CN102682304A - Multi-feature integrated passer-by detection method and device - Google Patents
Multi-feature integrated passer-by detection method and device Download PDFInfo
- Publication number
- CN102682304A CN102682304A CN2012100828465A CN201210082846A CN102682304A CN 102682304 A CN102682304 A CN 102682304A CN 2012100828465 A CN2012100828465 A CN 2012100828465A CN 201210082846 A CN201210082846 A CN 201210082846A CN 102682304 A CN102682304 A CN 102682304A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- characteristic
- zone
- gradient
- result
- 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
Images
Abstract
The invention discloses a multi-feature integrated passer-by detection method and device for the technical field of intelligent video monitoring. The method comprises the following steps of: utilizing a passers-by sample library in advance, extracting gradient feature and linear edge feature, integrating the gradient feature and linear edge feature into a new feature to be as input data, practicing and generating a passer-by classifier; according to the selected simple passer-by features, extracting a to-be-detected area to form an image processing result required by the feature; using the selected simple passer-by feature and the image processing result, removing a non-passer-by area in the to-be-detected area to obtain the initially located passer-by area; utilizing the passer-by classifier to judge and locate the passer-by to obtain the final passer-by detecting result. By adopting the method and the device, the passer-by detection speed is quicker and more accurate. From the perspective of the actual engineering application, the passer-by detection can be accurately detected through the monitoring video in real time.
Description
Technical field
The present invention relates to the intelligent video monitoring technical field, relate in particular to a kind of pedestrian detection method and device that merges many characteristics.
Background technology
Pedestrian detection based on video technique has market and application widely, occurs detecting like pedestrians disobeying traffic rule event detection in the intelligent transportation, pedestrian protection system, security area pedestrian's intrusion detection, expressway pedestrian in vehicle-mounted etc.
At present, based on the pedestrian detection method of video monitoring, mainly contain searching of pedestrian's profile and template matching method, people's face detection method, sorter judgement method.In practical application, these methods exist application scenarios not exist limitation, accuracy high or be difficult to the defective that reaches real-time, are difficult to be applied to complicated outdoor scene; Such as; Pedestrian's profile is sought and template matches, often receives the influence of target, illumination and weather etc. on every side, and accuracy of detection is lower; People's face detects and requires image detail comparatively clear, thereby causes the less target that can't detect than distant place in the large scene; Utilize pedestrian's sorter of the better performances of training gained, often because of feature extraction is comparatively complicated, the higher reason of sorter intrinsic dimensionality, computing is comparatively consuming time, is difficult to real-time use.
Summary of the invention
The purpose of the embodiment of the invention is the deficiency to existing pedestrian detection method based on video or image; And a kind of pedestrian detection method and device that merges many characteristics proposed; Make pedestrian detection rapid speed, accuracy higher; And from practical engineering application, can carry out accurately monitor video, real-time pedestrian detection.
In order to reach the foregoing invention purpose, a kind of pedestrian detection method that merges many characteristics that the embodiment of the invention proposes is realized through following technical scheme:
A kind of pedestrian detection method that merges many characteristics, said method comprises:
Utilize pedestrian's sample storehouse in advance, extract gradient characteristic and linear edge characteristic, and with said gradient characteristic and linear edge Feature Fusion be new feature as the input data, train generation pedestrian sorter;
According to selected simple pedestrian's characteristic, extract zone to be detected and form the required processing result image of this characteristic;
Utilize selected simple pedestrian's characteristic and said processing result image, get rid of and be not pedestrian's zone in the zone to be detected, obtain the pedestrian zone of Primary Location;
Utilize said pedestrian's sorter to carry out pedestrian's judgement and location, obtain final pedestrian detection result.
Further preferably, said according to selected simple pedestrian's characteristic, extract the required processing result image of zone this characteristic of formation to be detected and specifically comprise:
If adopt edge symmetry characteristic as preliminary pedestrian's characteristic, then extract the edge of image result, and/or, then extract the gradient result of image if adopt the histogram of gradients characteristic as preliminary pedestrian's characteristic, carry out preliminary Flame Image Process.
Further preferably, saidly utilize said pedestrian's grader to come it is carried out pedestrian judgement and location, obtain final pedestrian detection result and specifically comprise:
The pedestrian zone of expansion Primary Location, the convergent-divergent extended area is to normalize to the size of training sample, the gradient result and the linear edge result of the pedestrian's extended area after the extraction normalization;
The gradient result in the pedestrian zone in the structure pedestrian extended area is the gradient characteristic; Linear edge result in the structure extended area is the linear edge characteristic; And merge said gradient characteristic and linear edge is characterized as new feature; And utilize said pedestrian's sorter to come it is carried out pedestrian's judgement and location, obtain final pedestrian detection result.
Further preferably, said linear edge comprises line segment and/or can fit to the edge of line segment.
Further preferably, said gradient is characterized as various histogram of gradients.
In order to realize aforementioned goal of the invention, the embodiment of the invention has also proposed a kind of pedestrian detection device that merges many features, and said device is realized by the following technical programs:
A kind of pedestrian detection device that merges many characteristics, said device comprises:
The sorter generation module is used for utilizing in advance pedestrian's sample storehouse, extracts gradient characteristic and linear edge characteristic, and with said gradient characteristic and linear edge Feature Fusion be new feature as the input data, train generation pedestrian sorter;
Non-pedestrian's area filter module is used for not filtering and obtaining zone to be detected not containing the pedestrian zone in the input area;
Pedestrian's Primary Location module is used for treating surveyed area and carries out Flame Image Process according to selected simple pedestrian's characteristic, to obtain the pedestrian zone of Primary Location;
The final locating module of pedestrian is used for utilizing said pedestrian's sorter to carry out pedestrian's judgement and location, obtains final pedestrian detection result.
Further preferably, the final locating module of said pedestrian specifically comprises:
The feature extraction submodule is used for the pedestrian zone of Primary Location is expanded and size normalization, said extended area is extracted the gradient result and the region-wide extraction linear edge result in pedestrian zone;
The Feature Fusion submodule, the gradient result who is used for constructing the pedestrian zone of Primary Location be the gradient characteristic, the interior linear edge result of structure extended area is the linear edge characteristic, and merges said gradient characteristic and linear edge is characterized as new feature;
Pedestrian's locator module is used for according to said new feature, utilizes predefined pedestrian's sorter to adjudicate, and obtains meticulous pedestrian detection result.
Compared with prior art; The embodiment of the invention has reduced pedestrian precisely judgement and positioned area through preliminary processing and analysis; The gradient characteristic that the present invention the adopted interference of condition (illumination and weather etc.) to external world is insensitive; The edge feature that the embodiment of the invention adopted has been considered ubiquitous linear element structure on the rigidity target, can get rid of the interference of artificiality (like car, traffic lane line, road sign, lamp stand, barrier etc.) preferably, so detection speed is very fast, accuracy is higher.From the practical engineering application result, the present invention can carry out accurately monitor video, real-time pedestrian detection.
Description of drawings
Through the description of its exemplary embodiment being carried out below in conjunction with accompanying drawing, the above-mentioned feature and advantage of the present invention will become apparent and understand easily.
Fig. 1 is 1 one kinds of process flow diagrams that merge the pedestrian detection method of many characteristics of the embodiment of the invention;
Fig. 2 is the pedestrian detection device composition synoptic diagram that 2 one kinds of the embodiment of the invention merge many characteristics;
Fig. 3 is that the embodiment of the invention 3 another kind of pedestrian detection devices that merge many characteristics are formed synoptic diagram.
Embodiment
Below in conjunction with accompanying drawing the present invention is done further explain.
As shown in Figure 1, be 1 one kinds of pedestrian detection methods that merge many characteristics of the embodiment of the invention, said method comprises:
S101. utilize pedestrian's sample storehouse in advance, extract gradient characteristic and linear edge characteristic, and with said gradient characteristic and linear edge Feature Fusion be new feature as the input data, train generation pedestrian sorter;
S102. according to selected simple pedestrian's characteristic, extract zone to be detected and form the required processing result image of this characteristic;
S103. utilize selected simple pedestrian's characteristic and said processing result image, get rid of and be not pedestrian's zone in the zone to be detected, obtain the pedestrian zone of Primary Location;
S104. utilize said pedestrian's sorter to carry out pedestrian's judgement and location, obtain final pedestrian detection result.
Because with respect to background in the traffic scene and inhuman target; The pedestrian has comparatively obvious and simple characteristic and distinguishes; Through extracting and analyze this category feature; Can comparatively fast filter out a large amount of not being and confirm candidate's pedestrian zone in pedestrian's zone, and then carry out preliminary pedestrian's judgement and location.The embodiment of the invention can be through the interference of adopting each class methods (like the vertical gradient of background modeling and foreground segmentation, extraction and analysis image) and combination gets rid of fast the zone (like background area, vehicle region etc.) that comparatively significantly, does not have the pedestrian; The possible pedestrian zone that acquires present frame is with as zone to be detected; Purpose with the scope that reduces surveyed area also can detect full figure as zone to be detected.
But because the complicacy of scene; Only utilizing simple feature to accomplish detection can obtain a large amount of erroneous detection results and be difficult to obtain comparatively accurate pedestrian position; Therefore; Need adopt a kind of fusion manifold, more robust and sorter accurately in addition, come preliminary ruling and positioning result are further detected.
Therefore, it is following that the embodiment of the invention merges the pedestrian detection method of many characteristics:
Utilizing pedestrian's sample storehouse in advance, extract gradient characteristic and linear edge characteristic, is that a kind of new characteristic is as importing data, pedestrian's sorter of training generation with these two kinds of Feature Fusion.
Merge used gradient characteristic in manifold pedestrian detection method, its gradient that adopts can be the gradient of normal definitions, also can be for carrying out the gradient of various rectifications or mutation; The gradient characteristic that it adopts can be various histogram of gradients, but is not subject to histogram of gradients.
Merge used linear edge characteristic in manifold pedestrian detection method, its linear edge can be to adopt the line segments extraction method to extract the line segment that obtains, and also can be the edge of the approximate line segment that institute's match obtains behind the extraction edge; The calculating of its linear edge characteristic is to be used formed through the various information (like the direction value at line segment edge, the length of line segment, the position of line segment etc.) with linear edge.
The gradient characteristic is merged mutually (can adopt head and the tail to connect or other multiple modes) with the linear edge characteristic; Form a kind of new characteristic as the input data; Adopt various learning methods (like Boosting method, SVM method, decision tree classification method etc.) to carry out the automatic study of pedestrian's sorter, produce pedestrian's sorter model.
Secondly, extract the characteristic that zone to be detected is comparatively simple, be used for pedestrian's judgement, utilize said characteristic to treat detection zone and carry out preliminary judgement of pedestrian zone and location;
Then; To various features such as its gradient of candidate pedestrian's extracted region of Primary Location, linear edge; And fusion becomes a kind of characteristic of the pedestrian's of being used for judgement; Utilize again to pedestrian's sorter that this characteristic trained to come it is carried out meticulous pedestrian's judgement and location, obtain final pedestrian detection result.
Wherein, For confirming of pedestrian to be detected zone; Can full figure or specific zone be detected as pedestrian to be detected zone, also can obtain the vertical gradient and the next zone that comparatively significantly, does not have the pedestrian of getting rid of fast of combination thereof of foreground extraction and analysis image like the mode of background modeling through adopting each class methods; Interference like the shade of background area, vehicle region, motion, tiny target etc.; Acquire the possible pedestrian zone of present frame, the purpose with the scope that reduces surveyed area also can detect full figure as zone to be detected.
The following mode of pedestrian's zone passage to be detected is carried out preliminary judgement of pedestrian's characteristic and location: based on adopt simple pedestrian's characteristic of adjudicating; Extract zone to be detected and form the required processing result image of this characteristic; Such as; Adopt edge symmetry characteristic as simple pedestrian's characteristic, then extract the edge of image result; Adopt the histogram of gradients characteristic as simple pedestrian's characteristic, then extract the gradient result of image; Whether construct a plurality of pedestrians zone in the zone to be detected, utilize processing result image to form preliminary pedestrian's characteristic in this pedestrian zone, adjudicating it is the pedestrian, obtains preliminary pedestrian's judgement and positioning result.
Can multiple comparatively simple, the characteristic that can be easier to carry out rough pedestrian's judgement of pedestrian's extracted region to be detected be come the zone is carried out pedestrian's search, filtered according to the existing characteristic of pedestrian again, and preliminary judgement and location.Distribute and vertical sobel edge result such as, prospect capable of using, and pedestrian's vertical local edge (like symmetry, have a pair of above peak value etc.) carries out preliminary pedestrian's judgement and location; Can extract the HoG characteristic that the gradient result forms low dimension, utilize sorter to carry out preliminary pedestrian's judgement and location; The method of looking for the pedestrian head profile in the article of " Tracking multiple human in complex situations " that Zhao Tao capable of using delivered in 2004, by name is carried out preliminary pedestrian's judgement and location.
Can further filter the preliminary ruling result,,, but not be subject to this processing, also can not filter as pedestrian's preliminary ruling and positioning result to keep several zones with a high credibility.
Preferred embodiment down, can be through adopting the pedestrian dummy of different size size, be divided into a plurality of pedestrians zone of repeatedly constructing in the surveyed area; Also can acquire standard pedestrian's size results of priori through the method for demarcating, utilize standard pedestrian size to construct a plurality of pedestrians zone in the zone to be detected, to reduce operand.
Behind the pedestrian zone after acquiring Primary Location, carry out meticulous pedestrian's judgement and location in the following manner: the pedestrian of expansion Primary Location is regional, and pedestrian's extended area is carried out size normalization; The gradient result in the pedestrian zone of Primary Location in pedestrian's extended area after the extraction normalization, said gradient result can be gradient result original or that correct; Pedestrian's extended area after the normalization is extracted the linear edge result; Structure gradient result is the gradient characteristic, and linear edge result is the linear edge characteristic, and these two kinds of Feature Fusion are become new feature; The input new feature utilizes the pedestrian's sorter that has trained to come its judgement is obtained final pedestrian detection result.Its detailed process is following:
Pedestrian zone with Primary Location is the center, by the positive sample characteristics of training pedestrian sorter, suitable expansion (it is wide high 1.0 times to 3.0 times generally to expand to former pedestrian zone, and wide and high two directions can adopt different extensive ratio) is carried out in this zone.
According to the extended area size, expansion area field width and high extensive ratio and the pairing sample image size of training pedestrian sorter are carried out convergent-divergent with pedestrian's extended area, thereby are realized the pedestrian's extended area size normalization purpose with Primary Location.
The gradient result in the pedestrian zone in the pedestrian's extended area behind the calculating convergent-divergent forms the gradient characteristic;
Linear edge result in pedestrian's extended area behind the extraction convergent-divergent forms the linear edge characteristic; Gradient characteristic and the linear edge Feature Fusion extracted are formed new pedestrian's characteristic, utilize the pedestrian's sorter that has trained to come it is adjudicated, keep court verdict.
Judgement is handled for people's pedestrian zone to pedestrian's sorter, acquires the pedestrian detection result.In the situation of practical implementation, can be separated by the pedestrian zone nearer through removal in confidence level lower, the pedestrian zone that the screening back is remained is as the pedestrian detection result; Also can merge through a plurality of pedestrians zone that will adjudicate to the pedestrian, with amalgamation result as the pedestrian detection result; After pedestrian's sorter is adjudicated, obtain final pedestrian result, the embodiment of the invention does not limit concrete implementation method.
Wherein, said gradient characteristic, its gradient that adopts can be the gradients of normal definitions, also can be for carrying out the gradient of various rectifications or mutation; The gradient characteristic that it adopts can be various histogram of gradients, but is not subject to histogram of gradients.
Further preferably, merge the linear edge in manifold pedestrian detection method, be meant that line segment maybe can fit to the edge of line segment.
Said linear edge can be to adopt the line segments extraction method to extract the line segment that obtains, and also can be the edge of the approximate line segment that institute's match obtains behind the extraction edge; The calculating of its linear edge characteristic is to be used formed through the various information (like the direction value at line segment edge, the length of line segment, the position of line segment etc.) with linear edge.
As shown in Figure 2, in order to realize aforementioned goal of the invention, the embodiment of the invention 2 has also proposed a kind of pedestrian detection device that merges many characteristics, and said device is realized through following technical scheme:
A kind of pedestrian detection device that merges many characteristics, said device comprises:
The sorter generation module is used for utilizing in advance pedestrian's sample storehouse, extracts gradient characteristic and linear edge characteristic, and with said gradient characteristic and linear edge Feature Fusion be new feature as the input data, train generation pedestrian sorter;
Non-pedestrian's area filter module is used for not filtering and obtaining zone to be detected not containing the pedestrian zone in the input area;
Pedestrian's Primary Location module is used for treating surveyed area and carries out Flame Image Process according to selected simple pedestrian's characteristic, to obtain the pedestrian zone of Primary Location;
The final locating module of pedestrian is used for utilizing said pedestrian's sorter to carry out pedestrian's judgement and location, obtains final pedestrian detection result.
As shown in Figure 3, further preferably, in the embodiment of the invention 3, the final locating module of said pedestrian specifically comprises:
The feature extraction submodule is used for the pedestrian zone of Primary Location is expanded and size normalization, said extended area is extracted the gradient result and the region-wide extraction linear edge result in pedestrian zone;
The Feature Fusion submodule, the gradient result who is used for constructing the pedestrian zone of Primary Location be the gradient characteristic, the interior linear edge result of structure extended area is the linear edge characteristic, and merges said gradient characteristic and linear edge is characterized as new feature;
Pedestrian's locator module is used for according to said new feature, utilizes predefined pedestrian's sorter to adjudicate, and obtains meticulous pedestrian detection result.
The enforcement of said device such as method embodiment description, do not give unnecessary details one by one here.
Compared with prior art; The embodiment of the invention has reduced pedestrian precisely judgement and positioned area through preliminary processing and analysis; The gradient characteristic that the present invention the adopted interference of condition (illumination and weather etc.) to external world is insensitive; The edge feature that the embodiment of the invention adopted has been considered ubiquitous linear element structure on the rigidity target, can get rid of the interference of artificiality (like car, lamp stand, barrier etc.) preferably, so detection speed is very fast, accuracy is higher.From the practical engineering application result, the present invention can carry out accurately monitor video, real-time pedestrian detection.
One of ordinary skill in the art of the present invention are appreciated that; The above embodiment of the present invention is merely one of the preferred embodiments of the present invention; Be the length restriction; Here can not all embodiments of particularize, any enforcement that can embody claim technical scheme of the present invention is all in protection scope of the present invention.
It should be noted that; Above content is to combine concrete embodiment to further explain that the present invention did; Can not assert that embodiment of the present invention only limits to this; Under above-mentioned guidance of the present invention, those skilled in the art can carry out various improvement and distortion on the basis of the foregoing description, and these improve or distortion drops in protection scope of the present invention.
Claims (7)
1. pedestrian detection method that merges many characteristics is characterized in that said method comprises:
Utilize pedestrian's sample storehouse in advance, extract gradient characteristic and linear edge characteristic, and with said gradient characteristic and linear edge Feature Fusion be new feature as the input data, train generation pedestrian sorter;
According to selected simple pedestrian's characteristic, extract zone to be detected and form the required processing result image of this characteristic;
Utilize selected simple pedestrian's characteristic and said processing result image, get rid of and be not pedestrian's zone in the zone to be detected, obtain the pedestrian zone of Primary Location;
Utilize said pedestrian's sorter to carry out pedestrian's judgement and location, obtain final pedestrian detection result.
2. the pedestrian detection method of the many characteristics of fusion as claimed in claim 1 is characterized in that, and is said according to selected simple pedestrian's characteristic, extracts the required processing result image of zone this characteristic of formation to be detected and specifically comprises:
If adopt edge symmetry characteristic as preliminary pedestrian's characteristic, then extract the edge of image result, and/or, then extract the gradient result of image if adopt the histogram of gradients characteristic as preliminary pedestrian's characteristic, carry out preliminary Flame Image Process.
3. the pedestrian detection method of the many characteristics of fusion as claimed in claim 1 is characterized in that, saidly utilizes said pedestrian's sorter to come it is carried out pedestrian judgement and location, obtains final pedestrian detection result and specifically comprises:
The pedestrian zone of expansion Primary Location, the convergent-divergent extended area is to normalize to the size of training sample, the gradient result and the linear edge result of the pedestrian's extended area after the extraction normalization;
The gradient result in the pedestrian zone in the structure pedestrian extended area is the gradient characteristic; Linear edge result in the structure extended area is the linear edge characteristic; And merge said gradient characteristic and linear edge is characterized as new feature; And utilize said pedestrian's sorter to come it is carried out pedestrian's judgement and location, obtain final pedestrian detection result.
4. like the pedestrian detection method of any many characteristics of described fusion of claim 1 to 3, it is characterized in that said linear edge comprises line segment and/or can fit to the edge of line segment.
5. like the pedestrian detection method of any many characteristics of described fusion of claim 1 to 3, it is characterized in that said gradient is characterized as various histogram of gradients.
6. pedestrian detection device that merges many characteristics is characterized in that said device comprises:
The sorter generation module is used for utilizing in advance pedestrian's sample storehouse, extracts gradient characteristic and linear edge characteristic, and with said gradient characteristic and linear edge Feature Fusion be new feature as the input data, train generation pedestrian sorter;
Non-pedestrian's area filter module is used for not filtering and obtaining zone to be detected not containing the pedestrian zone in the input area;
Pedestrian's Primary Location module is used for treating surveyed area and carries out Flame Image Process according to selected simple pedestrian's characteristic, to obtain the pedestrian zone of Primary Location;
The final locating module of pedestrian is used for utilizing said pedestrian's sorter to carry out pedestrian's judgement and location, obtains final pedestrian detection result.
7. the pedestrian detection device of the many characteristics of fusion as claimed in claim 6 is characterized in that, the final locating module of said pedestrian specifically comprises:
The feature extraction submodule is used for the pedestrian zone of Primary Location is expanded and size normalization, said extended area is extracted the gradient result and the region-wide extraction linear edge result in pedestrian zone;
The Feature Fusion submodule, the gradient result who is used for constructing the pedestrian zone of Primary Location be the gradient characteristic, the interior linear edge result of structure extended area is the linear edge characteristic, and merges said gradient characteristic and linear edge is characterized as new feature;
Pedestrian's locator module is used for according to said new feature, utilizes predefined pedestrian's sorter to adjudicate, and obtains meticulous pedestrian detection result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012100828465A CN102682304A (en) | 2012-03-26 | 2012-03-26 | Multi-feature integrated passer-by detection method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012100828465A CN102682304A (en) | 2012-03-26 | 2012-03-26 | Multi-feature integrated passer-by detection method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102682304A true CN102682304A (en) | 2012-09-19 |
Family
ID=46814199
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2012100828465A Pending CN102682304A (en) | 2012-03-26 | 2012-03-26 | Multi-feature integrated passer-by detection method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102682304A (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103106409A (en) * | 2013-01-29 | 2013-05-15 | 北京交通大学 | Composite character extraction method aiming at head shoulder detection |
CN103559482A (en) * | 2013-11-05 | 2014-02-05 | 无锡慧眼电子科技有限公司 | Pedestrian detection method based on edge symmetry |
CN103745196A (en) * | 2013-12-27 | 2014-04-23 | 东软集团股份有限公司 | Method and apparatus of generalized pedestrian detection |
CN104318760A (en) * | 2014-09-16 | 2015-01-28 | 北方工业大学 | Crossing violation behavior intelligent detection method and system based on analog model |
CN104504394A (en) * | 2014-12-10 | 2015-04-08 | 哈尔滨工业大学深圳研究生院 | Dese population estimation method and system based on multi-feature fusion |
CN104700114A (en) * | 2013-12-09 | 2015-06-10 | 现代自动车株式会社 | Method and device for recognizing pedestrian and vehicle supporting the same |
CN105243356A (en) * | 2015-09-10 | 2016-01-13 | 北京大学 | Method of building pedestrian detection model and device and pedestrian detection method |
CN106295460A (en) * | 2015-05-12 | 2017-01-04 | 株式会社理光 | The detection method of people and equipment |
CN106570899A (en) * | 2015-10-08 | 2017-04-19 | 腾讯科技(深圳)有限公司 | Target object detection method and device |
CN107255818A (en) * | 2017-06-13 | 2017-10-17 | 厦门大学 | A kind of submarine target quick determination method of bidimensional multiple features fusion |
CN107992815A (en) * | 2017-11-28 | 2018-05-04 | 北京小米移动软件有限公司 | Eyeglass detection method and device |
CN108197622A (en) * | 2017-12-26 | 2018-06-22 | 新智数字科技有限公司 | A kind of detection method of license plate, device and equipment |
CN108520261A (en) * | 2018-03-01 | 2018-09-11 | 中国农业大学 | A kind of recognition methods of peanut kernels quantity and device |
CN109598301A (en) * | 2018-11-30 | 2019-04-09 | 腾讯科技(深圳)有限公司 | Detection zone minimizing technology, device, terminal and storage medium |
CN109766756A (en) * | 2018-12-10 | 2019-05-17 | 中国科学院自动化研究所南京人工智能芯片创新研究院 | Make a dash across the red light data processing method, device, computer equipment and storage medium |
CN113518201A (en) * | 2020-07-14 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Video processing method, device and equipment |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101551852A (en) * | 2008-04-01 | 2009-10-07 | 索尼(中国)有限公司 | Training system, training method and detection method |
CN101650728A (en) * | 2009-08-26 | 2010-02-17 | 北京邮电大学 | Video high-level characteristic retrieval system and realization thereof |
CN101887524A (en) * | 2010-07-06 | 2010-11-17 | 湖南创合制造有限公司 | Pedestrian detection method based on video monitoring |
WO2011033657A1 (en) * | 2009-09-18 | 2011-03-24 | 株式会社東芝 | Feature extracting device |
CN102043953A (en) * | 2011-01-27 | 2011-05-04 | 北京邮电大学 | Real-time-robust pedestrian detection method aiming at specific scene |
-
2012
- 2012-03-26 CN CN2012100828465A patent/CN102682304A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101551852A (en) * | 2008-04-01 | 2009-10-07 | 索尼(中国)有限公司 | Training system, training method and detection method |
CN101650728A (en) * | 2009-08-26 | 2010-02-17 | 北京邮电大学 | Video high-level characteristic retrieval system and realization thereof |
WO2011033657A1 (en) * | 2009-09-18 | 2011-03-24 | 株式会社東芝 | Feature extracting device |
CN101887524A (en) * | 2010-07-06 | 2010-11-17 | 湖南创合制造有限公司 | Pedestrian detection method based on video monitoring |
CN102043953A (en) * | 2011-01-27 | 2011-05-04 | 北京邮电大学 | Real-time-robust pedestrian detection method aiming at specific scene |
Non-Patent Citations (2)
Title |
---|
姚雪琴等: "《基于边缘对称性和HOG 的行人检测方法》", 《计算机工程》 * |
程广涛等: "《基于HOG 特征的行人视觉检测方法》", 《传感器与微系统》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103106409A (en) * | 2013-01-29 | 2013-05-15 | 北京交通大学 | Composite character extraction method aiming at head shoulder detection |
CN103559482A (en) * | 2013-11-05 | 2014-02-05 | 无锡慧眼电子科技有限公司 | Pedestrian detection method based on edge symmetry |
CN104700114A (en) * | 2013-12-09 | 2015-06-10 | 现代自动车株式会社 | Method and device for recognizing pedestrian and vehicle supporting the same |
US9519831B2 (en) | 2013-12-27 | 2016-12-13 | Neusoft Corporation | Method and apparatus for detecting generalized passerby by utilizing features of a wheel and upper body |
CN103745196A (en) * | 2013-12-27 | 2014-04-23 | 东软集团股份有限公司 | Method and apparatus of generalized pedestrian detection |
CN103745196B (en) * | 2013-12-27 | 2016-10-05 | 东软集团股份有限公司 | Broad sense pedestrian detection method and device |
CN104318760A (en) * | 2014-09-16 | 2015-01-28 | 北方工业大学 | Crossing violation behavior intelligent detection method and system based on analog model |
CN104318760B (en) * | 2014-09-16 | 2016-08-24 | 北方工业大学 | Crossing violation behavior intelligent detection method and system based on analog model |
CN104504394A (en) * | 2014-12-10 | 2015-04-08 | 哈尔滨工业大学深圳研究生院 | Dese population estimation method and system based on multi-feature fusion |
CN106295460A (en) * | 2015-05-12 | 2017-01-04 | 株式会社理光 | The detection method of people and equipment |
CN106295460B (en) * | 2015-05-12 | 2019-05-03 | 株式会社理光 | The detection method and equipment of people |
CN105243356A (en) * | 2015-09-10 | 2016-01-13 | 北京大学 | Method of building pedestrian detection model and device and pedestrian detection method |
CN105243356B (en) * | 2015-09-10 | 2019-02-12 | 北京大学 | A kind of method and device that establishing pedestrian detection model and pedestrian detection method |
CN106570899A (en) * | 2015-10-08 | 2017-04-19 | 腾讯科技(深圳)有限公司 | Target object detection method and device |
CN107255818A (en) * | 2017-06-13 | 2017-10-17 | 厦门大学 | A kind of submarine target quick determination method of bidimensional multiple features fusion |
CN107992815A (en) * | 2017-11-28 | 2018-05-04 | 北京小米移动软件有限公司 | Eyeglass detection method and device |
CN108197622A (en) * | 2017-12-26 | 2018-06-22 | 新智数字科技有限公司 | A kind of detection method of license plate, device and equipment |
CN108520261A (en) * | 2018-03-01 | 2018-09-11 | 中国农业大学 | A kind of recognition methods of peanut kernels quantity and device |
CN108520261B (en) * | 2018-03-01 | 2021-06-18 | 中国农业大学 | Method and device for identifying peanut kernel number |
CN109598301A (en) * | 2018-11-30 | 2019-04-09 | 腾讯科技(深圳)有限公司 | Detection zone minimizing technology, device, terminal and storage medium |
CN109766756A (en) * | 2018-12-10 | 2019-05-17 | 中国科学院自动化研究所南京人工智能芯片创新研究院 | Make a dash across the red light data processing method, device, computer equipment and storage medium |
CN113518201A (en) * | 2020-07-14 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Video processing method, device and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102682304A (en) | Multi-feature integrated passer-by detection method and device | |
Jensen et al. | Vision for looking at traffic lights: Issues, survey, and perspectives | |
US9953245B2 (en) | Multi-camera vehicle identification system | |
CN108090429B (en) | Vehicle type recognition method for graded front face bayonet | |
US9685079B2 (en) | Short-time stopping detection from red light camera evidentiary photos | |
CN103324930B (en) | A kind of registration number character dividing method based on grey level histogram binaryzation | |
CN105702048B (en) | Highway front truck illegal road occupation identifying system based on automobile data recorder and method | |
CN103279756B (en) | Vehicle detection based on integrated classifier analyzes system and determination method thereof | |
CN105160309A (en) | Three-lane detection method based on image morphological segmentation and region growing | |
CN103164958B (en) | Method and system for vehicle monitoring | |
Salarian et al. | A vision based system for traffic lights recognition | |
CN105184301B (en) | A kind of method that vehicle heading is differentiated using four-axle aircraft | |
CN104463232A (en) | Density crowd counting method based on HOG characteristic and color histogram characteristic | |
Chen et al. | A precise information extraction algorithm for lane lines | |
Ketcham et al. | Recognizing the Illegal Parking Patterns of Cars on the Road in Front of the Bus Stop Using the Support Vector Machine | |
Satzoda et al. | Robust extraction of lane markings using gradient angle histograms and directional signed edges | |
Park et al. | Vision-based surveillance system for monitoring traffic conditions | |
Coronado et al. | Detection and classification of road signs for automatic inventory systems using computer vision | |
CN104463238A (en) | License plate recognition method and system | |
Wen et al. | A rear-vehicle detection system for static images based on monocular vision | |
Priambada et al. | Levensthein distance as a post-process to improve the performance of ocr in written road signs | |
Deng et al. | An integrated forward collision warning system based on monocular vision | |
Borkar et al. | Detecting lane markers in complex urban environments | |
CN105206060A (en) | Vehicle type recognition device and method based on SIFT characteristics | |
Joshy et al. | Improved detection and tracking of lane marking using Hough transform |
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: 20120919 |