CN100552698C - Real-time detection method to airfield runway in the image of taking photo by plane - Google Patents
Real-time detection method to airfield runway in the image of taking photo by plane Download PDFInfo
- Publication number
- CN100552698C CN100552698C CNB2007101935251A CN200710193525A CN100552698C CN 100552698 C CN100552698 C CN 100552698C CN B2007101935251 A CNB2007101935251 A CN B2007101935251A CN 200710193525 A CN200710193525 A CN 200710193525A CN 100552698 C CN100552698 C CN 100552698C
- Authority
- CN
- China
- Prior art keywords
- image
- search
- runway
- straight line
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to the real-time detection method of image terminal guidance field target, especially a kind of real-time detection method to airfield runway in the image of taking photo by plane.The method that the present invention adopts Hough transformation to combine with direction encoding realized the taking photo by plane real-time detection of airfield runway in the image.At first adopt the direction encoding method that the image travel direction is scanned, obtain the information such as position, length, direction of straight-line segment, by adding up the main direction of all straight-line segments, determine the rough direction of runway, in the neighborhood of this direction, carry out the Hough transformation computing then, thereby dwindled the hunting zone of Hough transformation greatly, improved the accuracy rate and the arithmetic speed of algorithm.Show by a large amount of experiments, when rotation at any angle, convergent-divergent take place, block in image, when illumination variation, the fuzzy interference, this method all can effectively overcome the various variations of image, stable detection goes out straight-line target.
Description
Technical field
The present invention relates to the real-time detection method of image terminal guidance field target, especially a kind of real-time detection method to airfield runway in the image of taking photo by plane.
Background technology
The target detection of taking photo by plane in the image is the research hot issue of domestic and international common concern.The airport is the special target of a class as important transportation place, and the detection of airfield runway is all had great importance from civilian or military angle.Observation can see from the high-altitude of the internet airport image of taking photo by plane, and the take photo by plane background more complicated of image of airport is separated the airport and to be had certain degree of difficulty from background, and practical application needs requirement of real time.But can also see that from image airfield runway has than obvious characteristics in the image of taking photo by plane on airport: the length and width of airfield runway have certain scope, and the gray scale of runway itself comparatively steadily and the runway shape that is rectangle.Adopt traditional correlation matching algorithm, matching efficiency is very low, and reason is as follows:
1. because runway may be blocked by some vehicles such as things such as aircraft, truck, when features such as using major-minor runway junction was carried out relevant matches, matching efficiency was very low.
2. because runway inside has the homogeneity of gray scale and texture, cause the mistake coupling easily, produce the pole-climbing phenomenon.
3. when runway some areas are destroyed, use the relevant matches method to be difficult to accurately locate post position.
When image rotates, when convergent-divergent changes, the relevant matches method lost efficacy.
Therefore, at the special nature of airfield runway, the method identification runway target based on straight-line detection commonly used.At present, line detection algorithm mainly contains two big classes: 1. be by treatment of picture being obtained the frontier point set of target, utilizing the straight line on Hough transformation (Hough Transform is hereinafter to be referred as HT) the extraction object boundary then; 2. be after the image pre-service, directly obtain the boundary line set of target, in this set, carry out straight-line segment identification then.
1. method is the most classical line detection algorithm, is applied for patent by IBM soon by Paul Hough design proposition in 1962.The advantage of this method is to the insensitive for noise in the image, and indivedual non-boundary pixel points do not influence the straight-line detection result.Shortcoming is that calculated amount is bigger, is unfavorable for real-time application.2. method generally is earlier the object boundary travel direction to be encoded, and carries out line segment extraction then in the chain code set of strings that obtains.The advantage of this method is that calculated amount is little, and can obtain the information such as position, length, direction of straight-line segment simultaneously.Weak point is that algorithm performance is restricted by the track algorithm of object boundary, and is responsive to noise ratio.
Summary of the invention
The objective of the invention is in order to solve the defective that the airfield runway traditional detection disposal route of taking photo by plane at present exists, a kind of real-time detection method to airfield runway in the image of taking photo by plane is proposed, to improve accuracy of identification and the speed to the airfield runway position, the engineering of implementation method is used.
The present invention is to carry out according to the following steps under TMS320C6416 fixed-point dsp (TIX's production) system of 32bit fixed point, 1GHz clock to the real-time detection method of airfield runway in the image of taking photo by plane:
A. image pre-service:
At first image being done template size is that 3 * 3 mean filter cancelling noise disturbs, and uses the Suo Beier operator that image is carried out edge extracting then, by setting suitable threshold with image binaryzation.
B. the binary image travel direction is encoded:
When following the tracks of at the edge, for the ease of the information (as image coordinate, search mark or the like) of record marginal point, we adopt the mode of chained list to store marginal points information.Because storage space is very little in the digital signal processor (DSPs), if generate chained list in the dynamic assigning memory mode, the uncontrollable problem of crossing the border.The present invention reads in the image of DSPs to set up the list structure array, has just fixed the memory headroom size that needs after reading in image, thereby has effectively controlled the problem of crossing the border.List structure is
struct?node
{
unsigned?short?x；
unsigned?short?y；
unsigned?char?flag；
unsigned?char?gray；
struct?node*next；
}
Wherein, x, y are the coordinate figure of current point; Flag=0,1,2 ... 7 represent the corresponding codes direction respectively; Flag=8 represents to work as point and was not labeled, and flag=9 represents current some mark; Gray represents the gray-scale value of current point; Next is a pointer variable of pointing to the node type structure.
The present invention adopts famous Freeman chain code representation, and it is not a pointwise recording pixel coordinate, but the method for closure between the record neighbor.This coding can packed data and is provided convenience for subsequent treatment.The coding staff that Fig. 1 has shown pixel P neighborhood to, wherein pixel P is 0,1,2 to the direction of neighbor by counterclockwise sequential encoding ... 7.
In order all to search for 8 pixels of neighborhood at every turn, will encode with 1,5 is that diagonal line is divided into two parts up and down.Top is by 2,3, and 4,5 form, and the direction of search when the first half is done search is taken as+and 1; The latter half is by 6,7,0,1 forms, and the direction of search is taken as-1, when do not find pixel non-vanishing and that be not labeled in the latter half search procedure, then change the direction of search into 1, proceed the first half search, otherwise this point is charged to chained list, continue in its neighborhood, to keep original direction of search search next node; If do not find pixel non-vanishing and that be not labeled in the first half search equally, then the direction of search is changed into-1, proceed the latter half search, if 8 pixels all do not satisfy condition then the chain end of list (EOL).
C. determine the rough direction α of runway
For the chained list that obtains among the step b, at first reject length less than 10 chained list, in remaining chained list, add up the maximum coding of occurrence number in every chained list, then be the principal direction of this chained list.Add up occurrence number is maximum in the principal direction of all chained lists direction rough direction α as runway.0 ° of direction of 0,4 correspondence of wherein encoding; 1,5 corresponding 45 ° of directions; 2,6 corresponding 90 ° of directions; 3,7 corresponding 135 ° of directions.
D. in rough direction α neighborhood, use Hough transformation (HT) to detect parallel lines:
The principle that HT extracts straight line is in the two-dimensional coordinate plane, all process points (x, straight line y) can be described by ρ and θ parameter, and ρ denotation coordination initial point is to the distance of straight line, θ represents the vertical line of this straight line and the angle of X-axis, and its parametric equation is ρ=xcos θ+ysin θ.
Point on the straight line of image space, (ρ, θ) (ρ, θ), (ρ θ) in fact is exactly the number of impact point on certain direction to H to totalizer H to corresponding parameter space.Traditional HT method is all done from 0 to the 179 coordinate space conversion of spending to each pixel, and calculated amount is very big.The present invention only carries out the accurate location of straight line in [α-30 °, α+30 °] interval.Adopt layering thought simultaneously, calculate H (ρ, θ) value every 10 ° earlier, select maximum H (ρ, θ) the corresponding direction of value, then this direction ± 10 ° of intervals in every 1 ° of calculating H (ρ, θ) value is accurately located straight line, greatly reduces the calculated amount of direct detection.After the straight line of location, with the ρ of this straight line correspondence ± 9 neighborhoods in, θ ± H in 2 neighborhoods (ρ, θ) value is changed to 0, (ρ, θ) value is accurately located the second straight line to select maximum H.
The present invention is that the method that has adopted HT to combine with direction encoding realizes the airfield runway straight-line detection.Because the HT calculated amount is bigger, we at first adopt the direction encoding method that the image travel direction is scanned, obtain the information such as position, length, direction of straight-line segment, by adding up the main direction of all straight-line segments, determine the rough direction of runway, in the neighborhood of this direction, carry out the HT computing then, thereby dwindled the hunting zone of HT greatly, improved the accuracy rate and the arithmetic speed of algorithm.
Because HT is to the insensitive for noise in the image, indivedual non-boundary pixel points do not influence the straight-line detection result, even under the situation that the runway existence is blocked or many places are impaired, still can detect post position exactly.After position that detects runway and direction, by only calculating miss distance, rejected because runway inside has the homogeneity of gray scale and texture perpendicular to runway heading, cause the mistake coupling easily, produce the problem of pole-climbing phenomenon.When rotation at any angle, convergent-divergent take place, block in image, when illumination variation, the fuzzy interference, main straight line information in the image can not change, therefore can effectively overcome the various variations of image based on the runway recognition and tracking algorithm of straight-line detection, stable detection goes out target, solved when image rotates, when convergent-divergent changes, the problem that the relevant matches method lost efficacy.
Description of drawings
Fig. 1 is that the coding staff of this method pixel P neighborhood is to synoptic diagram;
Fig. 2 is the experimental result picture that given embodiment runway detects;
Wherein (a) is original-gray image, (b) is the bianry image behind the direction encoding, (c) is the pinpoint parallel lines of HT, (d) determines post position (+expression) for detection algorithm;
Fig. 3 is for existing the experimental result picture that blocks on the runway;
Wherein (a) (b) is the bianry image behind the direction encoding for adding the gray level image that blocks, and (c) is HT accurately located parallel straight line, (d) determines post position (+expression) for detection algorithm;
Fig. 4 is the experimental result picture of blurred picture;
Wherein (a) is fuzzy interfering picture; (b) be image behind 3 * 3 mean filters; (c) bianry image behind the direction encoding; (d) HT accurately located parallel straight line ,+expression target of attack point.
Embodiment
Below in conjunction with the embodiment that provides the inventive method is further elaborated.
To carrying out the real-time detection of runway, determine the position coordinates of runway from the runway image of taking photo by plane of internet.
The hardware environment that adopts: the TMS320C6416 fixed-point dsp of 32bit fixed point, 1GHz clock.
Software arrangements: Detect_Runway.c
Real-time detection method to airfield runway in the image of taking photo by plane, carry out as follows:
A. call in realtime graphic, the memory image gray-scale value, magnitude range is 0～255.
B. image is carried out pre-service.At first image being done template size is 3 * 3 mean filter, uses the Suo Beier operator that image is carried out edge extracting then, by setting suitable threshold with image binaryzation.
C. to binary image travel direction coding, the concrete practice is as follows:
1) set up the list structure array of bianry image,
struct?node
{
unsigned?short?x；
unsigned?short?y；
unsigned?char?flag；
unsigned?char?gray；
struct?node*next；
}
Wherein, x, y are the coordinate figure of current point; Flag=0,1,2 ... 7 represent the corresponding codes direction respectively; Flag=8 represents to work as point and was not labeled, and flag=9 represents current some mark; Gray represents the gray-scale value of current point; Next is a pointer variable of pointing to the node type structure.It is 8 that the flag initial value is set, and the next initial value is NULL.
2) to image from top to bottom, scanning from left to right, if find that gray-scale value is not 0, and the pixel A of flag=8, then set up a new chained list.Head pointer points to A, and flag is changed to 9, and putting A is current some C, puts direction of search s=-1.
3) seek next node B, counter n initial value is 0.If s=-1 skips to 4; Otherwise skip to 5.
4) be encoded to 6,7,0,1 pixel in the neighborhood according to counterclockwise direction search C.A point of every search B, counter n value adds 1.If flag is not 8, then continue search rest of pixels point; The gray-scale value of B is 0 else if, and then flag is changed to 9, continues remaining point of search, if the gray-scale value of B is not 0, then flag is changed to current corresponding codes, and the next pointed B of C is changed to current some C with B, skips to 3.If n=8, the next pointed NULL of C then, the chain end of list (EOL) skips to 2; Otherwise change direction of search s=1, skip to 5.
5) according to being encoded to 2,3,4,5 pixel in the neighborhood of counterclockwise searching for C.A point of every search B, counter n value adds 1.If flag is not 8, then continue search rest of pixels point; The gray-scale value of B is 0 else if, and then flag is changed to 9, continues remaining point of search, if the gray-scale value of B is not 0, then flag is changed to current corresponding codes, and the next pointed B of C is changed to current some C with B, skips to 3.If n=8, the next pointed NULL of C then, chain end of list (EOL); Otherwise change direction of search s=-1, skip to 4.
D. determine the rough direction of runway.At first rejecting length less than 10 chained list, in remaining chained list, add up the maximum coding of occurrence number in every chained list, then is the principal direction of this chained list.Add up occurrence number is maximum in the principal direction of all chained lists direction rough direction α as runway.
E. in the neighborhood of α, use HT to detect parallel lines.(each element initial value is 0 for ρ, array θ) to set up H.θ every 10 ° of variations, calculates the ρ value that each is not 0 pixel in [α-30 °, α+30 °] is interval, (ρ, θ) value adds 1 to while H.Select maximum H (ρ, θ) the corresponding θ of value, then θ ± calculate the ρ value that each is not 0 pixel every 1 ° in 10 ° of intervals, simultaneously H (ρ, θ) value adds 1, (ρ θ) is worth the accurate straight line of locating to select maximum H at last.After the straight line of location, with the ρ of this straight line correspondence ± 9 neighborhoods in, θ ± H in 2 neighborhoods (ρ, θ) value is changed to 0, (ρ, θ) value is accurately located the second straight line to select maximum H.
For validity and the real-time of verifying this algorithm, the present invention has carried out a large amount of experiments, provides experimental result below.
Fig. 2 has shown the experimental result that runway detects.(a) be original-gray image; (b) be bianry image behind the direction encoding, the rough direction of straight line that draws image is θ=45 °, and the image behind the direction encoding has been given prominence to the information of its cathetus as can be seen; (c) be the pinpoint parallel lines of HT; (d) determine post position (+expression) for detection algorithm.
Fig. 3 has shown the experimental result that existence is blocked on the runway.(a) for adding the gray level image that blocks; (b) be bianry image behind the direction encoding, block the main direction that can't influence runway behind the direction encoding as can be seen; (c) be HT accurately located parallel straight line; (d) determine post position (+expression) for detection algorithm.
Fig. 4 has shown the experimental result of blurred picture.(a) be fuzzy interfering picture; (b) be image behind the 3*3 mean filter; (c) bianry image behind the direction encoding; (d) HT accurately located parallel straight line ,+expression target of attack point, visible HT accurately detects straight line, is not subjected to the fuzzy influence of disturbing.
Aspect time performance, experiment drew in the direction encoding stage, and the method for the present invention's use is no more than 5ms operation time, has saved nearly 1 times of time than traditional whole searching algorithms of 8 neighborhoods; In the HT stage, layering thought of the present invention greatly reduces operation time of HT, guarantees that the time of accurately determining straight line is no more than 10ms.Therefore total algorithm realization time is no more than 5+10=15ms, for the image sampling speed of per second 25 frames, can satisfy the actual requirement of engineering of operation time less than 20ms.
From above-mentioned experimental result as can be seen, when rotation at any angle, convergent-divergent take place, block in image, when illumination variation, the fuzzy interference, main straight line information in the image can not change, therefore can effectively overcome the various variations of image based on the runway recognition and tracking algorithm of straight-line detection, stable detection goes out target.Algorithm has been applied to the engineering hardware platform at present, can satisfy real-time processing requirements.
Claims (1)
1. the real-time detection method to airfield runway in the image of taking photo by plane is characterized in that, carries out according to the following steps under the TMS320C6416 fixed-point dsp system of 32bit fixed point, 1GHz clock:
A. image pre-service:
At first image being done template size is that 3 * 3 mean filter cancelling noise disturbs, and uses the Suo Beier operator that image is carried out edge extracting then, by setting suitable threshold with image binaryzation;
B. the binary image travel direction is encoded:
Adopting list structure storage of array marginal points information, is 0,1,2,3,4,5,6,7 with the direction of 8 neighbors of pixel P by counterclockwise sequential encoding; And will to encode with 1,5 be that diagonal line is divided into up and down two parts, top is formed by 2,3,4,5, the direction of search of doing when search in the first half is taken as+and 1, the latter half is formed by 6,7,0,1, and the direction of search is taken as-1, when do not find pixel non-vanishing and that be not labeled in the latter half search procedure, then change the direction of search into 1, proceed the first half search, otherwise this point is charged to chained list, continue in its neighborhood, to keep original direction of search search next node; If do not find pixel non-vanishing and that be not labeled in the first half search equally, then the direction of search is changed into-1, proceed the latter half search, if 8 pixels all do not satisfy condition then the chain end of list (EOL);
C. determine the rough direction α of runway:
For the chained list that obtains among the step b, at first reject the chained list of length less than threshold value, in remaining chained list, add up the maximum coding of occurrence number in every chained list, then be the principal direction of this chained list, add up occurrence number is maximum in the principal direction of all chained lists direction rough direction α, 0, the 4 corresponding 0 ° of direction of wherein encoding as runway; 1,5 corresponding 45 ° of directions; 2,6 corresponding 90 ° of directions; 3,7 corresponding 135 ° of directions;
D. in rough direction α neighborhood, use Hough transformation to detect parallel lines:
Set up H (ρ according to Hough transformation parametric equation ρ=xcos θ+ysin θ, array θ), at [α-30 °, α+30 °] carry out the accurate location of straight line in the interval, calculate H (ρ every 10 ° earlier, θ) value is selected maximum H (ρ, θ) the corresponding direction of value, then this direction ± calculate H (ρ every 1 ° in 10 ° of intervals, θ) the accurate location straight line of value, after the straight line of location, with the ρ of this straight line correspondence ± 9 neighborhoods in, θ ± the interior H (ρ of 2 neighborhoods, θ) value is changed to 0, and (ρ, θ) value is accurately located the second straight line to select maximum H then.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007101935251A CN100552698C (en) | 2007-12-11 | 2007-12-11 | Real-time detection method to airfield runway in the image of taking photo by plane |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007101935251A CN100552698C (en) | 2007-12-11 | 2007-12-11 | Real-time detection method to airfield runway in the image of taking photo by plane |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101187981A CN101187981A (en) | 2008-05-28 |
| CN100552698C true CN100552698C (en) | 2009-10-21 |
Family
ID=39480363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2007101935251A Expired - Fee Related CN100552698C (en) | 2007-12-11 | 2007-12-11 | Real-time detection method to airfield runway in the image of taking photo by plane |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100552698C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961333A (en) * | 2018-06-21 | 2018-12-07 | 杭州晶智能科技有限公司 | Efficient calculation method for pixel area of image area |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011022784A (en) * | 2009-07-15 | 2011-02-03 | Sony Corp | Information processor, block detection method and program |
| CN102469323B (en) * | 2010-11-18 | 2014-02-19 | 深圳Tcl新技术有限公司 | Method for converting 2D (Two Dimensional) image to 3D (Three Dimensional) image |
| CN102496141B (en) * | 2011-12-12 | 2013-06-19 | 中国科学院长春光学精密机械与物理研究所 | Device used for transforming reference map into image being identical in view angle with real-time map |
| CN110132288B (en) * | 2019-05-08 | 2022-11-22 | 南京信息工程大学 | Micro vehicle vision navigation method for equal-width road surface |
| CN114419450A (en) * | 2022-03-29 | 2022-04-29 | 中国人民解放军96901部队 | Linear target damage efficiency rapid evaluation method based on image feature analysis |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006236055A (en) * | 2005-02-25 | 2006-09-07 | Victor Co Of Japan Ltd | Road recognition device and road recognition method |
| CN1979528A (en) * | 2005-12-02 | 2007-06-13 | 佳能株式会社 | Line detecting method and device |
-
2007
- 2007-12-11 CN CNB2007101935251A patent/CN100552698C/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006236055A (en) * | 2005-02-25 | 2006-09-07 | Victor Co Of Japan Ltd | Road recognition device and road recognition method |
| CN1979528A (en) * | 2005-12-02 | 2007-06-13 | 佳能株式会社 | Line detecting method and device |
Non-Patent Citations (2)
| Title |
|---|
| SAR图像中机场跑道的自动识别研究. 杨顺辽.武汉理工大学学报（交通科学与工程版），第30卷第1期. 2006 * |
| 基于直线特征的跑道识别与提取. 何巍，张平.中国航空学会飞行器控制与操纵第十次学术交流会暨专业委员会成立20周年大会论文集. 2001 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108961333A (en) * | 2018-06-21 | 2018-12-07 | 杭州晶智能科技有限公司 | Efficient calculation method for pixel area of image area |
| CN108961333B (en) * | 2018-06-21 | 2021-05-14 | 杭州晶一智能科技有限公司 | Efficient calculation method for pixel area of image area |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101187981A (en) | 2008-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhao et al. | Fusion of 3D LIDAR and camera data for object detection in autonomous vehicle applications | |
| CN100552698C (en) | Real-time detection method to airfield runway in the image of taking photo by plane | |
| CN106095104A (en) | Continuous gesture path dividing method based on target model information and system | |
| CN108491838A (en) | Pointer-type gauges registration read method based on SIFT and HOUGH | |
| Chen et al. | CGSANet: A contour-guided and local structure-aware encoder–decoder network for accurate building extraction from very high-resolution remote sensing imagery | |
| CN113920420A (en) | Building extraction method and device, terminal equipment and readable storage medium | |
| You et al. | OPD-Net: Prow detection based on feature enhancement and improved regression model in optical remote sensing imagery | |
| Zhu et al. | Arbitrary-oriented ship detection based on retinanet for remote sensing images | |
| Wang et al. | Fast vanishing point detection method based on road border region estimation | |
| Hu et al. | Generalized image recognition algorithm for sign inventory | |
| Zheng et al. | Learning orientation information from frequency-domain for oriented object detection in remote sensing images | |
| Kong et al. | Simultaneous Viewpoint-and Condition-Invariant Loop Closure Detection Based on LiDAR Descriptor for Outdoor Large-Scale Environments | |
| Naiemi et al. | MOSTL: An accurate multi-oriented scene text localization | |
| Han et al. | An effective method for bridge detection from satellite imagery | |
| Kim et al. | Improvement of traffic sign recognition by accurate ROI refinement | |
| Li et al. | Fast monocular visual place recognition for non-uniform vehicle speed and varying lighting environment | |
| CN102004921A (en) | Target identification method based on image characteristic analysis | |
| Yang et al. | Detection algorithm of airport runway in remote sensing images | |
| CN104574396A (en) | Straight line detection method and system | |
| Ren et al. | SAR image matching method based on improved SIFT for navigation system | |
| Ali et al. | SURF and LA with RGB vector space based detection and monitoring of manholes with an application to tri-rotor UAS images | |
| Geng et al. | Fast pedestrian detection using deformable part model and pyramid layer location | |
| Xie et al. | Building recognition and reconstruction from aerial imagery and lidar data | |
| Ji et al. | Accurate bounding box for ship detection on remote sensing images with complex background | |
| Hazelhoff et al. | Exploiting automatically generated databases of traffic signs and road markings for contextual co-occurrence analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20091021 Termination date: 20111211 |