CN102956032A - Target template updating method - Google Patents
Target template updating method Download PDFInfo
- Publication number
- CN102956032A CN102956032A CN2011102412854A CN201110241285A CN102956032A CN 102956032 A CN102956032 A CN 102956032A CN 2011102412854 A CN2011102412854 A CN 2011102412854A CN 201110241285 A CN201110241285 A CN 201110241285A CN 102956032 A CN102956032 A CN 102956032A
- Authority
- CN
- China
- Prior art keywords
- template
- histogram
- target
- passage
- illumination
- 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 target template updating method. The method includes the steps: step one, comparing a current frame and a last frame to determine whether a target position has illumination upheaval or not; step two, if illumination upheaval does not occur, updating a target template and characteristic passage templates thereof according to preset updating speed parameters of the target template, and if illumination upheaval occurs, recalculating the updating speed parameters of the target template and executing the step three; step three, leading the updating speed parameters at the step two into a selective submodel updating strategy and updating the target template and characteristic passage templates thereof; and step four, executing step one to three sequentially aiming at each frame. The target template updating method has the advantages that the template updating speed can be adjusted timely under the condition of occurrence of illumination upheaval and on the basis of the selective submodel updating method to adapt to the requirement on quick template updating under violet illumination. Compared with the traditional selective submodel updating method, the target template updating method can better adapt to variation of external scenes to complete more continuous tracking effects.
Description
Technical field
The present invention relates to video monitoring and follow the tracks of and technical field of image processing, relate in particular to a kind of update method of To Template.
Background technology
Increasingly mature along with the development of intelligent monitoring technology and image processing techniques, original employing manpower carries out the suspicious object monitoring and has not satisfied demand, and can remedy to a great extent the problem of shortage of manpower as main intelligent safety and defence system take technology such as artificial intelligence and video analysis.The intelligent safety and defence system advantage is that the security personnel need not to go on patrol on the spot, just can obtain on-the-spot actual conditions by the video that supervisory system is passed back, so that make a policy rapidly, therefore wide development space and huge potential market is arranged.They mainly are the methods of utilizing image to process, automatically detect the target invasion in reality scene, make computing machine have the ability of certain understanding and analysis video, thereby the behaviors such as target invasion are reported to the police.
The tracking of target is an important research direction in the digital video technology, its application is extensive completely, from the industrial detection to the security monitoring, from the medical image to the Military Application etc., many fields have and relate to, the for example automatic Pilot of the security personnel of the detection of the magnitude of traffic flow, important place, aviation and automobile or auxiliary driving, the weapon guidance in the military affairs and the aspect of control.Because how the uncertainty that the diversity of tracking target form and target signature change realizes that efficient target tracking is the focus of research under the various environment always.
For general video sequence target tracking algorism, because adopting single renewal speed parameter to control template characteristic upgrades, if the initial setting up of renewal speed parameter is less than normal, so in the situation of target environmental change of living in relatively violent (as: illumination strengthens suddenly or weakens), because target signature changes than very fast, the To Template feature occurs and upgrade the speed that target signature changes that do not catch up with, thereby cause the phenomenon of track rejection, if the initial setting up of renewal speed parameter is bigger than normal, although can adapt to the situation of illumination acute variation, but when target is blocked, easily with non-object information updating to template, thereby cause with wrong phenomenon, greatly limited the validity of following the tracks of.
Summary of the invention
The To Template drastic change that causes for the illumination acute variation causes the problem of track rejection, the invention provides a kind of new template renewal method, can adapt in real time, fast the illumination drastic change of following the tracks of in the scene, simultaneously can wrong more new template when eclipse phenomena occurs yet, for providing, the target following that continues provides powerful support for.
A kind of update method of To Template may further comprise the steps:
Step 1: in contrast present frame and the former frame, whether the target position illumination drastic change occurs;
Step 2: when illumination drastic change does not occur, upgrade To Template and each feature passage template thereof by the renewal speed parameter of default To Template;
When illumination drastic change occurs, recomputate the renewal speed parameter of To Template, execution in step 3;
Step 3: the renewal speed parameter of step 2 is incorporated into selectivity submodel update strategy, upgrades To Template and each feature passage template thereof;
Step 4: in order for each frame execution in step 1～3.
Comprise in the described step 1:
Step 101, in present frame, determine position and the scope at target (target image) place, can utilize existing algorithm, for example utilize mean shift iterative algorithm according to position and the scope at former frame target place, calculate position and the scope at present frame target place.
Step 102, the histogram of the feature passage of extraction target under this position, what described feature passage was chosen is R, G, B, brightness ((R+G+B)/3), five kinds of passages of tone (H).
Step 103 is compared the histogram of target signature passage with the histogram template of target signature passage, determine whether to occur illumination drastic change;
In the track algorithm of video object, generally all to set up first To Template and each feature passage template thereof, can obtain template histogram and the To Template histogram of target signature passage according to each self-corresponding histogram.The histogram template or the To Template histogram that also claim the target signature passage.
The histogram template of target signature passage and target histogram template are constantly updated, and when comparing, the histogram template of described target signature passage and target histogram template all refer to obtain after last the renewal.Be attached to the present invention, the histogram template of target signature passage comprises the histogram template of R, G, B, brightness, five kinds of passages of tone.
In the described step 103, when the histogram of target signature passage is compared with the histogram template of target signature passage, if satisfy condition simultaneously A and condition B, then think illumination drastic change occurs:
The histogram of the R of condition A, target, G, B, 4 passages of brightness is compared former frame and is occured obviously to reduce with the matching value of corresponding histogram template, and the histogram of tone passage is compared with the matching value of its histogram template and is changed not obviously, and the histogram center of R, G, B, 4 passages of brightness is consistent with respect to the offset direction of the histogram template center of correspondence.
Can utilize existing account form when calculating described matching value, then be normalized to 0～1 scope, can think the generation significant change more than or equal to 0.3 the time when matching value changes generally speaking, otherwise can think to change not obvious.
For example, in former frame (X frame), the matching value of the histogram template of the histogram of R passage and R passage (the histogram template of the R passage after the upper once renewal before this X frame) is 0.7;
In the present frame, the matching value of the histogram template of the histogram of R passage and R passage (the histogram template of the R passage after the upper once renewal before this X frame) is 0.3;
0.7-0.3=0.4, can think that then the histogram of R passage is compared obviously minimizing of former frame generation with the matching value of the histogram template of R passage.
The histogrammic similarity degree that the histogram that the histogrammic horizontal projection of condition B, each feature passage of present frame generates and the histogram template horizontal projection of each feature passage generate is greater than the threshold value of setting, and the histogram center of gravity of each feature passage is with after the center of gravity of corresponding histogram template is alignd, with the similarity degree of corresponding passage histogram template greater than the threshold value of setting.
Judging that similarity degree is to utilize Pasteur's coefficient to weigh, be judged to be similar in the time of between Pasteur's coefficient is in 0.9～1.The threshold value that is similarity degree is 0.9.
In the step 2, when illumination drastic change occurs, recomputate the renewal speed parameter a of To Template,
a＝(min(abs(σ))，0.3)+min((t-1)
3，9)*0.2*min(abs(σ)/0.5，1))/3 (1)
Wherein σ is the difference of the center of gravity of target histogram and target histogram template;
T is the lasting frame number of illuminance abrupt variation.
Formula can find out in (1) that the gain size of template renewal speed is common decision of frame number that is continued by the template drift yardstick under each feature passage of target and drift.The drift yardstick of To Template is the difference reflection of the center of gravity of current goal histogram and target histogram template, and the difference of center of gravity is larger, and then the proof drift is larger, otherwise the drift yardstick is less.
After illumination drastic change occuring and recomputating the renewal speed parameter, can utilize prior art, this renewal speed parameter is incorporated into selectivity submodel update strategy, upgrade To Template and each feature passage template thereof.
Beneficial effect of the present invention is, on the basis of selectivity submodel update method, can be in the situation that illumination drastic change occurs, adjust timely template renewal speed, adapt to upgrading demand fast of the violent situation lower bolster of illumination, compare traditional selectivity submodel update method, can better adapt to the variation of extraneous scene, finish more lasting tracking effect.
Description of drawings:
Fig. 1 is the track algorithm general frame process flow diagram of using update method of the present invention;
Fig. 3 is the process flow diagram of update method of the present invention;
Fig. 2 is selectivity submodel update strategy process flow diagram;
Fig. 4 is each passage property of the histogram synoptic diagram of the present invention;
Fig. 4 a is the histogram template under the R passage;
Fig. 4 b is the horizontal projection synoptic diagram of the histogram template among Fig. 4 a;
Fig. 4 c is the histogram that the horizontal projection among Fig. 4 b generates;
Fig. 5 a is the histogram of the R passage under the light intensity decreasing;
Fig. 5 b is the histogrammic horizontal projection synoptic diagram among Fig. 5 a;
Fig. 5 c is the histogram that the horizontal projection among Fig. 5 b generates;
Fig. 6 a is the histogram of the R passage under light intensity strengthens;
Fig. 6 b is the histogrammic horizontal projection synoptic diagram among Fig. 6 a;
Fig. 6 c is the histogram that the horizontal projection among Fig. 6 b generates.
Embodiment:
Fig. 1 shows track algorithm general frame process flow diagram of the present invention.Step is as follows:
Start the track algorithm module, the colored template of initialization target, To Template selects the colored template dimension of RGB passage to get 8 * 8 * 8, and extracting target R, G, B, brightness, five kinds of feature passages of tone and setting up dimension is 32 histogram template.
In the enterprising line search iteration of the RGB of target template, the final position of iteration is as the target location of present frame with mean shift algorithm.Specific algorithm is described can be with reference to pertinent literature, as: " based on the visual target tracking method summary of average drifting "-computer engineering.
Readjust the target window size, present embodiment is selected simple window size method of adjustment, i.e. positive negative float 10%, with the window of Pasteur's coefficient maximum as new window.
Upgrade To Template, present embodiment is to improve on the basis of traditional submodel update strategy, add renewal speed and adjust the factor, the present invention utilizes the histogram variation relation of several Color Channels, distinguishing accurately that To Template changes is owing to block or illuminance abrupt variation causes, and then the gain of calculating the adjusting template renewal speed is big or small, to reach the purpose of real-time update template.
Show that such as Fig. 2 concrete steps are as follows;
Above-mentioned window with Pasteur's coefficient maximum has namely obtained the current location of target as new window in the present frame, calculate the histogram under R, G, B, brightness, five feature passages of tone under the target current location.
Judging when front template drastic change whether to be caused by illumination drastic change, mainly is according to R, G, B, brightness, and after five kinds of passage histograms of tone were processed through vertical projections and horizontal projection, the characteristic that they have in theory during according to illumination drastic change determined.
When the histogram of target signature passage is compared with the histogram template of target signature passage, if satisfy condition simultaneously A and condition B, then think illumination drastic change occurs:
If comparing former frame with the matching value of corresponding histogram template, the histogram of the R of condition A target, G, B, 4 passages of brightness occurs obviously to reduce, and the histogram of tone passage is compared with the matching value of its histogram template and is changed not obviously, and the histogram center of R, G, B, 4 passages of brightness is consistent with respect to the offset direction of the histogram template center of correspondence.
The histogrammic similarity degree that the histogram that the histogrammic horizontal projection of condition B, each feature passage of present frame generates and the histogram template horizontal projection of each feature passage generate is greater than the threshold value of setting, and the histogram center of gravity of each feature passage is with after the center of gravity of corresponding histogram template is alignd, with the similarity degree of corresponding passage histogram template greater than the threshold value of setting.
Shown in Fig. 4 a～Fig. 6 c, be respectively under the normal state by row, under the illumination reduction state, the histogram that feature passage histogram under the illumination enhanced situation, horizontal projection and projection produce, can find out, when illumination drastic change occured, the histogram that the in the horizontal direction projection of histogram template after the variation generates possessed higher similarity, had equally higher similarity through level correction on the vertical direction.When target is blocked, will can not reach higher similarity.Calculate similarity Pasteur's coefficient calculations.The similar threshold value of setting is 0.9.
Calculate the renewal speed factor, main principle is exactly jointly to be determined by the frame number that the drift yardstick of the template weight under each passage of target and drift continue.If σ is the difference when front template and target histogram template center of gravity, t is the lasting frame number of illuminance abrupt variation, and velocity factor is calculated according to the following formula, and this formula is released by a large amount of test datas.
a＝(min(abs(σ))，0.3)+min((t-1)
3，9)*0.2*min(abs(σ)/0.5，1))/3 (1)
Velocity factor a with newly obtaining is brought into formula (3), and innovation of the present invention just is therefore to change speed into a dynamic parameter, can adapt to the environment of illumination drastic change.Then upgrade To Template according to idiographic flow shown in Figure 3.
Traditional selectivity submodel update strategy is: (Pasteur's coefficient is the computing method of judging two template similarity degrees at first to adopt Pasteur's coefficient ρ to weigh the validity of tracking results, scope is 0～1, value is to represent that similarity degree was the highest at 1 o'clock), if Pasteur's coefficient value is less than certain threshold value, illustrate that this tracking results has been subject to violent interference, this frame should not carry out template renewal, still in the basic enterprising line trace of original object template; If Pasteur's coefficient value is greater than certain threshold value, then (submodel refers to the histogrammic one-component of target to each submodel, for example: if the target histogram is 32 dimensions, the number of submodel is exactly 32 so) according to contribution degree (contribution degree refers to: each histogram component is to generating the influence degree of Pasteur's coefficient) ordering, before the higher submodel of several contribution degrees represent that "current" model and object module are more identical, for avoiding excessively upgrading the model skew that causes, the submodel higher to contribution degree do not upgrade, and only less to contribution degree is to upgrade with the identical relatively poor submodel of object module in the "current" model.
As shown in Figure 3, concrete steps are as follows:
Target histogram under the statistics current position coordinates is designated as p, obtains Pasteur's coefficient ρ with To Template histogram q coupling, if ρ＞0.9, the expression tracking results is not subject to violent interference, more new template, otherwise, can be understood as that eclipse phenomena occurs or other interference, not new template more.
Begin to carry out the selectivity submodel and upgrade, specific as follows:
p
uThe histogrammic submodel of expression present frame target, q
uExpression To Template histogram submodel, all histogrammic submodel q of traversal To Template
u, respectively to q
uUpgrade, carry out again at last normalization, obtain new To Template histogram.The submodel update rule is as follows:
1) q
u=p
u=0, this submodel does not all appear in expression present frame target histogram and the To Template histogram, will not upgrade it, more new formula is formula (2);
2) q
u＞0, p
u=0, there is this submodel in the expression To Template histogram and in present frame target histogram, do not have this submodel, the situation that this changes corresponding to target appearance usually, should upgrade submodel this moment, but too responsive to changing for fear of To Template, should adopt certain proportion to be weighted this moment, and more new formula is formula (3);
3) q
u=0, p
u＞0, do not have this submodel in the expression To Template histogram and have this submodel in present frame target histogram, the situation that this occurs corresponding to non-target occlusion usually also should be upgraded submodel this moment, and adopt equally certain proportion to be weighted, more new formula is formula (3);
4) q
u＞0, p
u＞0, and the contribution margin β of submodel
uLess than setting threshold k, β is set in the present embodiment
u=0.5 is the threshold value of contribution degree size, and should mate contribution according to it and be weighted renewal this moment, and the contribution degree size is calculated according to formula (6), and more new formula is formula (5);
5) q
u＞0, p
u＞0, and the contribution margin β of submodel
uGreater than setting threshold k, represent that this submodel is comparatively stable, should not upgrade submodel this moment, and more new formula is formula (4);
Specifically more new formula is as follows:
A is the model modification velocity factor in the formula, common value 0.1, C
qBe normalization coefficient.
β wherein
uBe the matching degree contribution margin, by formula calculate (6)
A in the formula (3) is the renewal speed factor, and it is a static parameter, and scope is 0 to 1, therefore can not adapt to the more new demand of illumination drastic change situation lower bolster.
Among the present invention, namely the improvement of conventional template update method is to change a in formula (2)～(6) variation (formula (1) shown in) of a dynamic parameter to adapt to the ambient light photograph into.
Because β
uWhen illumination drastic change, thus also should increase accordingly weight, therefore with the β in formula (2)～(6)
uChange β into
u`, β
u`=min (1, β
u+ a), a here is also shown in formula (1).
Tradition selectivity submodel update method is that velocity factor a is set to constant, and the present invention utilizes related channel program property of the histogram relation under the illumination drastic change condition, on this basis with a and β
uChange the variable that affected by σ and t into, can adapt to the scene demand, when illumination variation is violent, their value increases accordingly, accelerates the renewal speed of template, when illumination variation is mild, their value reduces, reduce template renewal speed, reach adaptive template renewal effect, the more lasting tracking effect of final realization.
Claims (7)
1. the update method of a To Template is characterized in that, may further comprise the steps:
Step 1: in contrast present frame and the former frame, whether the target position illumination drastic change occurs;
Step 2: when illumination drastic change does not occur, upgrade To Template and each feature passage template thereof by the renewal speed parameter of default To Template; When illumination drastic change occurs, recomputate the renewal speed parameter of To Template, execution in step 3;
Step 3: the renewal speed parameter of step 2 is incorporated into selectivity submodel update strategy, upgrades To Template and each feature passage template thereof;
Step 4: in order for each frame execution in step 1～3.
2. the update method of To Template as claimed in claim 1 is characterized in that, comprises in the described step 1:
Step 101, position and the scope at definite target place in present frame;
Step 102, the histogram of the feature passage of extraction target under this position;
Step 103 is compared the histogram of target signature passage with the histogram template of target signature passage, determine whether to occur illumination drastic change.
3. the update method of To Template as claimed in claim 2 is characterized in that, described feature passage is R, G, B, brightness, five kinds of passages of tone.
4. the update method of To Template as claimed in claim 3, it is characterized in that, in the described step 103, when the histogram of target signature passage is compared with the histogram template of target signature passage, A and condition B if satisfy condition simultaneously, then think illumination drastic change occurs:
The histogram of the R of condition A, target, G, B, 4 passages of brightness is compared former frame and is occured obviously to reduce with the matching value of corresponding histogram template, and the histogram of tone passage is compared with the matching value of its histogram template and is changed not obviously, and the histogram center of R, G, B, 4 passages of brightness is consistent with respect to the offset direction of the histogram template center of correspondence;
The histogrammic similarity degree that the histogram that the histogrammic horizontal projection of condition B, each feature passage of present frame generates and the histogram template horizontal projection of each feature passage generate is greater than the threshold value of setting, and the histogram center of gravity of each feature passage is with after the center of gravity of corresponding histogram template is alignd, with the similarity degree of corresponding passage histogram template greater than the threshold value of setting.
5. the update method of To Template as claimed in claim 4 is characterized in that, described matching value is 0～1 scope, when matching value changes more than or equal to 0.3 the time, thinks the generation significant change, otherwise can think to change not obvious.
6. the update method of To Template as claimed in claim 5 is characterized in that, judging that similarity degree utilizes Pasteur's coefficient to weigh, is judged to be similar in the time of between Pasteur's coefficient is in 0.9～1.
7. such as the update method of each described To Template of claim 1～6, it is characterized in that, in the step 2, when illumination drastic change occurs when, the renewal speed parameter a of To Template as shown in the formula:
a＝(min(abs(σ))，0.3)+min((t-1)
3，9)*0.2*min(abs(σ)/0.5，1))/3；
Wherein σ is the difference of the center of gravity of target histogram and target histogram template;
T is the lasting frame number of illuminance abrupt variation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011102412854A CN102956032A (en) | 2011-08-22 | 2011-08-22 | Target template updating method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011102412854A CN102956032A (en) | 2011-08-22 | 2011-08-22 | Target template updating method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102956032A true CN102956032A (en) | 2013-03-06 |
Family
ID=47764810
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011102412854A Pending CN102956032A (en) | 2011-08-22 | 2011-08-22 | Target template updating method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102956032A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104639843A (en) * | 2014-12-31 | 2015-05-20 | 小米科技有限责任公司 | Method and device for processing image |
| CN106991072A (en) * | 2016-01-21 | 2017-07-28 | 杭州海康威视数字技术股份有限公司 | Automatic measure on line event detection model update method and device |
| CN110580706A (en) * | 2018-06-11 | 2019-12-17 | 北京中科晶上超媒体信息技术有限公司 | Method and device for extracting video background model |
| CN116994215A (en) * | 2023-09-26 | 2023-11-03 | 上海闪马智能科技有限公司 | Road facility abnormality judgment method and device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101645171A (en) * | 2009-09-15 | 2010-02-10 | 湖北莲花山计算机视觉和信息科学研究院 | Background modeling method (method of segmenting video moving object) based on space-time video block and online sub-space learning |
-
2011
- 2011-08-22 CN CN2011102412854A patent/CN102956032A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101645171A (en) * | 2009-09-15 | 2010-02-10 | 湖北莲花山计算机视觉和信息科学研究院 | Background modeling method (method of segmenting video moving object) based on space-time video block and online sub-space learning |
Non-Patent Citations (3)
| Title |
|---|
| 沈志熙等: "《均值漂移算法中的目标模型更新方法研究》", 《自动化学报》 * |
| 王剑等: "《一种基于主动轮廓模型的自适应模板更新算法》", 《光电子技术》 * |
| 王天旭等: "《基于背景模型的车辆目标检测》", 《微电子学与计算机》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104639843A (en) * | 2014-12-31 | 2015-05-20 | 小米科技有限责任公司 | Method and device for processing image |
| CN106991072A (en) * | 2016-01-21 | 2017-07-28 | 杭州海康威视数字技术股份有限公司 | Automatic measure on line event detection model update method and device |
| US11030886B2 (en) | 2016-01-21 | 2021-06-08 | Hangzhou Hikvision Digital Technology Co., Ltd. | Method and device for updating online self-learning event detection model |
| CN106991072B (en) * | 2016-01-21 | 2022-12-06 | 杭州海康威视数字技术股份有限公司 | Online self-learning event detection model updating method and device |
| CN110580706A (en) * | 2018-06-11 | 2019-12-17 | 北京中科晶上超媒体信息技术有限公司 | Method and device for extracting video background model |
| CN116994215A (en) * | 2023-09-26 | 2023-11-03 | 上海闪马智能科技有限公司 | Road facility abnormality judgment method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102339390B (en) | Method and system for updating target template of video monitoring system | |
| US11308335B2 (en) | Intelligent video surveillance system and method | |
| CN110929578B (en) | Anti-shielding pedestrian detection method based on attention mechanism | |
| CN102722725B (en) | Object tracing method based on active scene learning | |
| CN103530893B (en) | Based on the foreground detection method of background subtraction and movable information under camera shake scene | |
| CN111967393A (en) | Helmet wearing detection method based on improved YOLOv4 | |
| CN109410129A (en) | A kind of method of low light image scene understanding | |
| CN103020992B (en) | A kind of video image conspicuousness detection method based on motion color-associations | |
| CN105404847A (en) | Real-time detection method for object left behind | |
| CN111131693B (en) | Face image enhancement method based on multi-exposure face detection | |
| CN102768757B (en) | Remote sensing image color correcting method based on image type analysis | |
| CN111027505B (en) | Hierarchical multi-target tracking method based on significance detection | |
| CN103310422A (en) | Image acquiring method and device | |
| CN105335701A (en) | Pedestrian detection method based on HOG and D-S evidence theory multi-information fusion | |
| CN102956032A (en) | Target template updating method | |
| CN113569882A (en) | Knowledge distillation-based rapid pedestrian detection method | |
| CN105469038A (en) | Safety cap video detection method for electric power switching station | |
| CN104598914A (en) | Skin color detecting method and device | |
| CN103106796A (en) | Vehicle detection method and device of intelligent traffic surveillance and control system | |
| CN102074000A (en) | Tracking method for adaptively adjusting window width by utilizing optimal solution of variance rate | |
| CN108053418B (en) | Animal background modeling method and device | |
| CN111260687A (en) | Aerial video target tracking method based on semantic perception network and related filtering | |
| CN102800105B (en) | Target detection method based on motion vector | |
| CN102592125A (en) | Moving object detection method based on standard deviation characteristic | |
| CN103020987A (en) | Quick foreground detection method based on multi-background model |
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: 20130306 |