CN106023180A - Unstructured road RGB entropy segmentation method - Google Patents
Unstructured road RGB entropy segmentation method Download PDFInfo
- Publication number
- CN106023180A CN106023180A CN201610322052.XA CN201610322052A CN106023180A CN 106023180 A CN106023180 A CN 106023180A CN 201610322052 A CN201610322052 A CN 201610322052A CN 106023180 A CN106023180 A CN 106023180A
- Authority
- CN
- China
- Prior art keywords
- road
- grid
- image
- search
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an unstructured road RGB entropy segmentation method and belongs to the technical field of intelligent vehicle auxiliary driving. The method comprises steps of: computing a RGB entropy value matrix of an unstructured road color image; searching a minimum histogram difference value by means of the histogram of RGB entropy value matrix and computing a threshold rough identification road area and non-road area to obtain a monochrome identification result image; dividing the monochrome identification result image into grids, computing a ratio of black and white pixels of each grid and converting the image into a binary image to obtain a new small image; searching the grids at the left and right edges of the road by using a contour tracing method, accurately searching the left and right edge lines of the road by using a side-to-side method so as to accurately separate the road area from the non-road area. The unstructured road RGB entropy segmentation method may well segment unstructured roads, is good in anti-interference capability, insensitive to road shapes, small in calculated quantity, and consistent with the real-time requirement of a unmanned driving system.
Description
Technical field
The present invention relates to a kind of unstructured road RGB entropy dividing method, belong to intelligent automobile auxiliary driving technology field.
Background technology
The automated driving system of vehicle is an important applied field of computer vision, and Road Detection algorithm is this field
One of key technology.Actual road can be divided into structured road and unstructured road two class.Structured road typically has
Significantly lane line and road boundary, Comparison between detecting methods is ripe.Unstructured road refers generally to the road that structuring degree is relatively low
Road, this kind of road does not has lane line and road boundary clearly, adds shade and the impact of water mark, and detection difficulty is relatively
Greatly.
At present, the detection algorithm of unstructured road be broadly divided into the method for feature based, method based on model, based on
The method of neutral net and method based on support vector machine.The Road Detection algorithm of feature based, mainly exists according to road
Difference in the features such as color, texture, gradient, makes a distinction with non-rice habitats.Its advantage is insensitive to road shape, needs
Priori is few, but the most sensitive to shade and water mark, and computational processing is bigger.Method based on model, it is necessary first to look for
To the road model mated most, the road area that the method detects is more complete, but for complicated road pavement form, it is impossible to
Set up effective model.Method based on neutral net and method based on support vector machine, need training substantial amounts of to sample
Collection.
In consideration of it, develop a kind of can the method in effective Accurate Segmentation unstructured road region particularly important.
Summary of the invention
It is an object of the invention to unstructured road border be difficult to accurately define, and easily by chaff interference in shade, road
Impact, it is proposed that a kind of unstructured road RGB entropy dividing method, it is intended to solve existing dry to shadows on the road, barrier etc.
Disturb, Accurate Segmentation unstructured road region, and amount of calculation meets Unmanned Systems's requirement of real-time.
The present invention is a kind of unstructured road RGB entropy dividing method, realizes especially by following steps:
Step 1, by vehicle-mounted single-lens ccd video camera obtain unstructured road image.
The RGB entropy matrix of step 2, first calculating unstructured road image.Then the Nogata of RGB entropy matrix is calculated
Figure, finds minimum histogram difference, and calculates threshold value Threshold.Finally according to threshold value, rough identification road area and Fei Dao
Region, road, obtains black and white identification image.
Step 2-1, to the road coloured image that size is M × N, calculate each component of normalization RGB.
RA (i, j)=R (i, j)/(R (i, j)+G (i, j)+B (i, j))
GA (i, j)=G (i, j)/(R (i, j)+G (i, j)+B (i, j))
BA (i, j)=B (i, j)/(R (i, j)+G (i, j)+B (i, j))
In formula, i=0 ..., M-1, j=0 ..., N-1.
Step 2-2, calculating RGB entropy matrix HRGB, and its value is transformed between [0,255], obtain the matrix of M × N
HRGB′。
HRGB(i, j)=-RA (i, j) log2RA (i, j)-GA (i, j) log2GA (i, j)-BA (i, j) log2BA
(i, j)
Step 2-3, to matrix HRGB' calculate its grey level histogram, obtain f=[f (0), f (1) ..., f (255)].
Step 2-4, calculating histogram difference cha.
In formula, t=1 ..., 255.
Step 2-5, successively decreasing from t=255 searches out t=1, by the method finding minimum histogram difference, calculates threshold value
Threshold。
Step 2-6, to the color road original image that size is M × N, road image is divided into two classes 0 (represent road pictures
Element), 255 (non-rice habitats pixels), obtain the bianry image WR that size is M × N, its pixel value be ω (i, j).
Wherein, (i j) is matrix H to hRGBValue in '.
Step 3, black white image to step 2-6 gained, be divided into multiple grid, and calculate monochrome pixels ratio in grid
Example, is finally converted into bianry image.
Step 3-1, it is M × N black and white binary image WR to the size obtained by step 2-6, is first removed isolated white
Colour vegetarian refreshments processes, and is then divided into multiple grid, as shown in Figure 2.Set initial value t, divide the image into q × r grid:
In formula,Represent rounding operation.
Be not necessarily integer in view of (M/t, N/t), it is clear that the number of lines of pixels that last column grid comprises be M-(q-1) ×
T, the pixel columns that last string grid comprises is N-(r-1) × t, and the number of pixels that other grid comprises is t × t.
Step 3-2, calculate each grid monochrome pixels ratio:
Wp (i, j)=NW(i, j)/Nz(i, j)
In formula, i=1 ..., q, j=1 ..., r, NW(i, j) and Nz(i, j) represents the i-th row respectively, white in j row grid
Pixel and total number of pixels.
Wp (i, j) value is between [0,1], wp be a dimension be the matrix of q × r, be multiplied by 255, can be with the shape of image
Formula shows its result.
Step 3-3, wp is converted into bianry image.First wp is removed isolated white boxes process, then sets
Switching threshold, converts images into bianry image:
Little in view of road area white point, therefore coefficient kcoefChoose smaller, be typically set at [0.01 0.1] it
Between.
Step 4, in step 3-3 bianry image, use Contour extraction method search road left and right edges grid.Search be
Size be q × r black and white binary image wt in carry out.For ease of distinguishing, a pixel in q × r image, referred to as one
Grid.Search road left and right edges grid technique is as shown in Figure 3.
Step 4-1, starting point choose midpoint, bottom, first look for the leftmost edge grid of road.By midpoint, bottom to the left
Search, until grid value is 255 or image the 1st row, write down respective party case put (i, j) in j value in LeftPoint.j,
I.e. write down the position of road left hand edge black box row;Similar, search for the right, and write down corresponding grid train value and arrive
In RightPoint.j.Simultaneously so that the grid value between this row [LeftPoint.j, RightPoint.j] is 0, other
Grid value is 255.Calculate and store the W=Rigth.j-LeftPoint.j that has a lot of social connections.
Step 4-2, upwards lattice, i.e. continue search for image lastrow.First it is the search of left side road edge.With step
Column position LeftPoint.j in 4-1 is starting point.If current grid (i-1, LeftPoint.j) is white boxes, the most to the right
Continue search for, until grid value is 0 end, write down corresponding grid (i-1, j) in j value in LeftPoint.j;If working as front
Lattice are black box, continue search for the most to the left, until grid value is 255 end, write down corresponding grid (i-1, j) in j value arrive
In LeftPoint.j.The search of the right road edge is similar to.Simultaneously so that this row [LeftPoint.j, RightPoint.j]
Between grid value be 0, other grid value is 255.Calculate and store the W=Rigth.j-LeftPoint.j that has a lot of social connections.
Step 4-3, repetition step 4-2, until having a lot of social connections W less than setting threshold value WthTill, terminate search.
Step 5, in the black and white result images of step 2-6, the result of integrating step 4, by the method for edge-to-edge
(side-to-side), precise search road left and right edges line and realize Accurate Segmentation road area and non-rice habitats region.In step
In rapid 4, (i, LeftPoint.j) is the result of search in the image that size is q × r, corresponds to the image that size is M × N
In, then it represents that the i-th row the LeftPoint.j grid, refer to the black box of white and black box intersection, such as Fig. 4 institute
Show.Each grid comprises several pixels.Here search road left and right edges line, refers in the image that size is M × N
(i.e. ω in step 2-6) searches for road left and right edges line, and extracts road area.Search road left and right edges line divides with region
Segmentation method is as shown in Figure 4.
Step 5-1, search left side road edge line, from the beginning of image base.With in step 4 search result grid (i,
LeftPoint.j) it is original position, from the beginning of grid bottommost high order end, searches for the left, until pixel value is 255 or image
1st row, terminate search, write down corresponding black picture element position (x, LeftPoint.y).Similar, search for the right, and write down corresponding
In black pixel point position (x, RightPoint.y).Herein for being easy to compare so that x row, [LeftPoint.y,
RightPoint.y] row between pixel value be 255, other pixel value is original path image.
Step 5-2, in this grid to lastrow, repeat step 5-1, until all provisional capitals of this grid have been searched for.
Step 5-3, upwards a grid, start search from grid bottommost, repeats step 5-1,5-2, until this grid
All provisional capitals have been searched for.
Step 5-4, repetition step 5-3, until the search of all of black box is complete.
Beneficial effect
The present invention a kind of unstructured road RGB entropy dividing method, by calculating the RGB entropy of road image, is distinguished roughly
Road area and non-rice habitats region, then use edge-to-edge (side-to-side) method accurately to split road area and non-
Road area.Compared with prior art, this dividing method energy Accurate Segmentation unstructured road, to shadows on the road, road barricade
Things etc. have stronger capacity of resisting disturbance, and insensitive to road shape.Amount of calculation is less simultaneously, meets Unmanned Systems real-time
Property requirement.
Accompanying drawing explanation
Fig. 1 is the implementing procedure figure of the present invention a kind of unstructured road RGB entropy dividing method;
Fig. 2 is that road image is divided into multiple grid technique figure;
Fig. 3 is to use Contour extraction method search road left and right edges grid technique figure.
Fig. 4 is to use edge-to-edge (side-to-side) method segmentation road area method figure
Fig. 5 unstructured road colour original
Fig. 6 is the rough segmentation effect figure of unstructured road based on RGB entropy
Fig. 7 is Fig. 6 to be divided into multiple grid and calculates the bianry image that its monochrome pixels ratio is changed into
Fig. 8 is to use Contour extraction method search road left and right edges grid design sketch
Fig. 9 is to use edge-to-edge (side-to-side) method Accurate Segmentation road area design sketch
Figure 10 is to use mid-to-side method segmentation road area design sketch
Detailed description of the invention
The present invention will be further described and describes in detail with embodiment below in conjunction with the accompanying drawings:
Embodiment:
In order to verify the effectiveness of this dividing method, under different types of unstructured road scene, do lane segmentation
Experiment, and compare with mid-to-side method Road image segmentation.The resolution of image is 695 × 950, specifically real
Applying method is:
One, unstructured road coloured image is obtained by vehicle-mounted single-lens ccd video camera, as shown in Figure 5.
Two, the RGB entropy of calculating unstructured road image, and calculate the threshold value of RGB entropy, identify roughly road area
With non-rice habitats region, as shown in Figure 6.
Three, unstructured road edge line and Accurate Segmentation road area are accurately searched
Step 1, the monochrome pixels ratio of calculating second step gained black and white binary image.
Step 1-1, the size obtaining second step are 695 × 950 black and white binary image, are first removed isolated white
Pixel processes, and is then divided into multiple grid.Set initial value t=8, divide the image into 87 × 119 grids.Consider
(M/t, N/t) is not necessarily integer, it is clear that the number of lines of pixels that last column grid comprises is 7 row, and last string grid comprises
Pixel columns is 6 row, and the number of pixels that other grid comprises is 8 × 8.
Step 1-2, each grid monochrome pixels ratio are calculated as follows:
Wp (i, j)=NW(i, j)/Nz(i, j)
In formula, i=1 ..., q, j=1 ..., r, NW(i, j) and Nz(i, j) represents the i-th row respectively, white in j row grid
Pixel and total number of pixels.
Step 2, wp is converted into bianry image.First wp is removed isolated white boxes process, then sets and turn
Change threshold value, convert images into bianry image:
Little in view of road area white point, therefore coefficient kcoefChoose smaller, take k herecoef=0.01.
Gained bianry image is as shown in Figure 7.
Step 3, employing Contour extraction method search road left and right edges grid.Search is to be the black and white of 87 × 119 in size
Bianry image wt is carried out.For ease of distinguishing, it is a pixel in 87 × 119 images size, referred to as one grid.Institute
Obtain result figure as shown in Figure 8.
Step 4, employing edge-to-edge (side-to-side) method precise search road left and right edges line and segmentation road
Region.(i, LeftPoint.j) is the result of search in the image that size is 87 × 119 in step 3, and corresponding to size is
In the image of 695 × 950, then it represents that the i-th row the LeftPoint.j grid, refer to white and black box intersection black
Color grid, as shown in Figure 4.Each grid comprises several pixels.Here search road left and right edges line, refer in size be
Search road left and right edges line in the image (i.e. Fig. 6) of 695 × 950, and split road area.Acquired results is as shown in Figure 9.For
Strengthening the contrast effect with original image, road area white represents, other region retains original image.In order to compare segmentation
Effect, Figure 10 gives the design sketch of mid-to-side method segmentation road image.
The above is presently preferred embodiments of the present invention, and the present invention should not be limited to this embodiment and accompanying drawing institute is public
The content opened.Every without departing from the equivalence completed under spirit disclosed in this invention or amendment, both fall within the model of present invention protection
Enclose.
Claims (5)
1. a unstructured road RGB entropy dividing method, it is characterised in that comprise the following steps:
Step 1, collection road coloured image.
Step 2, the RGB entropy of calculating unstructured road coloured image, and calculate the threshold value of RGB entropy, identify roughly road
Pixel and non-rice habitats pixel, obtain black and white identification image.
Step 3, to gained black and white identification image, be divided into multiple grid, and calculate monochrome pixels ratio in grid, obtain one
New less image, is converted into bianry image.
Step 4, in step 3 gained bianry image, use Contour extraction method search road left and right edges grid.
Step 5, in the black and white result images of step 2, the result of integrating step 4, by the method (side-to-of edge-to-edge
Side), precise search road left and right edges line, it is achieved precisely split road area.
Method the most according to claim 1, it is characterised in that described step 2 further includes steps of
Step 2-1, to the road coloured image that size is M × N, calculate each component of normalization RGB.
RA (i, j)=R (i, j)/(R (i, j)+G (i, j)+B (i, j))
GA (i, j)=G (i, j)/(R (i, j)+G (i, j)+B (i, j))
BA (i, j)=B (i, j)/(R (i, j)+G (i, j)+B (i, j))
In formula, i=0 ..., M-1, j=0 ..., N-1.
Step 2-2, calculating RGB entropy matrix HRGB, and its value is transformed between [0,255], obtain the matrix H of M × NRGB′。
HRGB(i, j)=-RA (i, j) log2RA (i, j)-GA (i, j) log2GA (i, j)-BA (i, j) log2BA (i, j)
Step 2-3, to matrix HRGB' calculate its grey level histogram, obtain f=[f (0), f (1) ..., f (255)].
Step 2-4, calculating histogram difference cha.
In formula, t=1 ..., 255.
Step 2-5, successively decreasing from t=255 searches out t=1, by the method finding minimum histogram difference, calculates threshold value
Threshold。
Step 2-6, to the color road original image that size is M × N, road image is divided into two classes 0 (expression road pixel),
255 (non-rice habitats pixels), obtain black and white binary image WR that size is M × N, its pixel value be ω (i, j).
Wherein, (i j) is matrix H to hRGBValue in '.
Method the most according to claim 1, it is characterised in that described step 3 further includes steps of
Step 3-1, to M × N black and white binary image WR, be first removed isolated white pixel point and process, be then divided into many
Individual grid.Set initial value t, divide the image into q × r grid:
In formula,Represent rounding operation.
Step 3-2, calculate each grid monochrome pixels ratio.Each grid monochrome pixels ratio is defined as:
Wp (i, j)=NW(i, j)/Nz(i, j)
In formula, i=1 ..., q, j=1 ..., r, NW(i, j) and Nz(i, j) represents the i-th row respectively, white pixel in j row grid
With total number of pixels.
Step 3-3, wp is converted into bianry image.First wp is removed isolated white boxes process, then sets conversion
Threshold value, converts images into bianry image:
Little in view of road area white point, therefore coefficient kcoefChoose smaller, be typically set between [0.01 0.1].
Method the most according to claim 1, it is characterised in that described step 4 further includes steps of
Step 4-1, search are to carry out in black and white binary image wt that size is q × r.For ease of distinguishing, q × r image
In a pixel, referred to as one grid.Starting point chooses midpoint, bottom, first looks for the leftmost edge grid of road.The end of by
Portion is searched at midpoint to the left, until grid value be 255 or image the 1st row, write down respective party case put (i, j) in j value arrive
In LeftPoint.j, i.e. write down the position of road left hand edge black box row;Similar, search for the right, and write down corresponding grid row
It is worth in RightPoint.j.Simultaneously so that the grid value between this row [LeftPoint.j, RightPoint.j] is 0,
Other grid value is 255.Calculate and store the W=Rigth.j-LeftPoint.j that has a lot of social connections.
Step 4-2, upwards lattice, i.e. continue search for image lastrow.First it is the search of left side road edge.With step 4-1
In column position LeftPoint.j be starting point.If current grid (i-1, LeftPoint.j) is white boxes, continue the most to the right
Continuous search, until grid value is 0 end, write down corresponding grid (i-1, j) in j value in LeftPoint.j;If current grid
For black box, continue search for the most to the left, until grid value is 255 end, write down corresponding grid (i-1, j) in j value arrive
In LeftPoint.j.The search of the right road edge is similar to.Simultaneously so that this row [LeftPoint.j, RightPoint.j]
Between grid value be 0, other grid value is 255.Calculate and store the W=Rigth.j-LeftPoint.j that has a lot of social connections.
Step 4-3, repetition step 4-2, until having a lot of social connections W less than setting threshold value WthTill, terminate search.
Method the most according to claim 1, it is characterised in that described step 5 further includes steps of
Step 5-1, search road left and right edges line, refer to search for road left and right edges in the bianry image WR that size is M × N
Line, and extract road area.Search left side road edge line, from the beginning of image base.With the result grid of search in step 4
(i, LeftPoint.j) is original position, from the beginning of grid bottommost high order end, searches for the left, until pixel value is 255 or figure
As the 1st row, terminate search, write down corresponding black picture element position (x, LeftPoint.y).Similar, search for the right, and write down phase
Answer in black pixel point position (x, RightPoint.y).For the ease of comparing, x row [LeftPoint.y, RightPoint.y]
Pixel value between row is 255, and other pixel value is original path image.
Step 5-2, in this grid to lastrow, repeat step 5-1, until all provisional capitals of this grid have been searched for.
Step 5-3, upwards a grid, start search from grid bottommost, repeats step 5-1,5-2, until this grid owns
Provisional capital has been searched for.
Step 5-4, repetition step 5-3, until the search of all of black box is complete.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610322052.XA CN106023180A (en) | 2016-05-17 | 2016-05-17 | Unstructured road RGB entropy segmentation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610322052.XA CN106023180A (en) | 2016-05-17 | 2016-05-17 | Unstructured road RGB entropy segmentation method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106023180A true CN106023180A (en) | 2016-10-12 |
Family
ID=57097322
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610322052.XA Pending CN106023180A (en) | 2016-05-17 | 2016-05-17 | Unstructured road RGB entropy segmentation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106023180A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107943061A (en) * | 2018-01-09 | 2018-04-20 | 辽宁工业大学 | A kind of model automobile automatic Pilot experimental provision and method based on machine vision |
CN110335322A (en) * | 2019-07-09 | 2019-10-15 | 成都理工大学 | Roads recognition method and road Identification device based on image |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130128001A1 (en) * | 2011-11-18 | 2013-05-23 | Ganmei YOU | Method and system for detecting object on a road |
CN104298775A (en) * | 2014-10-31 | 2015-01-21 | 北京工商大学 | Multi-feature content-based image retrieval method and system |
CN104700071A (en) * | 2015-01-16 | 2015-06-10 | 北京工业大学 | Method for extracting panorama road profile |
EP3076367A1 (en) * | 2013-07-22 | 2016-10-05 | Zhejiang University | Method for road detection from one image |
-
2016
- 2016-05-17 CN CN201610322052.XA patent/CN106023180A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130128001A1 (en) * | 2011-11-18 | 2013-05-23 | Ganmei YOU | Method and system for detecting object on a road |
EP3076367A1 (en) * | 2013-07-22 | 2016-10-05 | Zhejiang University | Method for road detection from one image |
CN104298775A (en) * | 2014-10-31 | 2015-01-21 | 北京工商大学 | Multi-feature content-based image retrieval method and system |
CN104700071A (en) * | 2015-01-16 | 2015-06-10 | 北京工业大学 | Method for extracting panorama road profile |
Non-Patent Citations (4)
Title |
---|
LIU HUA-JUN 等: "A Fast Method for Vanishing Point Estimation and Tracking and Its Application in Road Images", 《2006 6TH INTERNATIONAL CONFERENCE ON ITS TELECOMMUNICATIONS》 * |
柏猛 等: "基于图模型的道路检测方法", 《模式识别与人工智能》 * |
王晓栋: "基于彩色特征和B-spline的非结构化道路检测算法及应用研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
郭秋梅 等: "基于二维熵和轮廓特征的非结构化道路检测", 《计算机应用》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107943061A (en) * | 2018-01-09 | 2018-04-20 | 辽宁工业大学 | A kind of model automobile automatic Pilot experimental provision and method based on machine vision |
CN110335322A (en) * | 2019-07-09 | 2019-10-15 | 成都理工大学 | Roads recognition method and road Identification device based on image |
CN110335322B (en) * | 2019-07-09 | 2024-03-01 | 成都理工大学 | Road recognition method and road recognition device based on image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103093250B (en) | A kind of Adaboost method for detecting human face based on new Haar-like feature | |
CN104850850A (en) | Binocular stereoscopic vision image feature extraction method combining shape and color | |
Zheng et al. | Large-scale oil palm tree detection from high-resolution remote sensing images using faster-rcnn | |
CN107463892A (en) | Pedestrian detection method in a kind of image of combination contextual information and multi-stage characteristics | |
CN105354581B (en) | The color image feature extracting method of Fusion of Color feature and convolutional neural networks | |
CN103473785B (en) | A kind of fast multi-target dividing method based on three-valued image clustering | |
CN102708356A (en) | Automatic license plate positioning and recognition method based on complex background | |
CN107092871A (en) | Remote sensing image building detection method based on multiple dimensioned multiple features fusion | |
CN108876805B (en) | End-to-end unsupervised scene passable area cognition and understanding method | |
CN105005778A (en) | Expressway vehicle detection method based on visual background extraction | |
CN105389774A (en) | Method and device for aligning images | |
CN104732534B (en) | Well-marked target takes method and system in a kind of image | |
CN104517317A (en) | Three-dimensional reconstruction method of vehicle-borne infrared images | |
CN104217436B (en) | SAR image segmentation method based on multiple features combining sparse graph | |
CN107944437B (en) | A kind of Face detection method based on neural network and integral image | |
CN102982561B (en) | Method for detecting binary robust scale invariable feature of color of color image | |
CN110414385A (en) | A kind of method for detecting lane lines and system based on homography conversion and characteristic window | |
CN104851089A (en) | Static scene foreground segmentation method and device based on three-dimensional light field | |
CN105405138A (en) | Water surface target tracking method based on saliency detection | |
CN115661754B (en) | Pedestrian re-recognition method based on dimension fusion attention | |
CN104966054A (en) | Weak and small object detection method in visible image of unmanned plane | |
CN108388901B (en) | Collaborative significant target detection method based on space-semantic channel | |
CN102930292B (en) | A kind of object identification method based on p-SIFT feature | |
CN105787932A (en) | Stereo matching method based on segmentation cross trees | |
CN108319961B (en) | Image ROI rapid detection method based on local feature points |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20161012 |
|
WD01 | Invention patent application deemed withdrawn after publication |