CN102855500A - Haar and HoG characteristic based preceding car detection method - Google Patents
Haar and HoG characteristic based preceding car detection method Download PDFInfo
- Publication number
- CN102855500A CN102855500A CN2011101751678A CN201110175167A CN102855500A CN 102855500 A CN102855500 A CN 102855500A CN 2011101751678 A CN2011101751678 A CN 2011101751678A CN 201110175167 A CN201110175167 A CN 201110175167A CN 102855500 A CN102855500 A CN 102855500A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- haar
- strong classifier
- hog
- feature
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a Haar and HoG characteristic based preceding car detection method. The method includes: 1), manually selecting a large number of vehicular pictures and non-vehicular pictures as positive samples and negative samples of a training set, and formatting the positive samples and the negative samples to be under 24X24 pixel; 2), using Haar characteristics and HoG characteristics to respectively characterize each of the positive samples and the negative samples to form characteristic vectors; 3) respectively establishing a weak classifier aiming at two characteristic vectors formed according to the Haar characteristics and the HoG characteristics; 4), utilizing the cascaded Adaboost algorithm to train the weak classifiers to obtain a cascaded vehicular strong classifier; and 5), inputting sub-images of different sizes and different positions into the cascaded vehicular strong classifier for judgment according to preceding road video images obtained by a vehicular camera, and judging the sub-images where a vehicle is positioned to be a preceding vehicle. The Haar and HoG characteristic based preceding car detection method is good in instantaneity and high in robustness, and has active influences on guaranteeing vehicular safe traveling, human and property safety.
Description
Technical field
The invention belongs to area of pattern recognition, relate to the automobile active safety field, be specifically related to a kind of front truck detection method based on Haar and HoG feature.
Background technology
Be accompanied by popularizing of the progress of urbanization and automobile, the increasing severely of traffic hazard, and caused a large amount of personnel's injury and economic loss.How to have ensured vehicle on highway safe, travel fast, avoid the generation of traffic hazards such as knocking into the back, become the important topic of automotive field.The concept of automobile active safety is arisen at the historic moment, and the front truck on the road detects a focus that has become automobile active safety research.
At present, front truck detection method commonly used mainly comprises: based on method and the model-based methods two large classes of feature.
Method based on feature is called again Knowledge-Based Method, and it is that moving vehicle detects the most frequently used method, detects vehicle by the affirmation to local or whole some feature of vehicle.The place ahead moving vehicle shows some obvious features in visual pattern, mainly comprise: shape facility, and for example the vehicle tail shape is substantially rectangular, and satisfies certain Aspect Ratio; Boundary characteristic, for example horizontal line of shade, the vertical border of the vehicle tail left and right sides at the bottom of trunk bottom and the car; Symmetry feature and position feature etc.But directly utilize one or more features listed above to remove to detect the place ahead moving vehicle, be easy to be subject to the impact of environmental factor around road, sky, the trees etc., easily cause the flase drop of vehicle and undetected.
Model-based methods is set up two dimension or three-dimensional model for vehicle exactly, by carry out the detection that Model Matching realizes moving vehicle in image.But because vehicle is varied, the attitude of vehicle also can change in the process of moving, thereby makes meter for vehicle reveal various shape, is difficult to set up unified model and comes described.Simultaneously, detection method based on model is excessive to the dependence of auto model, in case Model Matching failure then can't correctly identify moving vehicle, though and set up the precision that meticulous auto model can improve coupling, but certainly will cause being multiplied of calculated amount in the matching process, be difficult to satisfy the requirement of real-time.
Therefore, study the road front truck detection method that a kind of real-time is good, robustness is high, to the support vehicles safety traffic, the safety of the protection person and property has active influence.
Summary of the invention
For overcoming deficiency of the prior art, the present invention aims to provide a kind of front truck detection method based on Haar and HoG feature.The method is to road front vehicles partial occlusion, and illumination variation has very strong robustness.
For realizing above-mentioned technical purpose, reach above-mentioned technique effect, the present invention is achieved through the following technical solutions:
A kind of front truck detection method based on Haar and HoG feature, it may further comprise the steps:
Step 1) manually selects a large amount of vehicle pictures and non-vehicle pictures as the positive and negative samples of training set, and positive and negative samples is normalized under 24 * 24 pixels;
Step 2) uses Haar feature and HoG feature respectively each width of cloth positive and negative samples after the normalization to be characterized, form proper vector;
Step 3) makes up respectively Weak Classifier for two feature vectors of Haar feature and HoG Characteristics creation;
Step 4) utilizes the Adaboost algorithm of cascade that Weak Classifier is trained, and obtains cascade vehicle strong classifier;
The road ahead video image that step 5) obtains for vehicle-mounted camera is judged in the subimage input cascade vehicle strong classifier with wherein various sizes, various positions, and the subimage place that is judged as vehicle is judged as front vehicles.
Compared with prior art, the present invention has following beneficial effect:
The present invention is the road front truck detection method that a kind of real-time is good, robustness is high, and to the support vehicles safety traffic, the safety of the protection person and property has active influence.
Above-mentioned explanation only is the general introduction of technical solution of the present invention, for can clearer understanding technological means of the present invention, and can be implemented according to the content of instructions, below with preferred embodiment of the present invention and cooperate accompanying drawing to be described in detail as follows.The specific embodiment of the present invention is provided in detail by following examples and accompanying drawing thereof.
Description of drawings
Accompanying drawing described herein is used to provide a further understanding of the present invention, consists of the application's a part, and illustrative examples of the present invention and explanation thereof are used for explaining the present invention, do not consist of improper restriction of the present invention.In the accompanying drawings:
Fig. 1 is the structural representation of system that the present invention is based on the front truck detection method of Haar and HoG feature.
Fig. 2 is the front truck detection method process flow diagram that the present invention is based on Haar and HoG feature.
Embodiment
Below with reference to the accompanying drawings and in conjunction with the embodiments, describe the present invention in detail.
Referring to Figure 1 shows that a kind of front truck detection system based on Haar and HoG feature, the video image of the road area C that vehicle-mounted vidicon B is captured is sent to embedded vehicle detector A, and embedded vehicle detector A utilizes the vehicle strong classifier of the cascade that the front truck detection method that the present invention is based on Haar and HoG feature trains to come real-time vehicle is detected.If detect vehicle is arranged on the road ahead, will mark in real time with a square white edge at vehicle-carrying display screen.
Referring to shown in Figure 2, a kind of front truck detection method based on Haar and HoG feature, it may further comprise the steps:
Step 1) manually selects a large amount of vehicle pictures and non-vehicle pictures as the positive and negative samples of training set, and positive and negative samples is normalized under 24 * 24 pixels;
Step 2) uses Haar feature and HoG feature respectively each width of cloth positive and negative samples after the normalization to be characterized, form proper vector;
Step 3) makes up respectively Weak Classifier for two feature vectors of Haar feature and HoG Characteristics creation;
Step 4) utilizes the Adaboost algorithm of cascade that Weak Classifier is trained, and obtains cascade vehicle strong classifier;
The road ahead video image that step 5) obtains for vehicle-mounted camera is judged in the subimage input cascade vehicle strong classifier with wherein various sizes, various positions, and the subimage place that is judged as vehicle is judged as front vehicles.
Concrete comprise following implementation step:
1) vehicle and the positive negative sample of non-vehicle are chosen
Use vehicle-mounted camera to take and store a large amount of road videos.Shooting condition comprises each period and weather condition, such as noon, at dusk, fine, greasy weather and rainy day etc.In these videos, manually intercept out the vehicle that wherein is positioned at the place ahead.The vehicle pictures that intercepts out is the positive sample of square conduct, and wherein vehicle center is positioned at picture center left and right vehicle wheel outward flange and is positioned at the picture left and right sides.Choosing with positive sample of negative sample is similar, and in the road video, manually intercepting out wherein is not the part of vehicle, such as road, sky, buildings, pedestrian and traffic sign etc.The non-vehicle pictures that intercepts out still is that square is as negative sample.Positive and negative samples pictures all is unified and is normalized to 24 * 24 pixels.In the present embodiment, choose altogether 3014 in positive sample, 8319 of negative samples.
2) use the Haar feature that image is characterized
The Haar feature is a kind of digital picture feature of using in pattern-recognition.It is because and the Formal Similarity of Haar small echo and by name Haar feature.The Haar feature is proposed at article " A general framework for object detection " by Papageorgiou the earliest.The pixel that rectangle Haar characterizing definition the most basic is two class subregions in the rectangular area and poor, this two classes subregion is part about rectangle respectively, size is half of rectangle.Certain Haar feature calculation formula is:
In the formula (1),
,
Represent respectively two class subregions,
Certain point in the expression rectangular area,
Coordinate is in the presentation video
The point
Gray-scale value.Viola and Jones expand on the basis to essential characteristic, have proposed by 3 and 4 Haar features that adjacent rectangle forms.On this basis, Lienhart and Maydt have proposed rotatable Haar feature, with original feature expansion to 45 degree direction.
Utilize formula (1) directly all Haar features of piece image to be calculated, calculated amount is large.To this, Viola and Jones have also proposed a quick calculation method.Corresponding any calculative image, they have set up an integral image consistent with this picture size.Any one element value of this integral image equals original image and is positioned at all pixel gray-scale value sums of this element same position upper left.By integrogram, the pixel gray-scale value sum in any rectangular area of original image just can be expressed as in the integrogram algebraic relation between 4 elements.Can represent with following formula:
In the formula (2), sum represents the pixel gray-scale value sum of certain rectangular area,
The value that represents four angle points in this rectangular area first vegetarian refreshments of corresponding same coordinate on integrogram.Like this, a basic Haar feature that is made of two rectangles only need to be used 6 points, and three rectangle Haar features only need to be used 8 points, have greatly saved calculated amount.
The mathematics form of expression of the Haar feature of any form, yardstick and location positioning on sample image is exactly a numerical value, and a width of cloth sample image can extract the Haar feature of various ways, yardstick and position.In the present embodiment, the sample of each 24 * 24 pixel has the Haar proper vector that forms one 87941 dimension.
3) use the HoG feature that image is characterized
Histograms of oriented gradients (HoG:Histogram of oriented gradient) descriptor is a kind of Feature Descriptor that is used for carrying out object detection in computer vision and image processing, is proposed by Navneet Dalal and Bill Triggs.The method has been used the gradient direction feature of image own.The method is similar to the edge orientation histogram method, the SIFT descriptor, but it is characterized in that it is calculating at an intensive big or small unified grid unit of grid, and used the normalized method of overlapping local contrast in order to improve degree of accuracy.
The core concept of histograms of oriented gradients descriptor is that presentation and the shape of the object in the piece image can be described well by the direction Density Distribution at gradient or edge.Its implementation is first image to be divided into the little connected region that is called the grid unit; Then gather gradient direction or the edge orientation histogram of each pixel in the grid unit; At last altogether just can the constitutive characteristic descriptor these set of histograms.In order to improve degree of accuracy, can also be degree of the comparing normalization in the larger block (block) of image of these local histograms, then the method does normalization according to this density value to each grid unit (cell) in the block by calculating first the density of each histogram in this block.After this normalization, can obtain better stability to illumination variation and shade.
Compare with other character description method, the HoG descriptor has many good qualities.At first, because the HoG method is to operate in the local grid unit of image, so it can both keep good unchangeability to image geometry with deformation optics, these two kinds of deformation only can appear on the larger space field.Secondly, under the sampling of thick spatial domain, meticulous direction sampling and stronger conditions such as indicative of local optical normalization, as long as vehicle can keep certain posture substantially, can allow that vehicle has some trickle attitudes to change, these trickle attitudes changes can be left in the basket and not affect the detection effect.
The concrete calculation procedure of HoG feature of using in the present embodiment is as follows:
Step 1: compute gradient value.Adopt the method for common usefulness, the discrete gradient masterplate of an one dimension is applied in respectively the horizontal and vertical direction gets on.Can use following convolution kernel to carry out convolution:
Perhaps
Step 2: set up the unit histogram.Each pixel in each unit is voted to the direction histogram.The shape of each unit can be rectangle or circle, and the direction value of direction histogram can be 0-180 degree or 0-360 degree.Experiment is found, direction is divided into 4 channels result best.The weight of ballot can be the amplitude of gradient or its function.In the reality test, effect is better usually for gradient amplitude itself.
Step 3: the unit is combined into block larger, that spatially link.The HOG descriptor is the vector of the histogrammic element of normalization grid, and histogram is made of the zone of all blocks.These blocks usually can be overlapping, means that each unit has affected last descriptor more than once.Usually use two main blocks: a kind of is the R-HoG block of rectangle, and another kind is circular C-HoG block.The characteristics that present rectangle based on vehicle, the present invention adopts the R-HoG block of rectangle more.
Step 4: with the normalization of gradient intensity part.Order
Represent the not normalized vector that contains all histogram informations of certain block, the normalization operator is:
Wherein,
Be a small value, solve the problem when denominator is zero in the normalization operator.
The statistics with histogram zone that the present invention adopts is 1 * 1,2 * 1,1 * 2 rectangular area for ratio, and scale size is respectively
, port number is 4.Like this, each sample common property of 24 * 24 is given birth to 2873 HoG features, and the dimension of each feature is 4.For example, it is 2 * 1 that Fig. 4 has provided a ratio, and yardstick is 4 HoG feature.
4) make up Weak Classifier
1. Haar differentiates the feature Weak Classifier
In the formula,
It is the Haar feature
Absolute value,
It is threshold value.For any one feature
, the Adaboost algorithm will determine a threshold value with optimum of minimum classification error.
2. HOG generating feature Weak Classifier
Different with the identification and classification device of Haar feature, will be used to the HOG feature based on the generation sorter of positive sample (vehicle).The positive histogrammic median of sample is considered to auto model:
, wherein
It is positive number of samples in the training set.
This generation sorter is by calculating image histogram to be judged
With the auto model histogram
Between distance as basis for estimation, as follows:
In the formula,
Be
With
Pasteur's distance
,
Be
The optimal threshold of feature.
5) use Adaboost Algorithm for Training vehicle classification device
The Adaboost algorithm is proposed by Yoav Freund and Robert Schapire, is a kind of algorithm of machine learning.It can consist of a strong classifier with several Weak Classifiers.The Adaboost algorithm of one two classification can be expressed as follows:
1. for
Individual sample
,
, wherein
The proper vector of expression sample;
2. for
, the initialization weights
With
, wherein
With
It is respectively positive and negative number of samples;
3. iteration
(a) normalized weight
(b) for each feature
, only with sorter of this features training
, its error in classification is used
Expression:
(c) find out a sorter of error in classification minimum
, its error is
(d) upgrade weight:
Wherein work as
When correctly being classified,
Otherwise,
。
4. export strong classifier:
Wherein
6) the cascade algorithm of belt restraining
The cascade algorithm is proved to be a fast and effectively detection algorithm.Final sorter is to be formed by several strong classifier cascades, and each strong classifier is to be obtained by the Adaboost training.Different from the past with maximum characteristic number
As strong classifier iteration stopping foundation, the present embodiment adopts two other parameter: the minimum verification and measurement ratio that allows
Allow false drop rate with maximum
When the strong classifier of every one-level
With
When all reaching the setting value before training, this grade training is namely finished.Next stage
The training negative sample
To from this grade, be produced in the negative sample of mis-classification.At what strong classifier early of cascade classifier, often use less feature just can distinguish more positive negative sample, weed out in other words the negative sample of more " simply ".Along with the increase of progression, strong classifier will increase the number of features of needs step by step.In addition, the rate of growth of the increase number of training characteristics and verification and measurement ratio is not linear in every one-level, but is exponential taper.Therefore, consider verification and measurement ratio and detection time, the present embodiment uses a upper limit function
Retrain every one-level
The number of use characteristic, wherein
Progression.When
The characteristic number that the Adaboost strong classifier of level uses
Arrive
The time, even
With
The value that does not reach prior setting is deconditioning also, and carries out
The training of level.Then the strong classifier that the Adaboost algorithm is formed carries out cascade, namely obtains a good cascade vehicle strong classifier that carries out vehicle and non-vehicle discriminating.
7) front vehicles detects
Vehicle-mounted camera is positioned on the vehicle mirrors, and camera lens is towards Vehicle Driving Cycle the place ahead direction.Use vehicle-mounted camera to obtain road ahead one two field picture, its resolution is 720 * 576.A newly-built search window, the size of window are 24 * 24
Doubly,
The search window of different size is mobile successively to lower right-most portion by the upper left of image, and 24 * 24 pixels are chosen and normalized to every movement once just with the image in the search window.Image after the normalization is characterized with Haar and HoG feature, and the proper vector after will characterizing is input to cascade vehicle classification device.If the output of cascade vehicle classification device is 1, represents that then image is vehicle image in this search window, otherwise then be not.So far, just finished the detection of front truck.
The above is the preferred embodiments of the present invention only, is not limited to the present invention, and for a person skilled in the art, the present invention can have various modifications and variations.Within the spirit and principles in the present invention all, any modification of doing, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (4)
1. the front truck detection method based on Haar and HoG feature is characterized in that, may further comprise the steps:
Step 1) manually selects a large amount of vehicle pictures and non-vehicle pictures as the positive and negative samples of training set, and positive and negative samples is normalized under 24 * 24 pixels;
Step 2) uses Haar feature and HoG feature respectively each width of cloth positive and negative samples after the normalization to be characterized, form proper vector;
Step 3) makes up respectively Weak Classifier for two feature vectors of Haar feature and HoG Characteristics creation;
Step 4) utilizes the Adaboost algorithm of cascade that Weak Classifier is trained, and obtains cascade vehicle strong classifier;
The road ahead video image that step 5) obtains for vehicle-mounted camera is judged in the subimage input cascade vehicle strong classifier with wherein various sizes, various positions, and the subimage place that is judged as vehicle is judged as front vehicles.
2. the front truck detection method based on Haar and HoG feature according to claim 1 is characterized in that, the concrete grammar of step 1 is as follows: use vehicle-mounted camera to take and store a large amount of road videos under each period and weather condition; In these videos, manually intercept out the vehicle that is positioned at the place ahead, the vehicle pictures that intercepts out is the positive sample of square conduct, wherein vehicle center is positioned at picture center left and right vehicle wheel outward flange and is positioned at the picture left and right sides; Manually intercepting out is not the part of vehicle wherein, and it is that square is as negative sample that the non-vehicle pictures that intercepts out is appointed right; The positive and negative samples picture all is unified and is normalized to 24 * 24 pixels.
3. according to claim 1 based on the front truck detection method of Haar and HoG feature, it is characterized in that, the concrete grammar of step 4 is as follows:
Each Weak Classifier obtains a strong classifier by the Adaboost Algorithm for Training, then adopts the minimum verification and measurement ratio that allows
Allow false drop rate with maximum
As strong classifier iteration stopping foundation, when the strong classifier of every one-level
With
When all reaching the setting value before training, this grade training is namely finished; The training negative sample of next stage strong classifier will be produced in the negative sample of mis-classification from this grade; The strong classifier that the Adaboost algorithm is formed carries out cascade, namely obtains a good cascade vehicle strong classifier that carries out vehicle and non-vehicle discriminating.
4. according to claim 1 and 2 or 3 described front truck detection methods based on Haar and HoG feature, it is characterized in that, the concrete grammar of step 5 is as follows: vehicle-mounted camera is positioned on the vehicle mirrors, and camera lens is towards Vehicle Driving Cycle the place ahead direction; Use vehicle-mounted camera to obtain road ahead one two field picture, its resolution is 720 * 576; A newly-built search window, the size of window are 24 * 24
Doubly,
The search window of different size is mobile successively to lower right-most portion by the upper left of image, moves and is spaced apart 2 pixels; 24 * 24 pixels are chosen and normalized to every movement once just with the image in the search window; Image after the normalization is characterized with Haar and HoG feature, and the proper vector after will characterizing is input to cascade vehicle strong classifier; If the output of cascade vehicle strong classifier is 1, represents that then image is vehicle image in this search window, otherwise then be not.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101751678A CN102855500A (en) | 2011-06-27 | 2011-06-27 | Haar and HoG characteristic based preceding car detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101751678A CN102855500A (en) | 2011-06-27 | 2011-06-27 | Haar and HoG characteristic based preceding car detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102855500A true CN102855500A (en) | 2013-01-02 |
Family
ID=47402076
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011101751678A Pending CN102855500A (en) | 2011-06-27 | 2011-06-27 | Haar and HoG characteristic based preceding car detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102855500A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103150903A (en) * | 2013-02-07 | 2013-06-12 | 中国科学院自动化研究所 | Video vehicle detection method for adaptive learning |
| CN103218621A (en) * | 2013-04-21 | 2013-07-24 | 北京航空航天大学 | Identification method of multi-scale vehicles in outdoor video surveillance |
| CN103530640A (en) * | 2013-11-07 | 2014-01-22 | 沈阳聚德视频技术有限公司 | Unlicensed vehicle detection method based on AdaBoost and SVM (support vector machine) |
| CN103559508A (en) * | 2013-11-05 | 2014-02-05 | 福建省视通光电网络有限公司 | Video vehicle detection method based on continuous Adaboost |
| CN103605977A (en) * | 2013-11-05 | 2014-02-26 | 奇瑞汽车股份有限公司 | Extracting method of lane line and device thereof |
| CN103679205A (en) * | 2013-12-23 | 2014-03-26 | 湖南大学 | Preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification |
| CN103871077A (en) * | 2014-03-06 | 2014-06-18 | 中国人民解放军国防科学技术大学 | Extraction method for key frame in road vehicle monitoring video |
| CN104268589A (en) * | 2014-09-16 | 2015-01-07 | 奇瑞汽车股份有限公司 | Front vehicle detection method |
| CN104809429A (en) * | 2015-04-01 | 2015-07-29 | 南京航空航天大学 | Front vehicle detecting method based on compressed sensing |
| CN106193144A (en) * | 2016-07-01 | 2016-12-07 | 蔡雄 | A kind of load wagon possessing Road Detection function |
| CN106372658A (en) * | 2016-08-30 | 2017-02-01 | 广东工业大学 | Vehicle classifier training method |
| CN107766789A (en) * | 2017-08-21 | 2018-03-06 | 浙江零跑科技有限公司 | A kind of vehicle detection localization method based on vehicle-mounted monocular camera |
| CN108182393A (en) * | 2017-12-22 | 2018-06-19 | 上海信耀电子有限公司 | A kind of automobile and its front truck tracking and system of application |
| CN109703460A (en) * | 2019-01-11 | 2019-05-03 | 合肥思艾汽车科技有限公司 | The complex scene adaptive vehicle collision warning device and method for early warning of multi-cam |
| CN109948610A (en) * | 2019-03-14 | 2019-06-28 | 西南交通大学 | A kind of vehicle fine grit classification method in the video based on deep learning |
| CN110682907A (en) * | 2019-10-17 | 2020-01-14 | 四川大学 | Automobile rear-end collision prevention control system and method |
| CN111126406A (en) * | 2019-12-17 | 2020-05-08 | 北京四维图新科技股份有限公司 | Vehicle driving area identification method and device |
| US11390249B2 (en) | 2020-08-05 | 2022-07-19 | Ford Global Technologies, Llc | Vehicle vision system |
| CN117523521A (en) * | 2024-01-04 | 2024-02-06 | 山东科技大学 | Vehicle detection method based on Haar features and improved HOG features |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101196996A (en) * | 2007-12-29 | 2008-06-11 | 北京中星微电子有限公司 | Image detection method and device |
| US20090287620A1 (en) * | 2008-05-16 | 2009-11-19 | Samsung Electronics Co., Ltd. | System and method for object detection and classification with multiple threshold adaptive boosting |
| CN101655914A (en) * | 2008-08-18 | 2010-02-24 | 索尼（中国）有限公司 | Training device, training method and detection method |
| CN101853389A (en) * | 2009-04-01 | 2010-10-06 | 索尼株式会社 | Detection device and method for multi-class targets |
-
2011
- 2011-06-27 CN CN2011101751678A patent/CN102855500A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101196996A (en) * | 2007-12-29 | 2008-06-11 | 北京中星微电子有限公司 | Image detection method and device |
| US20090287620A1 (en) * | 2008-05-16 | 2009-11-19 | Samsung Electronics Co., Ltd. | System and method for object detection and classification with multiple threshold adaptive boosting |
| CN101655914A (en) * | 2008-08-18 | 2010-02-24 | 索尼（中国）有限公司 | Training device, training method and detection method |
| CN101853389A (en) * | 2009-04-01 | 2010-10-06 | 索尼株式会社 | Detection device and method for multi-class targets |
Non-Patent Citations (1)
| Title |
|---|
| 李文波 等: "一种基于Adaboost算法的车辆检测方法", 《长春理工大学学报（自然科学版）》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103150903A (en) * | 2013-02-07 | 2013-06-12 | 中国科学院自动化研究所 | Video vehicle detection method for adaptive learning |
| CN103218621B (en) * | 2013-04-21 | 2015-12-02 | 北京航空航天大学 | The recognition methods of multiple dimensioned vehicle in a kind of life outdoor videos monitoring |
| CN103218621A (en) * | 2013-04-21 | 2013-07-24 | 北京航空航天大学 | Identification method of multi-scale vehicles in outdoor video surveillance |
| CN103559508A (en) * | 2013-11-05 | 2014-02-05 | 福建省视通光电网络有限公司 | Video vehicle detection method based on continuous Adaboost |
| CN103605977A (en) * | 2013-11-05 | 2014-02-26 | 奇瑞汽车股份有限公司 | Extracting method of lane line and device thereof |
| CN103559508B (en) * | 2013-11-05 | 2016-07-27 | 福建中庚视通信息科技有限公司 | A kind of based on continuous Adaboost video vehicle detection method |
| CN103530640A (en) * | 2013-11-07 | 2014-01-22 | 沈阳聚德视频技术有限公司 | Unlicensed vehicle detection method based on AdaBoost and SVM (support vector machine) |
| CN103679205A (en) * | 2013-12-23 | 2014-03-26 | 湖南大学 | Preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification |
| CN103679205B (en) * | 2013-12-23 | 2016-09-28 | 湖南大学 | Assume based on shade and the Foregut fermenters method of layering HOG symmetrical feature checking |
| CN103871077A (en) * | 2014-03-06 | 2014-06-18 | 中国人民解放军国防科学技术大学 | Extraction method for key frame in road vehicle monitoring video |
| CN103871077B (en) * | 2014-03-06 | 2016-06-15 | 中国人民解放军国防科学技术大学 | A kind of extraction method of key frame in road vehicles monitoring video |
| CN104268589A (en) * | 2014-09-16 | 2015-01-07 | 奇瑞汽车股份有限公司 | Front vehicle detection method |
| CN104268589B (en) * | 2014-09-16 | 2017-09-12 | 奇瑞汽车股份有限公司 | A kind of front vehicles detection method |
| CN104809429A (en) * | 2015-04-01 | 2015-07-29 | 南京航空航天大学 | Front vehicle detecting method based on compressed sensing |
| CN106193144A (en) * | 2016-07-01 | 2016-12-07 | 蔡雄 | A kind of load wagon possessing Road Detection function |
| CN106372658A (en) * | 2016-08-30 | 2017-02-01 | 广东工业大学 | Vehicle classifier training method |
| CN107766789B (en) * | 2017-08-21 | 2020-05-29 | 浙江零跑科技有限公司 | Vehicle detection positioning method based on vehicle-mounted monocular camera |
| CN107766789A (en) * | 2017-08-21 | 2018-03-06 | 浙江零跑科技有限公司 | A kind of vehicle detection localization method based on vehicle-mounted monocular camera |
| CN108182393A (en) * | 2017-12-22 | 2018-06-19 | 上海信耀电子有限公司 | A kind of automobile and its front truck tracking and system of application |
| CN109703460A (en) * | 2019-01-11 | 2019-05-03 | 合肥思艾汽车科技有限公司 | The complex scene adaptive vehicle collision warning device and method for early warning of multi-cam |
| CN109948610A (en) * | 2019-03-14 | 2019-06-28 | 西南交通大学 | A kind of vehicle fine grit classification method in the video based on deep learning |
| CN109948610B (en) * | 2019-03-14 | 2022-11-18 | 西南交通大学 | Deep learning-based method for classifying fine granularity of vehicles in video |
| CN110682907A (en) * | 2019-10-17 | 2020-01-14 | 四川大学 | Automobile rear-end collision prevention control system and method |
| CN110682907B (en) * | 2019-10-17 | 2021-06-01 | 四川大学 | Automobile rear-end collision prevention control system and method |
| CN111126406A (en) * | 2019-12-17 | 2020-05-08 | 北京四维图新科技股份有限公司 | Vehicle driving area identification method and device |
| CN111126406B (en) * | 2019-12-17 | 2023-04-07 | 北京四维图新科技股份有限公司 | Vehicle driving area identification method and device |
| US11390249B2 (en) | 2020-08-05 | 2022-07-19 | Ford Global Technologies, Llc | Vehicle vision system |
| CN117523521A (en) * | 2024-01-04 | 2024-02-06 | 山东科技大学 | Vehicle detection method based on Haar features and improved HOG features |
| CN117523521B (en) * | 2024-01-04 | 2024-04-02 | 山东科技大学 | Vehicle detection method based on Haar features and improved HOG features |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102855500A (en) | Haar and HoG characteristic based preceding car detection method | |
| CN103559791B (en) | A kind of vehicle checking method merging radar and ccd video camera signal | |
| WO2019223582A1 (en) | Target detection method and system | |
| CN105160309B (en) | Three lanes detection method based on morphological image segmentation and region growing | |
| Hwang et al. | Multispectral pedestrian detection: Benchmark dataset and baseline | |
| CN109711264B (en) | Method and device for detecting occupation of bus lane | |
| Mao et al. | Preceding vehicle detection using histograms of oriented gradients | |
| CN105335702B (en) | A kind of bayonet model recognizing method based on statistical learning | |
| CN107886080A (en) | One kind is parked position detecting method | |
| US9064157B2 (en) | Pedestrian detection systems and methods | |
| CN106326893A (en) | Vehicle color recognition method based on area discrimination | |
| CN105574552A (en) | Vehicle ranging and collision early warning method based on monocular vision | |
| US9928424B2 (en) | Side window detection through use of spatial probability maps | |
| CN104036284A (en) | Adaboost algorithm based multi-scale pedestrian detection method | |
| CN102902983B (en) | A kind of taxi identification method based on support vector machine | |
| CN103679205A (en) | Preceding car detection method based on shadow hypothesis and layered HOG (histogram of oriented gradient) symmetric characteristic verification | |
| CN106295605A (en) | Traffic lights detection and recognition methods | |
| CN106250824A (en) | Vehicle window localization method and system | |
| CN105184301B (en) | A kind of method that vehicle heading is differentiated using four-axle aircraft | |
| US20160321510A1 (en) | Apparatus and method for detecting bar-type traffic sign in traffic sign recognition system | |
| Elihos et al. | Comparison of image classification and object detection for passenger seat belt violation detection using NIR & RGB surveillance camera images | |
| CN107918775B (en) | Zebra crossing detection method and system for assisting safe driving of vehicle | |
| CN108182393A (en) | A kind of automobile and its front truck tracking and system of application | |
| CN104331708B (en) | A kind of zebra crossing automatic detection analysis method and system | |
| Hechri et al. | Lanes and road signs recognition for driver assistance system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130102 |