CN103413144A - Airport detection and recognition method based on local global feature joint decision - Google Patents
Airport detection and recognition method based on local global feature joint decision Download PDFInfo
- Publication number
- CN103413144A CN103413144A CN2013103234660A CN201310323466A CN103413144A CN 103413144 A CN103413144 A CN 103413144A CN 2013103234660 A CN2013103234660 A CN 2013103234660A CN 201310323466 A CN201310323466 A CN 201310323466A CN 103413144 A CN103413144 A CN 103413144A
- Authority
- CN
- China
- Prior art keywords
- runway
- positive
- decision
- line segment
- negative sample
- 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
Abstract
The invention discloses an airport detection and recognition method based on a local global feature joint decision. The technical problem that an existing airport detection and recognition method is low in recognition rate is solved. According to the technical scheme, on the basis of the linear feature of an airport, image sharpening preprocessing, maximum between-cluster variance partition, mathematical morphology method noise removal, canny edge detection and Hough linear detection are adopted; a posterior probability classifier formed by a 2bitBP local feature and the relative similarity calculated through a nearest neighbor positive and negative sample set are used for determining a cluster threshold value of the distance between straight lines in a joint mode, and accordingly the influence of surrounding interference lines is removed; affine transformation is used, normalizing to a unified direction is achieved, an HOG feature is extracted, and the threshold value is input to an SVM classifier for making classification judgment. Due to the fact that the clustering method based on the local global feature joint decision removes the influence of plenty of interference lines, the defect that a suspected airport area is obtained directly is overcome, and the recognition rate of an airport area is improved.
Description
Technical field
The present invention relates to a kind of airport and detect recognition methods, particularly recognition methods is detected on a kind of airport based on local global characteristics joint decision.
Background technology
Airport is as strategic objective important in Aerial Images, automatically detects with identification in military field or civil area all has important using value and meaning to it.Existing Airport recognition method major part is based on the detection recognition methods to airfield runway.
Document " based on airport and the water-free bridge target identification research and implementation of remote sensing images, Xi ' an:Xidian University, 2010 " discloses a kind of airport and has detected recognition methods.The image of the method utilization after to binary segmentation carries out skeletal extraction, after edge extracting, with Hough, extract straight line again, all satisfactory straight-line segments are carried out the runway extension and are connected to determine doubtful traffic pattern, finally extract the architectural feature of doubtful traffic pattern, send into the SVM judgement of classifying.But, in definite doubtful traffic pattern stage, during due to the extraction straight line, easily be subject to disturbing the impact of line segment on every side, therefore for different airfield runways, can together with disturbing line segment on every side, be confirmed as doubtful traffic pattern, affect the effect of follow-up Airport recognition part, even can not identify airport target.
Summary of the invention
In order to overcome existing airport, detect the low deficiency of recognition methods discrimination, the invention provides a kind of airport based on local global characteristics joint decision and detect recognition methods.The method is at first according to the linear feature on airport, by image sharpening pre-service, maximum between-cluster variance cut apart, the Mathematical Morphology method removes noise, canny rim detection and Hough detection of straight lines, and the straight line detected carried out the straight line extension, connects and process; Secondly, utilize posterior probability sorter that the 2bitBP local feature forms and the positive and negative sample set of arest neighbors to calculate the determining of cluster threshold value that relative similarity is carried out distance between the joint decision straight-line segment, thereby around removing, disturb the impact of line segment; Last aftertreatment normalizes to unified direction and extracts the HOG feature by affined transformation, is input in the svm classifier device judgement of classifying and exports recognition result.Due to the clustering method based on the overall joint decision in part, remove the impact of a large amount of interference line segments, made up the defect that directly obtains doubtful traffic pattern, can improve the discrimination of traffic pattern.
The technical solution adopted for the present invention to solve the technical problems is: recognition methods is detected on a kind of airport based on local global characteristics joint decision, is characterized in comprising the following steps:
Step 1, input one width Aerial Images, select the USM sharpening method to carry out sharpening to Aerial Images, then use minimum in class, the OTSU of inter-class variance maximum is cut apart Image Segmentation Using after sharpening, image is divided into to target and background two classes, obtains the bianry image of the coarse segmentation on airport.To the bianry image of airport coarse segmentation, adopt the opening and closing operation of Mathematical Morphology method to carry out denoising.Use afterwards canny operator extraction edge, on this basis, the straight line that the Hough conversion obtains detecting.
Runway extends: runway straight-line segment runway is determined in conversion through Hough, extends from the two ends search of runway, makes runway near actual landing airdrome length.Its extension method is as follows:
1) calculate the slope slope of runway runway, and the starting point of acquiescence runway is above terminal;
2) runway is parallel to the y axle, from the starting point of runway and terminal respectively towards directly over and under search, until run into black pixel point;
3) runway is parallel to the x axle, from starting point and the terminal of runway, towards front-left and front-right, searches for respectively, until run into black pixel point;
4) if slope > 1,
1. initialization flag1=flag2=false;
2. from the runway starting point, start, towards the upper left side search, to run into white point, make flag1=true;
3. directly over the runway starting point starts court, search for, run into white point, make flag2=true;
If 4. flag1 and flag2 are true, select a point of the most close runway starting point to be assigned to starting point; If have one to be true, the white point found on correspondence direction be assigned to starting point;
If 5. in flag1 and flag2, at least one is true, turn to step 1.
6. adopt step method 1.-5. from the terminal of runway start to lower right and under search;
5) if the method for step 4) is adopted in 0<slope<1, the runway starting point is to upper left side and front-left search, and terminal is to lower right and front-right search;
6) if slope=1, the runway starting point to upper left side, terminal searches for to lower right, until run into black pixel point; If slope=-1, the runway starting point left below, terminal searches for to upper right side, until run into black pixel point;
7) if slope<0 and abs (slope)<1, the runway starting point is below and front-left search left, terminal is searched for to upper right side and front-right; If slope<0 and abs (slope) > 1, adopt the method for step 4), the runway starting point left below and under search, terminal to upper right side and directly over search.
Runway connects: after runway extended the processing end, the straight-line segment close and that slope is close of adjusting the distance connected, and concrete grammar is as follows:
1) for line segment L1, angle of inclination is θ
1, two end points are designated as respectively P1 and P2, find the line segment L2 that is total to the yardstick limit with some end points of line segment L1, and angle of inclination is θ
2, end points is designated as respectively P3 and P4;
2) ask the distance between line segment L1 and line segment L2 two two-end-points, i.e. Dist (P1, P2), Dist (P1, P4), Dist (P3, P2), Dist (P3, P4);
3) minimumly in 1: four distance of condition apart from Min (Dist), be less than threshold value dist,
Condition 2:| θ
1-θ
2|<threshold value threshold, threshold gets 5;
In 3: four distances of condition, maximum is greater than the longest length in two straight lines apart from max (Dist)
4) when step 3) is set up, connection L1, two line segments of L2 are a long line segment, delete simultaneously and participate in the line segment L1 and the line segment L2 that connect.
5) when step 3) is false, continue search, attachable line segment is arranged, get back to step 1); There is no attachable line segment, finish.
Step 2, utilize the method for cluster, by the method for local and global characteristics joint decision, determine best threshold value, thereby obtain the doubtful traffic pattern of optimum.
The local feature decision-making:
1) produce the positive and negative sample set of system, wherein the positive sample set of system comprises various airports template image, and the set of system negative sample comprises that other do not comprise the image of traffic pattern;
2) produce the position of all 2bitBP features, the traversal image-region, produce all feature locations;
3) utilize the Weak Classifier formation final strong classifier of Adaboost method from minute rate minimum of selecting all 2bitBP features to make mistakes;
The 2bitBP feature extraction of the fixed position that the doubtful zone 4) subset corresponding to certain cluster threshold value produced has obtained, judge the posterior probability of corresponding decimal code, if if be greater than 50% of each sorter weight summation, be input to next step global characteristics decision-making treatment, otherwise abandon this doubtful zone, as shown in formula (1), formula (2):
Wherein, P
PositiveBe the probability that doubtful zone is judged as target, SP (i) is the posterior probability that the i stack features is corresponding;
W (i) is the weight of sorter corresponding to i group posterior probability, and it obtains in Adaboost method training process.
By formula (2), judge whether doubtful zone is traffic pattern.
The global characteristics decision-making: this part produces that the positive and negative sample set of an arest neighbors is incompatible carries out decision-making, treats that by tolerance the doubtful traffic pattern of decision-making and the relative similarity of the positive and negative sample set of arest neighbors carry out decision-making.
Relative similarity is defined as follows:
dN=1-max(nccN) (4)
dP=1-max(nccP) (5)
Wherein, conf
RSFor relative similarity, nccP and nccN are respectively the distance of each sample in image block to be detected and positive sample set, negative sample set.
The generation step of the positive and negative sample set of arest neighbors is as follows, and training is input as the positive and negative sample set of system, is output as the positive and negative sample set of arest neighbors:
1) the positive sample set of system and negative sample collection are carried out to random alignment, the first frame is placed on to first position all the time by the original positive sample that the user chooses here;
2) each sample in the systematic sample sequence arranged is carried out to following operation:
Calculate the relative similarity of other samples in this sample and sequence, and process in two kinds of situation: if the predicted value of this sample is true, and relative similarity is less than threshold value, joins in the positive sample set of arest neighbors; If predicted value is false, relative similarity is greater than another threshold value, joins the set of arest neighbors negative sample.
3) the positive and negative sample set of output arest neighbors.
Decision part such as formula (6), save are the marks whether this doubtful zone can continue to be judged as traffic pattern, are that 1 expression still is doubtful zone, airport; Be 0 and be negative sample, add the set of system negative sample, carry out the renewal of the follow-up positive and negative sample set of system.
To retain by doubtful traffic pattern and its corresponding cluster threshold value of local feature decision-making and global characteristics decision-making, selecting the doubtful zone of relative similarity maximum in the global characteristics decision-making is the input of identifying processing, and its corresponding cluster threshold value is the optimal threshold for selecting.
Doubtful traffic pattern in step 3, input joint decision, normalize to unified direction by the affined transformation of corresponding point by doubtful traffic pattern, then extracts the HOG feature, is input in the svm classifier device trained the judgement of classifying.
The invention has the beneficial effects as follows: the method is at first according to the linear feature on airport, by image sharpening pre-service, maximum between-cluster variance cut apart, the Mathematical Morphology method is removed noise, canny rim detection and Hough detection of straight lines, and the straight line detected is carried out the straight line extension, connects and process; Secondly, utilize posterior probability sorter that the 2bitBP local feature forms and the positive and negative sample set of arest neighbors to calculate the determining of cluster threshold value that relative similarity is carried out distance between the joint decision straight-line segment, thereby around removing, disturb the impact of line segment; Last aftertreatment normalizes to unified direction and extracts the HOG feature by affined transformation, is input in the svm classifier device judgement of classifying and exports recognition result.Due to the clustering method based on the overall joint decision in part, remove the impact of a large amount of interference line segments, made up the defect that directly obtains doubtful traffic pattern, improved the discrimination of traffic pattern.
Below in conjunction with embodiment, the present invention is elaborated.
Embodiment
The airport detection recognition methods concrete steps that the present invention is based on local global characteristics joint decision are as follows:
1, extract runway.
Inputting a width Aerial Images, is at first image sharpening, selects existing USM sharpening method, then uses minimum in class, and the OTSU of inter-class variance maximum is cut apart Image Segmentation Using after sharpening, and image is divided into to target and background two classes, obtains the coarse segmentation result on airport.To the bianry image of airport coarse segmentation, adopt the opening and closing operation of Mathematical Morphology method to carry out denoising.Use afterwards canny operator extraction edge, on this basis, the straight line that the Hough conversion obtains detecting.
But due to affected by noise, the situation that exists airfield runway partly to rupture, therefore, need to further extend and be connected the runway straight-line segment detected.
Runway extends: runway straight-line segment runway is determined in conversion through Hough, extends but need to bring in search from two of runway, makes runway near actual landing airdrome length.Its extension method is as follows:
1) calculate the slope slope of runway runway, and the starting point of acquiescence runway is above terminal;
2) runway is parallel to the y axle, from the starting point of runway and terminal respectively towards directly over and under search, until run into black pixel point;
3) runway is parallel to the x axle, from starting point and the terminal of runway, towards front-left and front-right, searches for respectively, until run into black pixel point;
4) if slope > 1,
1. initialization flag1=flag2=false;
2. from the runway starting point, start, towards the upper left side search, to run into white point, make flag1=true;
3. directly over the runway starting point starts court, search for, run into white point, make flag2=true;
If 4. flag1 and flag2 are true, select a point of the most close runway starting point to be assigned to starting point; If have one to be true, the white point found on correspondence direction be assigned to starting point;
If 5. in flag1 and flag2, at least one is true, turn to 1.
6. from the terminal of runway start to lower right and under search, 1. similar-5.;
5) if 0<slope<1, the runway starting point is to the search of upper left side and front-left, terminal is to lower right and front-right search, similar step 4);
6) if slope=1, the runway starting point to upper left side, terminal searches for to lower right, until run into black pixel point; If slope=-1, the runway starting point left below, terminal searches for to upper right side, until run into black pixel point.
7) if slope<0 and abs (slope)<1, the runway starting point is below and front-left search left, terminal is searched for to upper right side and front-right; If slope<0 and abs (slope) > 1, the runway starting point left the below and under the search, terminal to upper right side and directly over the search, similar step 4).
Runway connects: after runway extended the processing end, the straight-line segment close and that slope is close of adjusting the distance connected, and concrete grammar is as follows:
1) for line segment L1, angle of inclination is θ
1, two end points are designated as respectively P1 and P2, find the line segment L2 that is total to the yardstick limit with some end points of line segment L1, and angle of inclination is θ
2, end points is designated as respectively P3 and P4;
2) ask the distance between line segment L1 and line segment L2 two two-end-points, i.e. Dist (P1, P2), Dist (P1, P4), Dist (P3, P2), Dist (P3, P4);
3) minimumly in 1: four distance of condition apart from Min (Dist), be less than threshold value dist,
Condition 2:| θ
1-θ
2|<threshold value threshold, threshold gets 5;
In 3: four distances of condition, maximum is greater than the longest length in two straight lines apart from max (Dist)
4) when step 3) is set up, connection L1, two line segments of L2 are a long line segment, delete simultaneously and participate in the line segment L1 and the line segment L2 that connect.
5) when step 3) is false, continue search, attachable line segment is arranged, get back to step 1); There is no attachable line segment, finish.
2, determine doubtful traffic pattern.
By the image that extracts runway output, be the image that comprises the airfield runway straight line, wherein be mingled with the impact of disturbing line segment on every side.Determine that doubtful traffic pattern will and disturb the line segment Region Segmentation open by real airfield runway zone, utilize the method for cluster, choosing of cluster threshold value travels through all cluster threshold values, method by local and global characteristics joint decision is determined best threshold value, thereby obtains optimum doubtful traffic pattern.
The local feature decision-making: it is characterized by existing 2bitBP feature, concrete feature selecting and decision-making technique are as follows:
1) produce the positive and negative sample set of system, wherein the positive sample set of system comprises various airports template image, and the set of system negative sample comprises that other do not comprise the image of traffic pattern;
2) produce the position of all 2bitBP features, the traversal image-region, produce all feature locations;
3) utilize 20 Weak Classifiers of Adaboost method from minute rate minimum of selecting all 2bitBP features to make mistakes, the form of sorter is that to add up the 2bitBP Feature Conversion of certain feature in all samples be to account for total probability after decimal code.These 20 Weak Classifiers form a final strong classifier;
The 2bitBP feature extraction of the fixed position that the doubtful zone 4) subset corresponding to certain cluster threshold value produced has obtained, judge the posterior probability of corresponding decimal code, if if be greater than 50% of each sorter weight summation, be input to next step global characteristics decision-making treatment, otherwise abandon this doubtful zone, as shown in formula (1), formula (2):
Wherein, P
PositiveBe the probability that doubtful zone is judged as target, SP (i) is the posterior probability that the i stack features is corresponding; W (i) is the weight of sorter corresponding to i group posterior probability, and it obtains in Adaboost method training process.
By formula (2), judge whether doubtful zone is traffic pattern.
The global characteristics decision-making: this part produces that the positive and negative sample set of an arest neighbors is incompatible carries out decision-making, treats that by tolerance the doubtful traffic pattern of plan and the relative similarity of the positive and negative sample set of arest neighbors carry out decision-making.
Relative similarity is defined as follows:
dN=1-max(nccN) (4)
dP=1-max(nccP) (5)
Conf wherein
RSFor relative similarity, nccP and nccN are respectively the distance of each sample in image block to be detected and positive sample set, negative sample set.The distance here is based on the distance of normalized crosscorrelation (Normalized Cross Correlation, (NCC)).
The generation step of the positive and negative sample set of arest neighbors is as follows, and training is input as the positive and negative sample set of system, is output as the positive and negative sample set of arest neighbors:
1) the positive sample set of system and negative sample collection are carried out to random alignment, the first frame is placed on to first position all the time by the original positive sample that the user chooses here;
2) each sample in the systematic sample sequence arranged is carried out to following operation:
Calculate the relative similarity of other samples in this sample and sequence, and process in two kinds of situation: if the predicted value of this sample is true, and relative similarity is less than threshold value (getting 0.65 here), joins in the positive sample set of arest neighbors; If predicted value is false, relative similarity is greater than another threshold value (0.5), joins the set of arest neighbors negative sample.
3) the positive and negative sample set of output arest neighbors.
Decision part such as formula (6), save are the marks whether this doubtful zone can continue to be judged as traffic pattern, are that 1 expression still is doubtful zone, airport; Be 0 and be negative sample, add the set of system negative sample, carry out the renewal of the follow-up positive and negative sample set of system.
To retain by doubtful traffic pattern and its corresponding cluster threshold value of local feature decision-making and global characteristics decision-making, selecting the doubtful zone of relative similarity maximum in the global characteristics decision-making is the input of identifying processing, and its corresponding cluster threshold value is the optimal threshold for selecting.
3, identifying processing.
Doubtful traffic pattern in this step input joint decision, normalize to unified direction by the affined transformation of corresponding point by doubtful traffic pattern, then extracts the HOG feature, is input in the svm classifier device trained the judgement of classifying.
Claims (1)
1. recognition methods is detected on the airport based on local global characteristics joint decision, it is characterized in that comprising the following steps:
Step 1, input one width Aerial Images, select the USM sharpening method to carry out sharpening to Aerial Images, then use minimum in class, the OTSU of inter-class variance maximum is cut apart Image Segmentation Using after sharpening, image is divided into to target and background two classes, obtains the bianry image of the coarse segmentation on airport; To the bianry image of airport coarse segmentation, adopt the opening and closing operation of Mathematical Morphology method to carry out denoising; Use afterwards canny operator extraction edge, on this basis, the straight line that the Hough conversion obtains detecting;
Runway extends: runway straight-line segment runway is determined in conversion through Hough, extends from the two ends search of runway, makes runway near actual landing airdrome length; Its extension method is as follows:
1) calculate the slope slope of runway runway, and the starting point of acquiescence runway is above terminal;
2) runway is parallel to the y axle, from the starting point of runway and terminal respectively towards directly over and under search, until run into black pixel point;
3) runway is parallel to the x axle, from starting point and the terminal of runway, towards front-left and front-right, searches for respectively, until run into black pixel point;
4) if slope > 1,
1. initialization flag1=flag2=false;
2. from the runway starting point, start, towards the upper left side search, to run into white point, make flag1=true;
3. directly over the runway starting point starts court, search for, run into white point, make flag2=true;
If 4. flag1 and flag2 are true, select a point of the most close runway starting point to be assigned to starting point; If have one to be true, the white point found on correspondence direction be assigned to starting point;
If 5. in flag1 and flag2, at least one is true, turn to step 1.
6. adopt step method 1.-5. from the terminal of runway start to lower right and under search;
5) if the method for step 4) is adopted in 0<slope<1, the runway starting point is to upper left side and front-left search, and terminal is to lower right and front-right search;
6) if slope=1, the runway starting point to upper left side, terminal searches for to lower right, until run into black pixel point; If slope=-1, the runway starting point left below, terminal searches for to upper right side, until run into black pixel point;
7) if slope<0 and abs (slope)<1, the runway starting point is below and front-left search left, terminal is searched for to upper right side and front-right; If slope<0 and abs (slope) > 1, adopt the method for step 4), the runway starting point left below and under search, terminal to upper right side and directly over search;
Runway connects: after runway extended the processing end, the straight-line segment close and that slope is close of adjusting the distance connected, and concrete grammar is as follows:
1) for line segment L1, angle of inclination is θ
1, two end points are designated as respectively P1 and P2, find the line segment L2 that is total to the yardstick limit with some end points of line segment L1, and angle of inclination is θ
2, end points is designated as respectively P3 and P4;
2) ask the distance between line segment L1 and line segment L2 two two-end-points, i.e. Dist (P1, P2), Dist (P1, P4), Dist (P3, P2), Dist (P3, P4);
3) minimumly in 1: four distance of condition apart from Min (Dist), be less than threshold value dist,
Condition 2:| θ
1-θ
2|<threshold value threshold, threshold gets 5;
In 3: four distances of condition, maximum is greater than the longest length in two straight lines apart from max (Dist)
4) when step 3) is set up, connection L1, two line segments of L2 are a long line segment, delete simultaneously and participate in the line segment L1 and the line segment L2 that connect;
5) when step 3) is false, continue search, attachable line segment is arranged, get back to step 1); There is no attachable line segment, finish;
Step 2, utilize the method for cluster, by the method for local and global characteristics joint decision, determine best threshold value, thereby obtain the doubtful traffic pattern of optimum;
The local feature decision-making:
1) produce the positive and negative sample set of system, wherein the positive sample set of system comprises various airports template image, and the set of system negative sample comprises that other do not comprise the image of traffic pattern;
2) produce the position of all 2bitBP features, the traversal image-region, produce all feature locations;
3) utilize the Weak Classifier formation final strong classifier of Adaboost method from minute rate minimum of selecting all 2bitBP features to make mistakes;
The 2bitBP feature extraction of the fixed position that the doubtful zone 4) subset corresponding to certain cluster threshold value produced has obtained, judge the posterior probability of corresponding decimal code, if if be greater than 50% of each sorter weight summation, be input to next step global characteristics decision-making treatment, otherwise abandon this doubtful zone, as shown in formula (1), formula (2):
Wherein, P
PositiveBe the probability that doubtful zone is judged as target, SP (i) is the posterior probability that the i stack features is corresponding;
W (i) is the weight of sorter corresponding to i group posterior probability, and it obtains in Adaboost method training process;
By formula (2), judge whether doubtful zone is traffic pattern;
The global characteristics decision-making: this part produces that the positive and negative sample set of an arest neighbors is incompatible carries out decision-making, treats that by tolerance the doubtful traffic pattern of decision-making and the relative similarity of the positive and negative sample set of arest neighbors carry out decision-making;
Relative similarity is defined as follows:
dN=1-max(nccN) (4)
dP=1-max(nccP) (5)
Wherein, conf
RSFor relative similarity, nccP and nccN are respectively the distance of each sample in image block to be detected and positive sample set, negative sample set;
The generation step of the positive and negative sample set of arest neighbors is as follows, and training is input as the positive and negative sample set of system, is output as the positive and negative sample set of arest neighbors:
1) the positive sample set of system and negative sample collection are carried out to random alignment, the first frame is placed on to first position all the time by the original positive sample that the user chooses here;
2) each sample in the systematic sample sequence arranged is carried out to following operation:
Calculate the relative similarity of other samples in this sample and sequence, and process in two kinds of situation: if the predicted value of this sample is true, and relative similarity is less than threshold value, joins in the positive sample set of arest neighbors; If predicted value is false, relative similarity is greater than another threshold value, joins the set of arest neighbors negative sample;
3) the positive and negative sample set of output arest neighbors;
Decision part such as formula (6), save are the marks whether this doubtful zone can continue to be judged as traffic pattern, are that 1 expression still is doubtful zone, airport; Be 0 and be negative sample, add the set of system negative sample, carry out the renewal of the follow-up positive and negative sample set of system;
To retain by doubtful traffic pattern and its corresponding cluster threshold value of local feature decision-making and global characteristics decision-making, selecting the doubtful zone of relative similarity maximum in the global characteristics decision-making is the input of identifying processing, and its corresponding cluster threshold value is the optimal threshold for selecting;
Doubtful traffic pattern in step 3, input joint decision, normalize to unified direction by the affined transformation of corresponding point by doubtful traffic pattern, then extracts the HOG feature, is input in the svm classifier device trained the judgement of classifying.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013103234660A CN103413144A (en) | 2013-07-29 | 2013-07-29 | Airport detection and recognition method based on local global feature joint decision |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2013103234660A CN103413144A (en) | 2013-07-29 | 2013-07-29 | Airport detection and recognition method based on local global feature joint decision |
Publications (1)
Publication Number | Publication Date |
---|---|
CN103413144A true CN103413144A (en) | 2013-11-27 |
Family
ID=49606151
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2013103234660A Pending CN103413144A (en) | 2013-07-29 | 2013-07-29 | Airport detection and recognition method based on local global feature joint decision |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103413144A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105740871A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | Method for detecting polygonal ground object by using line section space relationship |
CN105931216A (en) * | 2016-04-05 | 2016-09-07 | 李红伟 | Image processing technology-based airport pavement distress detection method |
CN106770967A (en) * | 2017-01-06 | 2017-05-31 | 重庆大学 | Electronic Nose non-targeted interference Gas Distinguishing Method based on a class local expression model |
CN107590154A (en) * | 2016-07-08 | 2018-01-16 | 阿里巴巴集团控股有限公司 | Object similarity decision method and device based on image recognition |
CN116385475A (en) * | 2023-06-06 | 2023-07-04 | 四川腾盾科技有限公司 | Runway identification and segmentation method for autonomous landing of large fixed-wing unmanned aerial vehicle |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101944233A (en) * | 2010-09-25 | 2011-01-12 | 西北工业大学 | Method for quickly extracting airport target in high-resolution remote sensing image |
CN102436644A (en) * | 2011-11-02 | 2012-05-02 | 南京物联网研究院发展有限公司 | Unstructured road detection method based on adaptive edge registration |
CN103162669A (en) * | 2013-03-01 | 2013-06-19 | 西北工业大学 | Detection method of airport area through aerial shooting image |
-
2013
- 2013-07-29 CN CN2013103234660A patent/CN103413144A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101944233A (en) * | 2010-09-25 | 2011-01-12 | 西北工业大学 | Method for quickly extracting airport target in high-resolution remote sensing image |
CN102436644A (en) * | 2011-11-02 | 2012-05-02 | 南京物联网研究院发展有限公司 | Unstructured road detection method based on adaptive edge registration |
CN103162669A (en) * | 2013-03-01 | 2013-06-19 | 西北工业大学 | Detection method of airport area through aerial shooting image |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105740871A (en) * | 2016-02-01 | 2016-07-06 | 福建师范大学 | Method for detecting polygonal ground object by using line section space relationship |
CN105931216A (en) * | 2016-04-05 | 2016-09-07 | 李红伟 | Image processing technology-based airport pavement distress detection method |
CN107590154A (en) * | 2016-07-08 | 2018-01-16 | 阿里巴巴集团控股有限公司 | Object similarity decision method and device based on image recognition |
CN107590154B (en) * | 2016-07-08 | 2021-03-02 | 阿里巴巴集团控股有限公司 | Object similarity determination method and device based on image recognition |
CN106770967A (en) * | 2017-01-06 | 2017-05-31 | 重庆大学 | Electronic Nose non-targeted interference Gas Distinguishing Method based on a class local expression model |
CN116385475A (en) * | 2023-06-06 | 2023-07-04 | 四川腾盾科技有限公司 | Runway identification and segmentation method for autonomous landing of large fixed-wing unmanned aerial vehicle |
CN116385475B (en) * | 2023-06-06 | 2023-08-18 | 四川腾盾科技有限公司 | Runway identification and segmentation method for autonomous landing of large fixed-wing unmanned aerial vehicle |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Nienaber et al. | Detecting potholes using simple image processing techniques and real-world footage | |
CN105335743A (en) | Vehicle license plate recognition method | |
CN102509291B (en) | Pavement disease detecting and recognizing method based on wireless online video sensor | |
CN103413144A (en) | Airport detection and recognition method based on local global feature joint decision | |
CN101038686B (en) | Method for recognizing machine-readable travel certificate | |
CN104036262B (en) | A kind of method and system of LPR car plates screening identification | |
Bailo et al. | Robust road marking detection and recognition using density-based grouping and machine learning techniques | |
Danti et al. | An image processing approach to detect lanes, pot holes and recognize road signs in Indian roads | |
CN101221623B (en) | Object type on-line training and recognizing method and system thereof | |
CN109086687A (en) | The traffic sign recognition method of HOG-MBLBP fusion feature based on PCA dimensionality reduction | |
CN101339616A (en) | Roads recognition method and apparatus | |
CN101727570B (en) | Tracking method, track detection processing unit and monitor system | |
CN105913041A (en) | Pre-marked signal lights based identification method | |
CN102902967A (en) | Method for positioning iris and pupil based on eye structure classification | |
CN104021574A (en) | Method for automatically identifying pavement diseases | |
CN102314599A (en) | Identification and deviation-detection method for lane | |
CN103366190A (en) | Method for identifying traffic sign | |
CN103136528A (en) | Double-edge detection based vehicle license plate identification method | |
CN109886168B (en) | Ground traffic sign identification method based on hierarchy | |
Sharma et al. | A hybrid technique for license plate recognition based on feature selection of wavelet transform and artificial neural network | |
CN102073852A (en) | Multiple vehicle segmentation method based on optimum threshold values and random labeling method for multiple vehicles | |
CN108108703A (en) | Deceleration strip missing detection method, device and electronic equipment | |
CN101246553B (en) | Cab model recognizing method based on triangulation transformation | |
CN109285169A (en) | A kind of roadside assistance equipment side coil based on wheel identification drags and leads abductive approach | |
CN109543498A (en) | A kind of method for detecting lane lines based on multitask network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20131127 |