CN104866819B - A kind of classification of landform method based on trinocular vision system - Google Patents
A kind of classification of landform method based on trinocular vision system Download PDFInfo
- Publication number
- CN104866819B CN104866819B CN201510213891.3A CN201510213891A CN104866819B CN 104866819 B CN104866819 B CN 104866819B CN 201510213891 A CN201510213891 A CN 201510213891A CN 104866819 B CN104866819 B CN 104866819B
- Authority
- CN
- China
- Prior art keywords
- landform
- data
- color
- ground
- vision system
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of classification of landform methods based on trinocular vision system, comprising the following steps: tri-item stereo vision system samples landform and exports 3D data abundant first；Then 3D data transmission distinguishes landform eventually by color sorter to geometric classfication device, Combining with terrain color characteristic.The invention enables the vehicle homing guidances based on trinocular vision system to become simple and quick, avoids the training of big data quantity, does not need manually to mark landform；This system both can be used as the supplement of complicated vehicle automatic guidance system, and a simple article travelling bogie automatic guidance system can also be made, and have certain market prospects.
Description
Technical field
The present invention relates to vehicle homing guidance devices, and in particular to a kind of classification of landform side based on trinocular vision system
Method.
Background technique
Vehicle guide automatically be the core the most of vehicle automatic running study portion, the development of modern artificial intelligence technology
So that the automatic bootstrap technique of vehicle is more mature.Since Electronic company releases first automatic guide vehicle system and puts into
After industrial production use, the automatic guidance of vehicle is stepped into industrial and agricultural production and terrain detection, the goods transportation in warehouse, fire
The application such as automatic running of star detector is familiar with by everybody.Simultaneously using image recognition progress terrain detection technology also by
Step attracts attention, but still has shortcoming to the design of classifier.Traditional classifier can't the different ground of automatic identification
Shape needs to acquire a large amount of terrain data as sample database, is labeled manually to the landform in sample database by user, and
Complicated landform identification model is established, classifier is trained with sample database, to set up the recognition template of landform.It is basic to do so
It is not carried out the function of guiding automatically, the data volume for the landform for manually needing to mark is larger, and can not take the generality of landform into account,
Landform recognition performance relies on landform sample database, and since terrain type is various, landform sample database is difficult to cover the landform of all kinds,
The accuracy of landform identification is influenced big by classifier training process, and essentially all conventional classification of landform device all exists such
Problem.
Summary of the invention
It is an object of the invention to overcome problem above of the existing technology, provide a kind of based on trinocular vision system
Classification of landform method shoots landform by trinocular vision system to obtain 3D geometric data and color data, sends to point
Class device is classified.
To achieve the above object, reach above-mentioned technical effect, the invention is realized by the following technical scheme:
A kind of classification of landform method based on trinocular vision system, comprising the following steps:
Step 1) tri-item stereo vision system samples landform and exports 3D data and color data abundant；
Step 2 3D data and picture color data transmission to classifier.
Further, the 3D data in the step (1) include topographical surface geometric data.
Further, the classifier in the step (2) is divided into geometric classfication device and color sorter.
Further, the geometric classfication device extracts the geometrical characteristic vector of topographical surface 3D data, according to feature vector
Classify to landform, distinguishes travelable ground and can not running ground.
Further, the color sorter is extracted according to the color data of the topographic map of tri-item stereo vision system photographs
The color feature vector of landform, and the classification results of geometric classfication device is combined to carry out color mark to travelable ground noodles.
The beneficial effects of the present invention are:
The invention enables the vehicle homing guidances based on trinocular vision system to become simple and quick, avoids big data quantity
Training, do not need manually to mark landform；This system both can be used as the supplement of complicated vehicle automatic guidance system, can also make
At a simple article travelling bogie automatic guidance system, there are certain market prospects.
The above description is only an overview of the technical scheme of the present invention, in order to better understand the technical means of the present invention,
And can be implemented in accordance with the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention and the accompanying drawings.
A specific embodiment of the invention is shown in detail by following embodiment and its attached drawing.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present invention, constitutes part of this application, this hair
Bright illustrative embodiments and their description are used to explain the present invention, and are not constituted improper limitations of the present invention.In the accompanying drawings:
Fig. 1 is present system structural schematic diagram；
Fig. 2 is the function treatment flow chart of present system；
Fig. 3 is the specific steps flow chart of present system.
Specific embodiment
Below with reference to the accompanying drawings and in conjunction with the embodiments, design feature and technology implementation process of the invention be described in detail:
As shown in Figure 1, this system by trinocular vision system to landform carry out Image Acquisition, the 3 of tri-item stereo vision system
A video camera is placed along the same direction, and is placed apart from identical.Tri-item stereo vision system has a narrow baseline and one
A wide baseline, narrow baseline is on the left of use and the baseline of the 0.12m of center camera, wide baseline are to be imaged using left and right side
The baseline of the 0.24m of machine.Narrow baseline increases two common field ranges of video camera, and wide baseline is in each visual distance
There is biggish absolute visual field range.3 video cameras of tri-item stereo vision system shoot three width terrain graphs from different perspectives, so
The depth information of landform is obtained using algorithm for stereo matching afterwards.After obtaining topographical surface 3D information, data are after pretreatment
It is transmitted to geometric classfication device by sensor, 3D information is divided into the minizone that size is 0.4m × 0.4m by geometric classfication device, several
What classifier extracts the geometrical characteristic vector of certain points from section, is labeled according to geometrical characteristic vector to landform.Geometry
Feature vector is expressed as, whereinFor the angle in least square face and horizontal plane,It is data point from minimum
Two multiply the mean square deviation in face,For the variance of the coordinate of data point,For the average value of the coordinate of data point.Geometric classfication device can
Classified according to the value of element each in geometrical characteristic vector to landform.The face on color sorter extraction image mesorelief surface
Color characteristic vector, and travelable ground noodles different in image are marked according to color feature vector.Color data be pixel in it is red,
Green, the intensity value of blue three primary colours, it can be directly obtained from color image, the shadow due to stereo visual system vulnerable to intensity of illumination
It rings, it is proposed that another method display color feature, that is, use vectorTo show color characteristic.
(2)
(3)
(4)
In formulaFor three components of pixel.Since there are many pixel in figure, if each pixel calculates meeting
Very big resource is occupied, so need to only calculate the average value of certain pixels.In automatic recognition system, color sorter
Automatically color feature vector is associated with the landform that geometric classfication device is marked.
Geometric classfication device distinguished according to threshold value can travel in landform ground and can not running ground, the determination of threshold value uses height
This mixed model carrys out threshold value to terrain modeling, further according to mahalanobis distance.
Fig. 2 is the function treatment flow chart of present system.Since 3D data include landform geological information, geometric classfication device
The landform geometrical characteristic vector for extracting 3D data maps whether geometrical characteristic vector is that can travel ground according to gauss hybrid models
Face Gaussian Profile and can not running ground Gaussian Profile, the geometrical characteristic vector of landform sample meets which kind of Gaussian Profile is just classified as
Corresponding terrain type, output can not running ground class, can travel ground noodles need color sorter to do further mark to it.Face
Colour sorting device carries out color mark to travelable ground subclass according to the terrain colors data in picture, so that can travel ground noodles
Shown with different colours, for can not running ground class then marked with unified fixed color.With the traveling of vehicle, three
The topographical surface data of item stereo vision system acquisition are constantly updated, and the tag along sort of whole system is also constantly updated accordingly.
In this way, vehicle can travel always on it can travel ground, the function of vehicle homing guidance is realized.
Specific process:
System is initialized first, by shooting the ground image of a width spaciousness clear, establishes the 3D of this ground image
Figure, is divided into minizone for 3D figure, extracts the geometrical characteristic vector of each minizone, establish initial characteristics vector library.Then, with
The movement of vehicle, tri-item stereo vision system landform is shot, establish landform 3D figure.3D figure is divided into minizone,
The geometrical characteristic vector for extracting each section is identified as travelable ground with initial characteristics vector storehouse matching if successful match,
Barrier is identified as if matching is unsuccessful.The terrain colors feature vector in picture is extracted, the classification on travelable ground is believed
Breath is combined with its color feature vector, makes color differentiation to the travelable ground of difference, and store it is emerging in picture can
The feature vector of running ground.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (4)
1. a kind of classification of landform method based on trinocular vision system, which comprises the following steps:
Step 1) tri-item stereo vision system samples landform and exports 3D data and color data abundant；
To classifier, classifier includes geometric classfication device and color sorter for step 2 3D data and picture color data transmission,
Geometric classfication device extracts the landform geometrical characteristic vector of 3D data, geometrical characteristic vector is mapped according to gauss hybrid models whether
For can travel ground Gaussian Profile and can not running ground Gaussian Profile, which kind of Gauss the geometrical characteristic vector of landform sample meet
Distribution is just classified as corresponding terrain type, output can not running ground class, can travel ground noodles need color sorter it is done into
One step mark, color sorter carries out color mark to travelable ground subclass according to picture color data, so that can travel ground
Noodles show with different colours, for can not running ground class then marked with unified fixed color, with the traveling of vehicle,
The topographical surface data of acquisition are constantly updated, tag along sort is also constantly updated accordingly, and vehicle travels always feasible
It sails on ground, constitutes the homing guidance of vehicle.
2. the classification of landform method according to claim 1 based on trinocular vision system, which is characterized in that the step 1)
In 3D data include topographical surface geometric data.
3. the classification of landform method according to claim 1 based on trinocular vision system, which is characterized in that the geometry point
Class device extracts the geometrical characteristic vector of topographical surface 3D data, is classified according to feature vector to landform, is distinguished feasible
Sail ground and can not running ground.
4. the classification of landform method according to claim 1 based on trinocular vision system, which is characterized in that the color point
Class device extracts the color feature vector of landform according to the color data of the topographic map of tri-item stereo vision system photographs, and combines several
The classification results of what classifier carry out color mark to travelable ground noodles.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510213891.3A CN104866819B (en) | 2015-04-30 | 2015-04-30 | A kind of classification of landform method based on trinocular vision system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510213891.3A CN104866819B (en) | 2015-04-30 | 2015-04-30 | A kind of classification of landform method based on trinocular vision system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104866819A CN104866819A (en) | 2015-08-26 |
| CN104866819B true CN104866819B (en) | 2018-12-14 |
Family
ID=53912640
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510213891.3A Expired - Fee Related CN104866819B (en) | 2015-04-30 | 2015-04-30 | A kind of classification of landform method based on trinocular vision system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104866819B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107255446B (en) * | 2017-08-01 | 2020-01-07 | 南京农业大学 | Dwarfing close-planting fruit tree canopy three-dimensional map construction system and method |
| CN108844618B (en) * | 2018-06-12 | 2019-07-23 | 中国科学技术大学 | A kind of landform cognitive method |
| DK180774B1 (en) * | 2018-10-29 | 2022-03-04 | Motional Ad Llc | Automatic annotation of environmental features in a map during navigation of a vehicle |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103017739A (en) * | 2012-11-20 | 2013-04-03 | 武汉大学 | Manufacturing method of true digital ortho map (TDOM) based on light detection and ranging (LiDAR) point cloud and aerial image |
| CN103389103A (en) * | 2013-07-03 | 2013-11-13 | 北京理工大学 | Geographical environmental characteristic map construction and navigation method based on data mining |
| CN103645480A (en) * | 2013-12-04 | 2014-03-19 | 北京理工大学 | Geographic and geomorphic characteristic construction method based on laser radar and image data fusion |
-
2015
- 2015-04-30 CN CN201510213891.3A patent/CN104866819B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103017739A (en) * | 2012-11-20 | 2013-04-03 | 武汉大学 | Manufacturing method of true digital ortho map (TDOM) based on light detection and ranging (LiDAR) point cloud and aerial image |
| CN103389103A (en) * | 2013-07-03 | 2013-11-13 | 北京理工大学 | Geographical environmental characteristic map construction and navigation method based on data mining |
| CN103645480A (en) * | 2013-12-04 | 2014-03-19 | 北京理工大学 | Geographic and geomorphic characteristic construction method based on laser radar and image data fusion |
Non-Patent Citations (1)
| Title |
|---|
| 基于多传感器融合的复杂地形分类算法研究;左亮,王美玲,杨毅;《第三十二届中国控制会议论文集（D卷）》;20130726;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104866819A (en) | 2015-08-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104850850B (en) | A kind of binocular stereo vision image characteristic extracting method of combination shape and color | |
| Liu et al. | Improved kiwifruit detection using pre-trained VGG16 with RGB and NIR information fusion | |
| CN106707296B (en) | It is a kind of based on the unmanned machine testing of Based on Dual-Aperture photo electric imaging system and recognition methods | |
| CN109791554B (en) | System and method for automatically generating metadata for a media document | |
| Wegner et al. | Cataloging public objects using aerial and street-level images-urban trees | |
| CN110175576A (en) | A kind of driving vehicle visible detection method of combination laser point cloud data | |
| Recky et al. | Windows detection using k-means in cie-lab color space | |
| CN103077521B (en) | A kind of area-of-interest exacting method for video monitoring | |
| CN101701818B (en) | Method for detecting long-distance barrier | |
| CN109102547A (en) | Robot based on object identification deep learning model grabs position and orientation estimation method | |
| CN110443898A (en) | A kind of AR intelligent terminal target identification system and method based on deep learning | |
| CN102999892A (en) | Intelligent fusion method for depth images based on area shades and red green blue (RGB) images | |
| CN109949593A (en) | A kind of traffic lights recognition methods and system based on crossing priori knowledge | |
| CN105678318B (en) | The matching process and device of traffic sign | |
| CN102147867B (en) | Method for identifying traditional Chinese painting images and calligraphy images based on subject | |
| CN103035013A (en) | Accurate moving shadow detection method based on multi-feature fusion | |
| CN108537239A (en) | A kind of method of saliency target detection | |
| CN105631852B (en) | Indoor human body detection method based on depth image contour | |
| CN104077605A (en) | Pedestrian search and recognition method based on color topological structure | |
| CN105005766A (en) | Vehicle body color identification method | |
| CN106600643B (en) | A kind of demographic method based on trajectory analysis | |
| CN104866819B (en) | A kind of classification of landform method based on trinocular vision system | |
| CN104517095A (en) | Head division method based on depth image | |
| CN102915433A (en) | Character combination-based license plate positioning and identifying method | |
| CN106503170B (en) | It is a kind of based on the image base construction method for blocking dimension |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181214 Termination date: 20190430 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |