CN106469293A - The method and system of quick detection target - Google Patents
The method and system of quick detection target Download PDFInfo
- Publication number
- CN106469293A CN106469293A CN201510516705.3A CN201510516705A CN106469293A CN 106469293 A CN106469293 A CN 106469293A CN 201510516705 A CN201510516705 A CN 201510516705A CN 106469293 A CN106469293 A CN 106469293A
- Authority
- CN
- China
- Prior art keywords
- image
- target object
- picture
- detection
- pixels
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/12—Fingerprints or palmprints
- G06V40/1347—Preprocessing; Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
This application provides in a kind of computer picture interesting target object Fast Detection Technique, be mainly used in improve computing capability limited by computer system in target object detection rates.This application provides a kind of method of quick detection target object, including:Read sequence of pictures to be detected, according to the time of advent of picture and previous testing result, image feature vector is calculated to every width picture;According to described complex characteristic vector, detection target object whether there is in the middle of picture.Compared with prior art, this application provides a kind of based on the interdependence between neighbouring picture, rationally setting scanning area, the method in scanning moment effectively reduce redundancy sweep test in traditional method, reduce averagely every frame process time and maintain higher recall rate and the speed of response.
Description
Technical field
The application is related to video image identification field, more particularly, to quickly detects the method and system of target in video image object.
Background technology
The detection of interesting target object and tracking technique are the important branch of computer vision, are the hot issues that current intelligent image processes with Video processing, have broad application prospects and long-range economic worth.For different types of target object, industry has pointed out multiple detection methods, such as the method based on geometric properties, method based on statistical nature etc..And the method based on statistical nature is just widely used by industry.Do multiple dimensioned dense scanning yet with needing to treat detection image, operand needed for the detection process of target object is quite surprising, so that real-time calculating performance only can just can be reached on high performance desk computer, cannot real time execution on small-sized portable computing device.
The equipment based on single photographic head or other exportable continuous images for the application is as input equipment, and is processed by corresponding computing unit, thus quickly positioning the position of target object interested in captured real scene image, quantity and size.This technology can be used and be used photographic head or video file to detect in the scene of typing as target object any.The application improves, using the technology of innovation, the speed that in scene, target object detects and ensure that higher recall rate level simultaneously, effectively reduces this demand to computing resource for target object detecting system.
Content of the invention
This application provides in a kind of computer picture interesting target object Fast Detection Technique, be mainly used in improve computing capability limited by computer system in target object detection rates.
According to an embodiment of the application, provide a kind of method of quick detection target object, including:Read sequence of pictures to be detected, according to the time of advent of picture and previous testing result, image feature vector is calculated to every width picture;According to described complex characteristic vector, detection target object whether there is in the middle of picture.
Compared with prior art, this application provides a kind of based on the interdependence between neighbouring picture, rationally setting scanning area, the method in scanning moment effectively reduce redundancy sweep test in traditional method, reduce averagely every frame process time and maintain higher recall rate and the speed of response.
Brief description
Accompanying drawing described herein is used for providing further understanding of the present application, constitutes the part of the application, and the schematic description and description of the application is used for explaining the application, does not constitute the improper restriction to the application.
Fig. 1 is the broad flow diagram of the fast target object detecting method of one embodiment of the application.
Fig. 2 is the primary structure block diagram of the fast target object detecting system of one embodiment of the application.
Specific embodiment
The main thought of the application is based on complicated image feature, the fast target object detection of video image interframe relation, it applies complicated image feature such as HARR and LBP feature, compares the grader with the series of parameters data that can be identified after data sample trains grader such as support vector machine etc. to obtain training in a large number by collection(Hereinafter referred to as evaluator), this evaluator done with image and mates with detection object, and the application mates the moment by classifying rationally, matching area, coupling yardstick improve the sweep speed of image and keep higher Detection accuracy.
Purpose, technical scheme and advantage for making the application are clearer, below in conjunction with amplitude and specific embodiment, itself are but described in further detail.
Fig. 1 shows main flow Figure 100 of the fast target object detecting method of an embodiment of the application.
Step S110, inputs the sequence of video images of target object to be detected.
Generally utilize the digital picture of the image capture device collection scene to be detected of system.Target object may be comprised in these digital pictures and be likely to there is no target object, target object is likely to there is different attitudes simultaneously.
In step S110, a frame of digital image of output, judges whether to need to do whole scan, Rule of judgment includes 1 at step S120)System runs to target object detecting step for the first time;2)Apart from the previous whole scan threshold value time of advent;3)Tracking module has triggered whole scan request.When meeting the either condition of this 3 Rule of judgment, system will do a whole scan.So-called whole scan refers to as area-of-interest, entire image be calculated characteristics of image S130 and does characteristic matching S140.
If not meeting the whole scan condition of step S120, system enters step S160, and this step will choose region interested and graphical rule based on following strategy, including 1)The region of priority scan, such as bottom 1/3rd region of image or central area etc. are set according to the region that the obtained highest possibility of reference data occurs.Now still need to the multiple dimension calculation characteristics of image in area-of-interest(Step S130)And do characteristic matching(Step S140);2)The results area of previous detection is done and suitably expands and be set to area-of-interest, now might have multiple semi-cylindrical hills and also need to area-of-interest is done merging treatment and repeat to process to avoid some regions overlap to cause.Characteristics of image now can be calculated on the image of single yardstick(Step S130)And do characteristic matching(Step S140);3)Adjacent dimension calculation characteristics of image to previous testing result(Step S130)And do characteristic matching(Step S140).
Feature extraction shown in step S130 works in standard picture(As 24x24)On, this image is calculated with its local binary patterns feature, this standard picture is divided into multiple blocks, to each pixel in block, be compared with its eight field pixels, certain neighborhood is more than for center pixel, is set to;It is otherwise provided as 0, so can obtain the feature that octet is for this position, whole block is calculated rectangular histogram normalization, the characteristic vector of current detection window after so the rectangular histogram in block being strung, can be obtained.In general, the dimension of characteristic vector is that comparison is high, can be by some conventional dimension reduction methods, such as pivot analysis(PCA)Deng reduction dimension to improve the match cognization speed in step S140.Meanwhile, for ensureing the success rate of characteristic matching S140, reduce loss, feature extraction also needs to carry out on multiple yardsticks of image, image done and repeatedly reduce.
After obtaining characteristic image, characteristic image and known target object feature are done matching step S140, characteristic matching is completed with the distance of target object feature by calculating characteristic image.The distance between characteristic vector generally adopts Euclidean distance or Chebyshev's distance (Chebyshev Distance) definition to be calculated.
Fig. 2 shows main flow Figure 200 of the fast target object detecting system of an embodiment of the application.
Sample collection step S200, gathers substantial amounts of view data sample in advance and these view data samples is split and labelling.Thus, it is possible to obtain the substantial amounts of single information on target object image (sample image) classified.
Before information on the image to be detected to input is scanned detection, need to carry out pretreatment, to obtain the sample database that target object to be detected can be detected and evaluator (grader with supplemental characteristic after training).
A kind of specifically mode, can be gather in advance substantial amounts of then these view data samples are split containing target object image data sample to be detected, the single target object that obtains (sample image will be called in the following text after segmentation)It is marked.By this segmentation, by the target object in view data sample separately single target object image (sample image) can be formed as positive sample.
This segmentation, labelling(I.e. view data sample analyses)Can be by modes such as artificial or machine algorithms it is intended that sample image (target object image to be detected after the segmentation of collection).
Here, acquisition mode, can provide target object image to be detected to carry out view data sample collection by picture catching/capturing apparatus.
For example:By the image of the target object to be detected of image acquisition device sample, these images include image under the conditions of different attitudes difference illumination etc. for the target object to be detected.Next these view data samples can be analyzed with (segmentation and labelling), such as manual type analysis, can be specifically to use image editing tools, as Photoshop, GIMP etc., target object region to be detected can be found on these view data samples collecting, separately save as an image, that is, segmentation figure as data sample be single target object image (sample image to be identified), and it is labeled as positive sample image to be identified(S220).
Need in step S225 to prepare the negative sample suitable with positive sample number, the selection of negative sample can be obtained by intercepting parts of images in the image of scene without target object to be detected for the image acquisition device it is also possible to pass through to intercept the image acquisition in the video file similar with target object scene to be detected.
Feature extraction shown in step S230 uses step S130 methods described.
Training grader step S240, using the feature of the sample image extracting in characteristic extraction step S230, as the feature samples of the detection process of image information, and with these sample training graders, to obtain corresponding grader(Step S280), i.e. the To Template of matching detection.
Grader, such as using support vector machines or AdaBoost etc..
In target object detection process, in similar Fig. 1, step S250 input picture, and step S260 then selects scanning window, inputted to grader S280 based on the characteristics of image that scanning window and step S230 are extracted, output category result step S290 after matching primitives.
The application employ according to historical data make rational planning for scanning area, scanning yardstick method control targe object detection in matching operation amount, improve detection speed and the response speed of target object, and, the testing result degree of reliability be guaranteed.
In a typical configuration, computing device includes one or more processors(CPU), input/output interface, network interface and internal memory.Internal memory potentially includes the volatile memory in computer-readable medium, random access memory (RAM)And/or the form such as Nonvolatile memory, such as read only memory(ROM)Or flash memory(Flash RAM).Internal memory is the example of computer-readable medium.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media and can realize information Store by any method or technique.Information can be computer-readable instruction, data structure, the module of program or other data.The example of the storage medium of computer includes, but are not limited to phase transition internal memory(PRAM), static RAM(SRAM), dynamic random access memory(DRAM), other kinds of random access memory(RAM), read only memory(ROM), Electrically Erasable Read Only Memory(EEPR0M), fast flash memory bank or other memory techniques, read-only optical disc read only memory (CD-ROM), digital versatile disc(DVD) or other optical storage, magnetic cassette tape, the storage of tape magnetic rigid disk or other magnetic storage apparatus or any other non-transmission medium, can be used for storing the information that can be accessed by a computing device.Define according to herein, computer-readable medium does not include temporary computer readable media(Transitory media), the such as data signal of modulation and carrier wave.
Each embodiment in this specification is typically described by the way of going forward one by one, and what each embodiment stressed is the difference with other embodiment, between each embodiment identical similar partly mutually referring to.
The application can be described in the general context of computer executable instructions, such as program module or unit.Usually, program module or unit can include execution particular task or the routine realizing particular abstract data type, program, object, assembly, data structure etc..In general, program module or unit can be by software, hardware or both be implemented in combination in.The application can also be put into practice in a distributed computing environment, in these distributed computing environment, execute task by the remote processing devices connected by communication network.In a distributed computing environment, program module or unit may be located in the local and remote computer-readable storage medium including storage device.
Finally, it can further be stated that, term " inclusion ", " comprising " or its any other variant are intended to comprising of nonexcludability, so that including a series of process of key elements, method, commodity or equipment not only include those key elements, but also include other key elements of being not expressly set out, or also include for this process, method, commodity or the intrinsic key element of equipment.In the absence of more restrictions, it is not excluded that also there is other identical element in process, method, commodity or the equipment including described key element in the key element being limited by sentence " including ... ".
Those skilled in the art are it should be appreciated that embodiments herein can be provided as method, system or computer program.Therefore, the application can be in the form of complete hardware embodiment, complete software embodiment or the embodiment combining software and hardware aspect.And, the application can be using in one or more computer-usable storage medium (including but not limited to disk memory, CD-ROM, optical memory etc. wherein including computer usable program code)The form of the computer program of upper enforcement.
Specific case used herein is set forth to the principle of the application and embodiment, and the explanation of above example is only intended to help and understands the present processes and its main thought;Simultaneously for one of ordinary skill in the art, according to the thought of the application, all will change in specific embodiments and applications, in sum, this specification content should not be construed as the restriction to the application.
Claims (14)
1. a kind of fast target object identification method is it is characterised in that include:Target object interested in detection, the image of positioning belt identification, and preserve this result;Select, reduce scanning area according to reference data, previous testing result;Detection positioning, using the method for characteristic matching, is searched in the picture, and feature typically selects local binary feature.
2. method according to claim 1, it is characterised in that including detection, the target object interested in the image of positioning belt identification, and preserves this result;Select, reduce scanning area according to reference data, previous testing result, reduce the yardstick level of image down;Detection positioning, using the method for characteristic matching, is searched in the picture, and feature typically selects local binary feature.
3. method according to claim 1, according to reference data, previous testing result select, reduce scanning area it is characterised in that, including planning multiple scanning windows in advance, when processing every two field picture, according to certain rule, as sequential order, the window of scanogram.
4. method according to claim 2, select, reduce scanning area according to reference data, previous testing result, reduce the yardstick level of image down, it is characterized in that, including the result according to front one-time detection, regional choice candidate region in its vicinity, selects corresponding graphical rule level simultaneously, i.e. the dimension scale of this image and original input picture.
5. method according to claim 1 includes it is characterised in that the target object image obtaining being carried out with complex characteristic vector and extracting:Based on an acquired target object image of positioning, size criteria is unified to described target object image;
A default block of pixels, the image node-by-node algorithm characteristics of image to the described block of pixels of each in described target object image, to obtain the complex characteristic vector of described character information image.
6. method according to claim 5 is it is characterised in that described node-by-node algorithm characteristics of image includes:The block of pixels that described block of pixels is adjacent forms block, and described block is chosen with the characteristic vector description of 476 dimensions;In described target object image respectively transversely, vertically move, combination obtain the higher-dimension to described target object image characteristic vector description.
7. method according to claim 1 is it is characterised in that also include:Pre-treatment step, wherein, collection has the image tagged of target object to be detected in a large number, to obtain labelling target image;To described labelling target image, it is normalized, to obtain image;A default block of pixels, the image node-by-node algorithm characteristics of image to the described block of pixels of each in described image,
To obtain the local binary feature of described target object;Using described feature samples, carry out classifier training, to obtain the described evaluator of identifying processing.
8. method according to claim 1 is it is characterised in that also include:Input step, inputs altimetric image to be checked using picture catching mode;
Verification step, is verified to the target object detecting to confirm final detection result.
9. a kind of fast target object identification system is it is characterised in that include:Scanning area planning unit, selects the scan position in every two field picture according to reference data, testing result or predefined rule;Feature extraction unit, calculates characteristics of image to image to be detected;Target object detection, positioning unit, according to target object feature and images match to be detected obtain target object whether there is and size information.
10. according to the system described in claim 9 it is characterised in that including:Scanning area planning unit, selects the scan position in every two field picture, and the yardstick level selecting image scanning according to the result of previous scan detection according to reference data, testing result or predefined rule;Feature extraction unit, calculates characteristics of image to image to be detected;Target object detection, positioning unit, according to target object feature and images match to be detected obtain target object whether there is and size information.
11. systems according to claim 9, according to reference data, previous testing result select, reduce scanning area it is characterised in that, including planning multiple scanning windows in advance, when processing every two field picture, according to certain rule, as sequential order, the window of scanogram;The scan matching that multiple yardstick levels after also needing in scanning window this window is reduced are carried out.
12. systems according to claim 9, select, reduce scanning area according to reference data, previous testing result, reduce the yardstick level of image down, it is characterized in that, including the result according to front one-time detection, regional choice candidate region in its vicinity, selects corresponding graphical rule level simultaneously, i.e. the dimension scale of this image and original input picture.
13. systems according to claim 9 include it is characterised in that the target object image obtaining being carried out with complex characteristic vector and extracting:Based on an acquired target object image of positioning, size criteria is unified to described target object image;
A default block of pixels, the image node-by-node algorithm characteristics of image to the described block of pixels of each in described target object image, to obtain the complex characteristic vector of described character information image.
14. systems according to claim 9 are it is characterised in that described node-by-node algorithm characteristics of image includes:The block of pixels that described block of pixels is adjacent forms block, and described block is chosen with the characteristic vector description of 476 dimensions;In described target object image respectively transversely, vertically move, combination obtain the higher-dimension to described target object image characteristic vector description;Image is done multiple reducing and by preceding method calculating high dimensional feature vector description.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510516705.3A CN106469293A (en) | 2015-08-21 | 2015-08-21 | The method and system of quick detection target |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510516705.3A CN106469293A (en) | 2015-08-21 | 2015-08-21 | The method and system of quick detection target |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106469293A true CN106469293A (en) | 2017-03-01 |
Family
ID=58229765
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510516705.3A Pending CN106469293A (en) | 2015-08-21 | 2015-08-21 | The method and system of quick detection target |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106469293A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271870A (en) * | 2018-08-21 | 2019-01-25 | 平安科技(深圳)有限公司 | Pedestrian recognition methods, device, computer equipment and storage medium again |
CN109829486A (en) * | 2019-01-11 | 2019-05-31 | 新华三技术有限公司 | Image processing method and device |
CN110874547A (en) * | 2018-08-30 | 2020-03-10 | 富士通株式会社 | Method and device for identifying object from video |
US11010613B2 (en) | 2018-11-29 | 2021-05-18 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for target identification in video |
CN112836759A (en) * | 2021-02-09 | 2021-05-25 | 重庆紫光华山智安科技有限公司 | Method and device for evaluating machine-selected picture, storage medium and electronic equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102592147A (en) * | 2011-12-30 | 2012-07-18 | 深圳市万兴软件有限公司 | Method and device for detecting human face |
CN102831439A (en) * | 2012-08-15 | 2012-12-19 | 深圳先进技术研究院 | Gesture tracking method and gesture tracking system |
CN104217203A (en) * | 2013-06-03 | 2014-12-17 | 支付宝(中国)网络技术有限公司 | Complex background card face information identification method and system |
CN104268584A (en) * | 2014-09-16 | 2015-01-07 | 南京邮电大学 | Human face detection method based on hierarchical filtration |
-
2015
- 2015-08-21 CN CN201510516705.3A patent/CN106469293A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102592147A (en) * | 2011-12-30 | 2012-07-18 | 深圳市万兴软件有限公司 | Method and device for detecting human face |
CN102831439A (en) * | 2012-08-15 | 2012-12-19 | 深圳先进技术研究院 | Gesture tracking method and gesture tracking system |
CN104217203A (en) * | 2013-06-03 | 2014-12-17 | 支付宝(中国)网络技术有限公司 | Complex background card face information identification method and system |
CN104268584A (en) * | 2014-09-16 | 2015-01-07 | 南京邮电大学 | Human face detection method based on hierarchical filtration |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109271870A (en) * | 2018-08-21 | 2019-01-25 | 平安科技(深圳)有限公司 | Pedestrian recognition methods, device, computer equipment and storage medium again |
CN109271870B (en) * | 2018-08-21 | 2023-12-26 | 平安科技(深圳)有限公司 | Pedestrian re-identification method, device, computer equipment and storage medium |
CN110874547A (en) * | 2018-08-30 | 2020-03-10 | 富士通株式会社 | Method and device for identifying object from video |
CN110874547B (en) * | 2018-08-30 | 2023-09-12 | 富士通株式会社 | Method and apparatus for identifying objects from video |
US11010613B2 (en) | 2018-11-29 | 2021-05-18 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for target identification in video |
CN109829486A (en) * | 2019-01-11 | 2019-05-31 | 新华三技术有限公司 | Image processing method and device |
CN112836759A (en) * | 2021-02-09 | 2021-05-25 | 重庆紫光华山智安科技有限公司 | Method and device for evaluating machine-selected picture, storage medium and electronic equipment |
CN112836759B (en) * | 2021-02-09 | 2023-05-30 | 重庆紫光华山智安科技有限公司 | Machine-selected picture evaluation method and device, storage medium and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110569721B (en) | Recognition model training method, image recognition method, device, equipment and medium | |
US8509526B2 (en) | Detection of objects in digital images | |
Ping et al. | A deep learning approach for street pothole detection | |
CN107330027B (en) | Weak supervision depth station caption detection method | |
CN106469293A (en) | The method and system of quick detection target | |
CN110460838B (en) | Lens switching detection method and device and computer equipment | |
CN108765386A (en) | A kind of tunnel slot detection method, device, electronic equipment and storage medium | |
CN105868708A (en) | Image object identifying method and apparatus | |
CN111583180B (en) | Image tampering identification method and device, computer equipment and storage medium | |
CN111723815B (en) | Model training method, image processing device, computer system and medium | |
CN110827292B (en) | Video instance segmentation method and device based on convolutional neural network | |
CN104301630A (en) | Video image splicing method and device | |
He et al. | Aggregating local context for accurate scene text detection | |
CN113706481A (en) | Sperm quality detection method, sperm quality detection device, computer equipment and storage medium | |
CN113780145A (en) | Sperm morphology detection method, sperm morphology detection device, computer equipment and storage medium | |
CN112052702A (en) | Method and device for identifying two-dimensional code | |
KR20170126300A (en) | Method, apparatus and computer program for matching between the images | |
Luo et al. | Saliency density maximization for object detection and localization | |
Ke et al. | Vehicle logo recognition with small sample problem in complex scene based on data augmentation | |
CN114078106A (en) | Defect detection method based on improved Faster R-CNN | |
Chang et al. | Semi-supervised learning for YOLOv4 object detection in license plate recognition system | |
Tang et al. | Salient Feature Pyramid Network for Ship Detection in SAR Images | |
CN113297514A (en) | Image processing method, image processing device, electronic equipment and computer storage medium | |
Li et al. | Detection of Surface Defects of Magnetic Tiles Based on Improved YOLOv5 | |
CN107886102B (en) | Adaboost classifier training method and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170301 |