information.
Summary of the invention
In order to solve the problems, such as to be previously mentioned in above-mentioned background technique, the invention proposes one kind to be used for mobile robot target
The detection target matching method of ranging.
The technical problems to be solved by the invention are realized using following technical scheme:
A kind of detection target matching method for mobile robot object ranging, includes the following steps:
(1) image is acquired using binocular camera;
(2) establish and training deep learning training pattern, according to train completion deep learning training pattern to left and right
Image carries out target detection simultaneously;
(3) obtaining detection target in left image is A1, A2, A3 ..., and detection target is B1 in An-1, An and right image,
B2, B3 ..., Bm-1, Bm randomly select a detection target in left image, are denoted as Ai;
(4) will test the detection target B1, B2, B3 ... in target Ai and right image, Bm-1, Bm carry out category identification, if
It is unidentified to detection target Ai type identical detection target when, jump directly in step (7);If recognizing and detection mesh
When marking the identical detection target of Ai type, progress step (5);
(5) detection target identical with detection target Ai type is one or more, if detection target is one,
Direct successful match, when if more than one, by comparing the bounding box center point coordinate and each detection mesh for detecting target Ai
The distance between target bounding box center point coordinate selects the distance detection closer inspection of target Ai bounding box center point coordinate
Target is surveyed, Bj will be denoted as with the detection matched detection target of target Ai;
(6) the detection target Ai in left image determines detection target with after the detection target Bj successful match in right image
Ai and detection target Bj is the same object in real space, is denoted as Ci, and allow and calculate object Ci by parallax method
Depth distance, be denoted as Zi;
(7) whether i is more than or equal to n at this time for judgement, if i is more than or equal to n, terminates this algorithm flow, is equal to if i is not more than
I+1 is then assigned to i by n, is jumped in (4) and is continued to execute again, until i stops when being more than or equal to n.
I initial value in step (3) in detection target Ai is 1, and is positive integer.
As a further improvement of the present invention, in step (2) deep learning training pattern to establish training step as follows:
(1) image data acquired in collection step (1) constitutes image data set;
(2) the deep learning frame of prior art maturation is chosen;
(3) using the image data set in the deep learning frame training step (1) chosen;
(4) the deep learning training pattern for being used for processes captured image data is obtained.
Improved as another kind of the invention, in step (2) deep learning training pattern to establish training step as follows:
(A) image data acquired in collection step (1) constitutes image data set;
(B) deep learning neural network framework is established;
(C) deep learning neural network framework is trained in advance, obtains the deep learning frame of oneself;
(D) using the image data set trained in the deep learning frame training step (1) completed;
(E) the deep learning training pattern for being used for processes captured image data is obtained.
The beneficial effects of the present invention are:
The present invention can not only realize that binocular camera obtains image and carries out the left image after object detection process and the right side
The Auto-matching of target is detected in image, and is also equipped with the rapidity of deep learning target detection, can be realized Auto-matching
Target is detected, detects target depth convenient for processing equipment automatic measurement.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples:
Fig. 1 is flow diagram of the invention;
Fig. 2 is deep learning training pattern Establishing process figure in the present invention;
Fig. 3 is the original image schematic diagram that binocular camera obtains in the present invention;
Fig. 4 is object detection results schematic diagram in the present invention;
Fig. 5 is that left image detects object delineation in the present invention;
Fig. 6 is that right image detects object delineation in the present invention.
Specific embodiment
In order to be easy to understand the technical means, the creative features, the aims and the efficiencies achieved by the present invention, tie below
It closes attached drawing and the present invention is further described in embodiment.
A kind of detection target matching method for mobile robot object ranging, includes the following steps:
(1) binocular camera is connected first with laptop, notebook is mounted in mobile robot, then camera
The realtime graphic in front of robot is acquired, as shown in Figure 3.
(2) establish and training deep learning training pattern, according to train completion deep learning training pattern to left and right
Image carries out target detection simultaneously.
There are two types of the modes of the deep learning training pattern, specific as follows:
Mode one:
(1) image data acquired in collection step (1) constitutes image data set;
(2) the deep learning frame of prior art maturation is chosen;
(3) using the image data set in the deep learning frame training step (1) chosen;
(4) the deep learning training pattern for being used for processes captured image data is obtained.
Mode two:
(A) image data acquired in collection step (1) constitutes image data set;
(B) deep learning neural network framework is established;
(C) deep learning neural network framework is trained in advance, obtains the deep learning frame of oneself;
(D) using the image data set trained in the deep learning frame training step (1) completed;
(E) the deep learning training pattern for being used for processes captured image data is obtained.
When specifically used, it can choose any one kind of them according to the actual situation.
(3) obtaining detection target in left image is A1, A2, A3, and detection target is B1, B2, B3, B4 in A4 and right image,
A detection target in left image is randomly selected, Ai, Schilling i=1 are denoted as.
(4) it will test target A1, with detection target B1, B2, B3, the B4 in right image, category identification carried out, if unidentified
It when to the detection identical detection target of target A1 type, jumps directly in step (7), if recognizing and detecting target A1 kind
When the identical detection target of class, carry out step (7).
(5) as shown in Figure 5 and Figure 6, recognize with survey target A1 type it is identical there are two detect target B1, B3, therefore
Need to compare the distance between detection target A1 bounding box center point coordinate and detection target B1, B3 bounding box center point coordinate,
And select the distance detection closer detection target of target A1.
As shown in Figure 5 and Figure 6, A1 bounding box center point coordinate (430,381) are found and right image detects target B1, the side B3
Boundary's frame center point coordinate (416.5,378.5), the distance between (623,348) d, d | A1-B1 |=13.73, d | A1-B3 |=
204.68, d | A1-B1 | < d | A1-B3 |, therefore B1 bounding box central point distance A1 bounding box central point is closer, therefore in left image
It detects target A1 and right image detects target B1 successful match.
(6) after detection target A1 and detection target B1 successful match, detection target A1 and the right figure in left image are determined
It is the same object C1 in real space as detecting target B1, the depth distance z1 of object C1 is calculated by parallax method, this
Binocular camera baseline length b in embodimentlineFor 120mm, calibrated focal length f is 731px, according to formula z=f*bline/ diff,
Middle z is Object Depth distance, and diff is parallax, i.e., the horizontal coordinate difference of match point in binocular image, and parallax diff is positive
It counts, the parallax diff of detection target A1 central point and right image detection target B1 central point in the left image is 430-
416.5=13.5 therefore z1=6497.78mm can be obtained.
(7) whether i is more than or equal to n at this time for judgement, if i is more than or equal to n, terminates this algorithm flow;If i is not more than and is equal to
N, then be assigned to i for i+1, jumps in step (4) and continues to execute again, until i stops when being more than or equal to n.
I=1 at this time, n=4, i, which are not more than, is equal to n, therefore i+1 is then assigned to i, this stylish i is 2, and algorithm flow jumps
It is continued to execute to step (4).
(8) as shown in Figure 5 and Figure 6, only one detection target B2, therefore B2 identical with detection target A2 type are recognized
Bounding box central point distance A2 bounding box central point it is closer, therefore detect target A2 and right image in left image and detect target B2
Successful match.
(9) identical as step (6), determine that detection target A2 and right image detection the target B2 in left image are in practical sky
Between in same object C2, z2=3189.82mm is calculated by parallax method.
(10) step (7) are repeated, at this time i=2, n=4, i, which are not more than, is equal to n, therefore i+1 is then assigned to i, this stylish i
It is 3, algorithm flow jumps to step (4) and continues to execute.
(11) as shown in Figure 5 and Figure 6, recognize with detection target A3 type it is identical there are two detect target B1 and B3.
As shown in Figure 5 and Figure 6, A3 bounding box center point coordinate (649.5,344.5) are found and right image detects target B1,
B3 bounding box center point coordinate (416.5,378.5), the distance between (632,348) d, d | A3-B1 |=235.47, d | A3-B2
|=17.85, d | A3-B3 | < d | A3-B1 |, therefore B3 bounding box central point distance A3 bounding box central point is closer, therefore left image
Middle detection target A3 and right image detect target B3 successful match.
(12) identical as step (6), determine that detection target A3 and right image detection the target B3 in left image are in reality
Same object C3 in space, calculates z3=5012.57mm by parallax method.
(13) step (7) are repeated, at this time i=3, n=4, i, which are not more than, is equal to n, therefore i+1 is then assigned to i, this stylish i
It is 4, algorithm flow jumps to step (4) and continues to execute.
(14) as shown in Figure 5 and Figure 6, only one detection target B4 identical with detection target A4 type is recognized, therefore
The bounding box central point distance A4 bounding box central point of B4 is closer, therefore target A4 and right image detection target are detected in left image
B4 successful match.
(15) identical as step (6), determine that detection target A4 and right image detection the target B4 in left image are in reality
Same object C4 in space, calculates z4=17544mm by parallax method.
(16) step (7) are repeated, at this time i=4, n=4, i are equal to n, then terminate this algorithm flow.
It it should be noted that the initial value of the i is 1, and is positive integer;In addition, a bounding box in above-described embodiment
Center point coordinate value is the coordinate value as unit of pixel p x, and coordinate origin is in the upper left corner of Fig. 5 and Fig. 6.
The basic principles, main features and advantages of the present invention have been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and what is described in the above embodiment and the description is only the present invention
Principle, without departing from the spirit and scope of the present invention, various changes and improvements may be made to the invention, these variation and
Improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appended claims and its equivalent
Object defines.
Claims (4)
1. a kind of detection target matching method for mobile robot object ranging, characterized by the following steps:
(1) image is acquired using binocular camera;
(2) establish and training deep learning training pattern, according to train completion deep learning training pattern to left and right image
Target detection is carried out simultaneously;
(3) obtaining detection target in left image is A1, A2, A3 ..., and detection target is B1, B2 in An-1, An and right image,
B3 ..., Bm-1, Bm randomly select a detection target in left image, are denoted as Ai;
(4) detection the target B1, B2, B3 ... in target Ai and right image be will test, Bm-1, Bm carry out category identification, if not knowing
It when being clipped to detection target identical with detection target Ai type, jumps directly in step (7), if recognizing and detection target Ai
When the identical detection target of type, carry out step (5);
(5) detection target identical with detection target Ai type is one or more, if detection target is one, directly
Successful match, when if more than one, by comparing the bounding box center point coordinate and each detection target for detecting target Ai
The distance between bounding box center point coordinate selects the distance detection closer detection mesh of target Ai bounding box center point coordinate
Mark will be denoted as Bj with the detection matched detection target of target Ai;
(6) the detection target Ai in the left image determines detection target with after the detection target Bj successful match in right image
Ai and detection target Bj is the same object in real space, is denoted as Ci, and allow and calculate object Ci by parallax method
Depth distance, be denoted as Zi;
(7) whether i is more than or equal to n at this time for judgement, if i is more than or equal to n, terminates this algorithm flow, if i, which is not more than, is equal to n,
I+1 is assigned to i, jumps in (4) and continues to execute again, until i stops when being more than or equal to n.
2. a kind of detection target matching method for mobile robot object ranging according to claim 1, feature
Be: the i initial value in the step (3) in detection target Ai is 1, and is positive integer.
3. a kind of detection target matching method for mobile robot object ranging according to claim 1, feature
Be: in the step (2) deep learning training pattern to establish training step as follows:
(1) image data acquired in collection step (1) constitutes image data set;
(2) the deep learning frame of prior art maturation is chosen;
(3) using the image data set in the deep learning frame training step (1) chosen;
(4) the deep learning training pattern for being used for processes captured image data is obtained.
4. a kind of detection target matching method for mobile robot object ranging according to claim 1, feature
Be: in the step (2) deep learning training pattern to establish training step as follows:
(A) image data acquired in collection step (1) constitutes image data set;
(B) deep learning neural network framework is established;
(C) deep learning neural network framework is trained in advance, obtains the deep learning frame of oneself;
(D) using the image data set trained in the deep learning frame training step (1) completed;
(E) the deep learning training pattern for being used for processes captured image data is obtained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910389574.5A CN110136186B (en) | 2019-05-10 | 2019-05-10 | Detection target matching method for mobile robot target ranging |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910389574.5A CN110136186B (en) | 2019-05-10 | 2019-05-10 | Detection target matching method for mobile robot target ranging |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110136186A true CN110136186A (en) | 2019-08-16 |
CN110136186B CN110136186B (en) | 2022-09-16 |
Family
ID=67573358
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910389574.5A Active CN110136186B (en) | 2019-05-10 | 2019-05-10 | Detection target matching method for mobile robot target ranging |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110136186B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112489186A (en) * | 2020-10-28 | 2021-03-12 | 中汽数据(天津)有限公司 | Automatic driving binocular data perception algorithm |
CN114018268A (en) * | 2021-11-05 | 2022-02-08 | 上海景吾智能科技有限公司 | Indoor mobile robot navigation method |
CN114049394A (en) * | 2021-11-23 | 2022-02-15 | 智道网联科技(北京)有限公司 | Monocular distance measuring method, device, equipment and storage medium |
WO2022041737A1 (en) * | 2020-08-28 | 2022-03-03 | 北京石头世纪科技股份有限公司 | Distance measuring method and apparatus, robot, and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101516040A (en) * | 2008-02-20 | 2009-08-26 | 深圳华为通信技术有限公司 | Video matching method, device and system |
WO2018086348A1 (en) * | 2016-11-09 | 2018-05-17 | 人加智能机器人技术(北京)有限公司 | Binocular stereo vision system and depth measurement method |
WO2018098915A1 (en) * | 2016-11-29 | 2018-06-07 | 深圳市元征科技股份有限公司 | Control method of cleaning robot, and cleaning robot |
CN109029363A (en) * | 2018-06-04 | 2018-12-18 | 泉州装备制造研究所 | A kind of target ranging method based on deep learning |
CN109084724A (en) * | 2018-07-06 | 2018-12-25 | 西安理工大学 | A kind of deep learning barrier distance measuring method based on binocular vision |
-
2019
- 2019-05-10 CN CN201910389574.5A patent/CN110136186B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101516040A (en) * | 2008-02-20 | 2009-08-26 | 深圳华为通信技术有限公司 | Video matching method, device and system |
WO2018086348A1 (en) * | 2016-11-09 | 2018-05-17 | 人加智能机器人技术(北京)有限公司 | Binocular stereo vision system and depth measurement method |
WO2018098915A1 (en) * | 2016-11-29 | 2018-06-07 | 深圳市元征科技股份有限公司 | Control method of cleaning robot, and cleaning robot |
CN109029363A (en) * | 2018-06-04 | 2018-12-18 | 泉州装备制造研究所 | A kind of target ranging method based on deep learning |
CN109084724A (en) * | 2018-07-06 | 2018-12-25 | 西安理工大学 | A kind of deep learning barrier distance measuring method based on binocular vision |
Non-Patent Citations (2)
Title |
---|
张琦等: "改进Fast-RCNN的双目视觉车辆检测方法", 《应用光学》 * |
潘琪等: "基于双目视觉的工件测量研究", 《智能计算机与应用》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022041737A1 (en) * | 2020-08-28 | 2022-03-03 | 北京石头世纪科技股份有限公司 | Distance measuring method and apparatus, robot, and storage medium |
CN112489186A (en) * | 2020-10-28 | 2021-03-12 | 中汽数据(天津)有限公司 | Automatic driving binocular data perception algorithm |
CN112489186B (en) * | 2020-10-28 | 2023-06-27 | 中汽数据(天津)有限公司 | Automatic driving binocular data sensing method |
CN114018268A (en) * | 2021-11-05 | 2022-02-08 | 上海景吾智能科技有限公司 | Indoor mobile robot navigation method |
CN114049394A (en) * | 2021-11-23 | 2022-02-15 | 智道网联科技(北京)有限公司 | Monocular distance measuring method, device, equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110136186B (en) | 2022-09-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110136186A (en) | A kind of detection target matching method for mobile robot object ranging | |
CN103424112B (en) | A kind of motion carrier vision navigation method auxiliary based on laser plane | |
CN110503040B (en) | Obstacle detection method and device | |
CN105700525B (en) | Method is built based on Kinect sensor depth map robot working environment uncertainty map | |
US11841434B2 (en) | Annotation cross-labeling for autonomous control systems | |
CN104899855A (en) | Three-dimensional obstacle detection method and apparatus | |
KR20190134231A (en) | Apparatus and method for estimating location of vehicle and computer recordable medium storing computer program thereof | |
CN112698306A (en) | System and method for solving map construction blind area by combining multiple laser radars and camera | |
CN103487034A (en) | Method for measuring distance and height by vehicle-mounted monocular camera based on vertical type target | |
CN107796373B (en) | Distance measurement method based on monocular vision of front vehicle driven by lane plane geometric model | |
CN103499337A (en) | Vehicle-mounted monocular camera distance and height measuring device based on vertical target | |
CN113160327A (en) | Method and system for realizing point cloud completion | |
CN105785989A (en) | System for calibrating distributed network camera by use of travelling robot, and correlation methods | |
CN113936198A (en) | Low-beam laser radar and camera fusion method, storage medium and device | |
CN112053324A (en) | Complex material volume measurement method based on deep learning | |
Kang et al. | Accurate fruit localisation using high resolution LiDAR-camera fusion and instance segmentation | |
CN113850102A (en) | Vehicle-mounted vision detection method and system based on millimeter wave radar assistance | |
CN108169743A (en) | Agricultural machinery is unmanned to use farm environment cognitive method | |
CN108957024A (en) | A kind of method, apparatus and electronic equipment of tachometric survey | |
CN113791074A (en) | Unmanned aerial vehicle bridge crack inspection system and method based on multi-sensor fusion | |
CN113988197A (en) | Multi-camera and multi-laser radar based combined calibration and target fusion detection method | |
CN107607939B (en) | Optical target tracking and positioning radar device based on real map and image | |
CN111126363B (en) | Object recognition method and device for automatic driving vehicle | |
KR20160125803A (en) | Apparatus for defining an area in interest, apparatus for detecting object in an area in interest and method for defining an area in interest | |
CN111256651B (en) | Week vehicle distance measuring method and device based on monocular vehicle-mounted camera |
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 |