CN103514596A - Method and device for processing image - Google Patents
Method and device for processing image Download PDFInfo
- Publication number
- CN103514596A CN103514596A CN201210227213.9A CN201210227213A CN103514596A CN 103514596 A CN103514596 A CN 103514596A CN 201210227213 A CN201210227213 A CN 201210227213A CN 103514596 A CN103514596 A CN 103514596A
- Authority
- CN
- China
- Prior art keywords
- cluster centre
- cluster
- image
- data
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses a method and device for processing an image. The segmentation effect can better conform to the visual effect, the accuracy rate of the segmentation effect is improved, and the number of color varieties generated after the segmentation is as small as possible. The method for processing the image comprises the steps of converting RGB data of the collected image into LUV data, setting the number of the color varieties expected after the segmentation, using the set number of the color varieties as the maximum value of the number used in a clustering center during the segmentation, and utilizing the K-MEANS mode for conducting clustering and segmentation on the LUV data of the image.
Description
Technical field
The present invention relates to digital audio/video technical field, the method and apparatus that particularly a kind of image is processed.
Background technology
It is a basis and important problem that image is segmented in graph and image processing field always, to the further operation of graph image result based on cutting apart all conventionally, the good and bad net result that directly affects graphics process of cutting apart.If under steam road is cut apart inaccurately as the pilotless automobile of controlling by robot, will directly affect automotive performance, even likely can endanger road safety; And skin color segmentation can help people automatically to isolate the body parts such as people's face, hand from background, in the processing of carrying out recognition of face, Expression Recognition, hand tracking, man-machine interaction, movement human target following or porny filtration etc., there is important using value.
Existing scheme one adopts carries out image based on pyramid algorith, mean-shif algorithm and watershed algorithm and cuts apart; Existing scheme two adopts based on some color space, as RGB color space, and in conjunction with the mode of machine learning, carries out image and cuts apart.
In prior art, at least there is following defect:
For scheme one, it is exactly that the color category number obtaining after cutting apart can not be specified flexibly that the partitioning algorithm that existing scheme is used has a common shortcoming, the color category splitting is too much, cause on the one hand follow-up image processing operations complexity, also reflect that the result of cutting apart is not accurate enough on the one hand.And for watershed algorithm, if reach reasonable segmentation effect, also need manual intervention, be not easy to operation, cut apart efficiency lower.
For scheme two, machine learning needs a large amount of samples, and choosing of sample is very complicated, trains also time-consuming.In a word, the cost that current machine learning mode is trained is very high, and calculation of complex efficiency is low, and the same kind of the color after cutting apart that exists too much causes subsequent treatment complexity, cuts apart the problem that accuracy rate is lower.
Summary of the invention
The invention provides the method and apparatus that a kind of image is processed, to solve color category after existing scheme is cut apart, too much to cause successive image to process complexity higher, cuts apart problem that accuracy rate is lower etc.
For achieving the above object, the embodiment of the present invention has adopted following technical scheme:
A kind of method that the embodiment of the present invention provides image to process, comprising:
The RGB data of the image collecting are converted to LUV data;
Rear desired color category number is cut apart in setting;
Using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilize K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
The device that the embodiment of the present invention also provides a kind of image to process, comprising:
Space converting unit, for being converted to LUV data by the RGB data of the image collecting;
Setting unit, cuts apart rear desired color category number for arranging;
Cutting unit, for using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilizes K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
The beneficial effect of the embodiment of the present invention is:
The embodiment of the present invention is cut apart rear desired color category number by setting in advance, and use LUV color space in conjunction with the technological means of the partitioning algorithm based on K-MEANS mode according to set color category number, can make segmentation effect and visual effect more meet, improve the accuracy rate of segmentation effect; And can control flexibly the quantity of cutting apart rear color category, the quantity that rear color category is cut apart in assurance is less, greatly reduces the complexity that successive image is processed.
Accompanying drawing explanation
The method flow diagram that Fig. 1 processes for a kind of image that the embodiment of the present invention one provides;
The structural representation of the device that Fig. 2 processes for a kind of image that the embodiment of the present invention three provides;
Fig. 3 (a) is the original image in experiment one;
Fig. 3 (b) adopts the segmentation result figure after this programme is cut apart to the image in Fig. 3 (a);
Fig. 4 (a) is the original image in experiment two;
Fig. 4 (b) adopts the segmentation result figure after this programme is cut apart to the image in Fig. 4 (a).
Embodiment
For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.
The embodiment of the present invention has adopted image from RGB(RGB) color space conversion is to LUV color space, carries out the mode that image is cut apart in LUV color space.
RGB color space is a kind of conventional color space.Many hardware devices, as physical display, video camera etc., the view data of institute's acquisition and processing is all RGB data.Yet, RGB is mainly the color model towards hardware, this model is generally used for color monitor or colored shooting etc., the color error ratio of itself and the perception of human eye vision institute is larger, and because RGB color space is by the expression of putting together of tone, brightness, three amounts of saturation degree, be difficult to separately, its details is difficult to carry out digitized adjustment, is not easy to carry out concrete image processing operations.
LUV(L represents brightness, luminance; U and V are chromaticity coordinates) color space is CIE
A kind of color space that is different from RGB standard that (international luminous lighting the council) proposes, its objective is and set up the color space (vision appreciable color distinction carried out the coding of unit) unified with vision, be widely used at present computer color vision process field, for general image, the span of chromaticity coordinate U and V is-100 to+100, and brightness L value is 0 to 100.
Because LUV color space is and the unified color space of people's vision, the present embodiment adopts and at LUV color space, to carry out image and cut apart, the effect of cutting apart to improve image.
Referring to Fig. 1, the method that a kind of image providing for the embodiment of the present invention one is processed, comprising:
11: the RGB data of the image collecting are converted to LUV data;
12: arrange and cut apart rear desired color category number;
13: using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilize K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
Execution succession to above-mentioned steps 11 and 12 does not limit.
The embodiment of the present invention is cut apart rear desired color category number by setting in advance, and use LUV color space in conjunction with the technological means of the partitioning algorithm based on K-MEANS mode according to set color category number, can make segmentation effect and visual effect more meet, improve the accuracy rate of segmentation effect; And can control flexibly the quantity of cutting apart rear color category, the quantity that rear color category is cut apart in assurance is less, greatly reduces the complexity that successive image is processed.
The method that the image below embodiment of the present invention two being provided is processed describes, and specifically comprises:
11: the RGB data of the image collecting are converted to LUV data.
This programme is to carry out cluster segmentation on LUV color space basis, but the image obtaining in general collecting device is to use RGB color space, need to carry out space conversion, the method of conversion first becomes XYZ color space (RGB=> XYZ) from RGB color space conversion, then becomes LUV(XYZ=from XYZ color space conversion > LUV).Introduce respectively RGB=below > conversion and the XYZ=of XYZ > conversion of LUV.
One), RGB color space conversion becomes XYZ color space
The concrete operations of conversion can be with following coded representation:
var_R=(R/255) //R from 0 to 255
var_G=(G/255) //G from 0 to 255
var_B=(B/255) //B from 0 to 255
if(var_R>0.04045)var_R=((var_R+0.055)/1.055)^2.4
else var_R=var_R/12.92
if(var_G>0.04045)var_G=((var_G+0.055)/1.055)^2.4
else var_G=var_G/12.92
if(var_B>0.04045)var_B=((var_B+0.055)/1.055)^2.4
else var_B=var_B/12.92
var_R=var_R*100
var_G=var_G*100
var_B=var_B*100
X=var_R*0.4124+var_G*0.3576+var_B*0.1805
Y=var_R*0.2126+var_G*0.7152+var_B*0.0722
Z=var_R*0.0193+var_G*0.1192+var_B*0.9505
Two), XYZ color space conversion becomes LUV color space
var_U=(4*X)/(X+(15*Y)+(3*Z))
var_V=(9*Y)/(X+(15*Y)+(3*Z))
var_Y=Y/100
if(var_Y>0.008856)var_Y=var_Y^(1/3)
else
var_Y=(7.787*var_Y)+(16/116)
ref_X=95.047
ref_Y=100.000
ref_Z=108.883
ref_U=(4*ref_X)/(ref_X+(15*ref_Y)+(3*ref_Z))
ref_V=(9*ref_Y)/(ref_X+(15*ref_Y)+(3*ref_Z))
CIE-L*=(116*var_Y)–16
CIE-u*=13*CIE-L**(var_U-ref_U)
CIE-v*=13*CIE-L**(var_V-ref_V)
12: arrange and cut apart rear desired color category number.
Preferably, the color category number that the present embodiment sets in advance is to be less than 12 numerical value.
13: using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilize K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
The ultimate principle of K-MEANS mode is as follows: accept input quantity k, then n data object is divided into k cluster to obtained cluster is met: the object similarity in same cluster is higher; And object similarity in different clusters is less.Cluster similarity is to utilize the average of object in each cluster to obtain " center object " (center of attraction) to calculate.
In the present embodiment, the main process of K-MEANS mode is as follows: first from n data object (pixel image), select arbitrarily k (as 2) object as initial cluster center; And for other object of be left,, according to the similarity (distance) of they and these cluster centres, respectively they are distributed to (cluster centre representative) cluster the most similar to it; And then calculate the cluster centre (average of all objects in this cluster) of each new cluster that obtains; Constantly repeat this process until meet end condition.
The concrete steps of the K-MEANS mode that the present embodiment provides according to above-mentioned ultimate principle comprise:
S1: choose initial cluster center, the number of described initial cluster center is less than set color category number.The number of cluster centre with cut apart after the piece number of the image block that obtains consistent, also with cut apart after the color category number that obtains consistent.
Exemplary, in the present embodiment, the value of initial cluster center can be defaulted as 2.The present embodiment adopts from less cluster centre, increases gradually the mode of cluster centre in iteration, and the mode than starting just to adopt the cluster centre of the number identical with default color category number, can improve Clustering Effect, thereby improves segmentation accuracy.
S2: each pixel and the distance of initial cluster center described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance.The pixel that is dispensed to same cluster centre forms an image block.
Initial cluster center can be the pixel of choosing at random in image.
The present embodiment is the cutting operation of carrying out under LUV space, adopts the LUV data of pixel to calculate during calculating.Above-mentioned distance adopts Euclidean distance, and concrete account form can be expressed as follows:
Wherein, W
iweight, r
iand s
ito be respectively value and the non-central point value of cluster centre point.
For example, for pixel A, calculate respectively the Euclidean distance of A and cluster centre 1 and cluster centre 2, if the distance of A and cluster centre 1 is less than A and cluster centre 2 distances, the distance that is A and cluster centre 1 is the minimum value in the distance calculating, A is dispensed to cluster centre 1, A belongs to the distance of cluster centre 1 correspondence.
S3: upgrade described initial cluster center and increase the new cluster centre of predetermined number, obtain the cluster centre of iteration for the second time.
Calculate the average of all pixels in the image block that cluster has formed, upgrade described initial cluster center.If the cluster centre number of current use is simultaneously less than the color category number of setting, again select randomly a pixel, to increase a new cluster centre.Be that this enforcement adopts from initial cluster center, during each iteration, only increase the mode of a new cluster centre, to improve the degree of accuracy of cluster.
S4: each pixel and the distance of the cluster centre of iteration for the second time described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance.
Except cluster centre changes, the operation in step S4 is the same with the operation in step S2, is a kind of iterative operation.
S5: judge in each cluster, whether the distance between pixel and cluster centre is less than distance threshold, if, end is cut apart, if not, the cluster centre of iteration be less than the new cluster centre that increases predetermined number when set color category is counted in the number of used cluster centre for the second time described in renewal, obtain the cluster centre of next iteration, carry out next iteration.
In step S5, adopt the mean value of the numerical value of all pixels in the cluster having obtained is upgraded to original cluster centre (cluster centre of iteration for the second time).
If the number of the cluster centre using equals set color category while counting, can not increase again new cluster centre, the number that keeps the cluster centre that last iteration used, after the cluster centre of last iteration is upgraded, continue to carry out iterative process, utilize the average of pixel number certificate in image block to upgrade cluster centre, with the cluster centre after upgrading, repartition image block, in the image block of repartitioning, the average of pixel is upgraded cluster centre again, repetitive operation, until the distance between pixel and this cluster centre is all less than distance threshold in each cluster.Preferably, the numerical value of the distance threshold in the present embodiment is 0.0001.
The present embodiment is when iteration, distance between pixel in each cluster (image block) and cluster centre is less than distance threshold as final end condition, when meeting this end condition, the number of the cluster centre using is less than default color category number or equals default color category number.
The device that the embodiment of the present invention also provides a kind of image to process, referring to Fig. 2, described device comprises:
Setting unit 22, cuts apart rear desired color category number for arranging;
Cutting unit 23, for using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilizes K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
Further, described space converting unit 21, specifically for the RGB data of described image are converted to XYZ data, is converted to LUV data by described XYZ data.
Further, described cutting unit 23 comprises iteration module,
Described iteration module, for choosing initial cluster center, the number of described initial cluster center is less than set color category number; Each pixel and the distance of initial cluster center described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance; Upgrade described initial cluster center and increase the new cluster centre of predetermined number, obtain the cluster centre of iteration for the second time; Each pixel and the distance of the cluster centre of iteration for the second time described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance; Judge in each cluster, whether the distance between pixel and cluster centre is less than distance threshold, if, end is cut apart, if not, the cluster centre of iteration be less than the new cluster centre that increases predetermined number when set color category is counted in the number of used cluster centre for the second time described in renewal, obtain the cluster centre of next iteration, carry out next iteration.
Further, described cutting unit 23 comprises cluster centre update module, for calculating the mean value of the data of all pixels of the corresponding cluster of cluster centre, using this mean value as new cluster centre.
Further, described cluster centre update module, also for choosing at random a pixel as described new cluster centre at image.
The specific works mode of apparatus of the present invention embodiment Zhong Ge unit and module, can be referring to embodiment of the method for the present invention.
From the above mentioned, the embodiment of the present invention is cut apart rear desired color category number by setting in advance, and use LUV color space in conjunction with the technological means of the partitioning algorithm based on K-MEANS mode according to set color category number, can make segmentation effect and visual effect more meet, improve the accuracy rate of segmentation effect; And can control flexibly the quantity of cutting apart rear color category, the quantity that rear color category is cut apart in assurance is less, greatly reduces the complexity that successive image is processed.
Below in conjunction with experimental result picture, further illustrate the beneficial effect of the embodiment of the present invention.
Referring to Fig. 3 (a), show the original image in experiment one.Cutting apart front default color category number is 5 kinds, and distance threshold is 0.0001.Fig. 3 (b) shows the image in Fig. 3 (a) is adopted to the segmentation result figure after this programme is cut apart, and the color category after cutting apart is 5 kinds, and iterations is 14 times.
Referring to Fig. 4 (a), show the original image in experiment two.Cutting apart front default color category number is 7 kinds, and distance threshold is 0.0001.Fig. 4 (b) shows the image in Fig. 4 (a) is adopted to the segmentation result figure after this programme is cut apart, and the color category after cutting apart is 7 kinds, and iterations is 19 times.
The results show, this programme adopts LUV can make segmentation result meet natural vision effect, and on this basis, the kind of color of controlling cluster based on K-MEANS mode is as far as possible little, generally in 12 kinds, thereby has reduced the complexity that successive image is processed.It is fast that this programme can reach image segmentation speed, the technique effect that segmentation effect is good.
The foregoing is only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.All any modifications of doing within the spirit and principles in the present invention, be equal to replacement, improvement etc., be all included in protection scope of the present invention.
Claims (10)
1. the method that image is processed, is characterized in that, described method comprises:
The RGB data of the image collecting are converted to LUV data;
Rear desired color category number is cut apart in setting;
Using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilize K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
2. method according to claim 1, is characterized in that, the described RGB data by the image collecting are converted to LUV data and comprise:
The RGB data of described image are converted to XYZ data;
Described XYZ data are converted to LUV data.
3. method according to claim 1, is characterized in that, described using the maximal value of set color category number number of the cluster centre that uses when cutting apart, and utilizes K-MEANS mode to carry out cluster segmentation to the LUV data of described image and comprises:
Choose initial cluster center, the number of described initial cluster center is less than set color category number;
Each pixel and the distance of initial cluster center described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance;
Upgrade described initial cluster center and increase the new cluster centre of predetermined number, obtain the cluster centre of iteration for the second time;
Each pixel and the distance of the cluster centre of iteration for the second time described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance;
Judge in each cluster, whether the distance between pixel and cluster centre is less than distance threshold, if, end is cut apart, if not, the cluster centre of iteration be less than the new cluster centre that increases predetermined number when set color category is counted in the number of used cluster centre for the second time described in renewal, obtain the cluster centre of next iteration, carry out next iteration.
4. method according to claim 3, is characterized in that, upgrades in the following way cluster centre:
Calculate the mean value of the data of all pixels in the corresponding cluster of cluster centre, using this mean value as new cluster centre.
5. according to the method described in claim 1 to 4 any one, it is characterized in that, the new cluster centre of described increase predetermined number comprises: in image, choose at random a pixel as described new cluster centre.
6. the device that image is processed, is characterized in that, described device comprises:
Space converting unit, for being converted to LUV data by the RGB data of the image collecting;
Setting unit, cuts apart rear desired color category number for arranging;
Cutting unit, for using the maximal value of set color category number number of the cluster centre that uses when cutting apart, utilizes K-MEANS mode to carry out cluster segmentation to the LUV data of described image.
7. device according to claim 6, is characterized in that, described space converting unit, specifically for the RGB data of described image are converted to XYZ data, is converted to LUV data by described XYZ data.
8. device according to claim 6, is characterized in that, described cutting unit comprises iteration module,
Described iteration module, for choosing initial cluster center, the number of described initial cluster center is less than set color category number; Each pixel and the distance of initial cluster center described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance; Upgrade described initial cluster center and increase the new cluster centre of predetermined number, obtain the cluster centre of iteration for the second time; Each pixel and the distance of the cluster centre of iteration for the second time described in each in computed image, and pixel is dispensed to the minimum cluster centre of distance; Judge in each cluster, whether the distance between pixel and cluster centre is less than distance threshold, if, end is cut apart, if not, the cluster centre of iteration be less than the new cluster centre that increases predetermined number when set color category is counted in the number of used cluster centre for the second time described in renewal, obtain the cluster centre of next iteration, carry out next iteration.
9. device according to claim 8, is characterized in that, described cutting unit comprises cluster centre update module,
Described cluster centre update module, for calculating the mean value of the data of all pixels of the corresponding cluster of cluster centre, using this mean value as new cluster centre.
10. according to the device described in claim 6 to 9 any one, it is characterized in that, described cluster centre update module, also for choosing at random a pixel as described new cluster centre at image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210227213.9A CN103514596A (en) | 2012-06-30 | 2012-06-30 | Method and device for processing image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210227213.9A CN103514596A (en) | 2012-06-30 | 2012-06-30 | Method and device for processing image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103514596A true CN103514596A (en) | 2014-01-15 |
Family
ID=49897274
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210227213.9A Pending CN103514596A (en) | 2012-06-30 | 2012-06-30 | Method and device for processing image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103514596A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105427260A (en) * | 2015-11-27 | 2016-03-23 | 优渊商贸（上海）有限公司 | Image processing method and apparatus |
| CN106104577A (en) * | 2014-03-07 | 2016-11-09 | 高通股份有限公司 | Photo management |
| CN106568782A (en) * | 2016-11-05 | 2017-04-19 | 上海大学 | Machine vision-based method for conducting color quantization of colored bottle cap image |
| CN109948548A (en) * | 2019-03-20 | 2019-06-28 | 齐鲁工业大学 | A kind of the lipstick recommended method and system of the match colors based on machine learning |
| CN110751654A (en) * | 2019-08-30 | 2020-02-04 | 稿定（厦门）科技有限公司 | Image matting method, medium, equipment and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6012058A (en) * | 1998-03-17 | 2000-01-04 | Microsoft Corporation | Scalable system for K-means clustering of large databases |
| CN102222234A (en) * | 2011-07-14 | 2011-10-19 | 苏州两江科技有限公司 | Image object extraction method based on mean shift and K-means clustering technology |
| CN102509097A (en) * | 2011-09-29 | 2012-06-20 | 北京新媒传信科技有限公司 | Method and device for image segmentation |
-
2012
- 2012-06-30 CN CN201210227213.9A patent/CN103514596A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6012058A (en) * | 1998-03-17 | 2000-01-04 | Microsoft Corporation | Scalable system for K-means clustering of large databases |
| CN102222234A (en) * | 2011-07-14 | 2011-10-19 | 苏州两江科技有限公司 | Image object extraction method based on mean shift and K-means clustering technology |
| CN102509097A (en) * | 2011-09-29 | 2012-06-20 | 北京新媒传信科技有限公司 | Method and device for image segmentation |
Non-Patent Citations (2)
| Title |
|---|
| 周明全 等: "《基于内容图像检索技术》", 31 July 2007, article "CIE颜色空间", pages: 19-22 * |
| 谢娟英 等: "一种改进的全局K-均值聚类算法", 《陕西师范大学学报（自然科学版）》, vol. 38, no. 2, 31 March 2010 (2010-03-31), pages 18 - 22 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106104577A (en) * | 2014-03-07 | 2016-11-09 | 高通股份有限公司 | Photo management |
| CN106104577B (en) * | 2014-03-07 | 2020-07-03 | 高通股份有限公司 | Photo management |
| CN105427260A (en) * | 2015-11-27 | 2016-03-23 | 优渊商贸（上海）有限公司 | Image processing method and apparatus |
| CN106568782A (en) * | 2016-11-05 | 2017-04-19 | 上海大学 | Machine vision-based method for conducting color quantization of colored bottle cap image |
| CN109948548A (en) * | 2019-03-20 | 2019-06-28 | 齐鲁工业大学 | A kind of the lipstick recommended method and system of the match colors based on machine learning |
| CN110751654A (en) * | 2019-08-30 | 2020-02-04 | 稿定（厦门）科技有限公司 | Image matting method, medium, equipment and device |
| CN110751654B (en) * | 2019-08-30 | 2022-06-28 | 稿定（厦门）科技有限公司 | Image matting method, medium, equipment and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112418722B (en) | Non-invasive load identification method based on V-I (velocity-amplitude) trajectory graph and neural network | |
| CN107145889B (en) | Target identification method based on double CNN network with RoI pooling | |
| CN110472467A (en) | The detection method for transport hub critical object based on YOLO v3 | |
| CN102968782B (en) | In a kind of coloured image, remarkable object takes method automatically | |
| CN103514596A (en) | Method and device for processing image | |
| CN111864896B (en) | Power load monitoring method and system | |
| CN102509099B (en) | Detection method for image salient region | |
| CN103473550B (en) | Based on the leaf image dividing method of Lab space and local dynamic threshold | |
| CN103699900B (en) | Building horizontal vector profile automatic batch extracting method in satellite image | |
| CN104143186B (en) | A kind of SLIC super-pixel piecemeal optimization method | |
| CN112215795B (en) | Intelligent detection method for server component based on deep learning | |
| CN107068044B (en) | Image gamma revision method, scanning card and display system | |
| CN110276764A (en) | K-Means underwater picture background segment innovatory algorithm based on the estimation of K value | |
| CN111833590B (en) | Traffic signal lamp control method and device and computer readable storage medium | |
| CN109766828A (en) | A kind of vehicle target dividing method, device and communication equipment | |
| CN112906529A (en) | Face recognition light supplementing method and device, face recognition equipment and face recognition system | |
| CN113255434A (en) | Apple identification method fusing fruit features and deep convolutional neural network | |
| CN106295645A (en) | A kind of license plate character recognition method and device | |
| CN113691954B (en) | Comprehensive energy consumption management system based on cloud computing | |
| CN109740553B (en) | Image semantic segmentation data screening method and system based on recognition | |
| CN104794726B (en) | A kind of underwater picture Parallel segmentation method and device | |
| CN111931724A (en) | Signal lamp abnormity identification method and device, electronic equipment and road side equipment | |
| CN113923839A (en) | Light adjusting method and system | |
| CN112132232A (en) | Medical image classification labeling method and system and server | |
| CN102043963A (en) | Method for recognizing and counting number of people in image |
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: 20140115 |