CN102760289A - Embedded complex connected domain searching method - Google Patents
Embedded complex connected domain searching method Download PDFInfo
- Publication number
- CN102760289A CN102760289A CN2011101092562A CN201110109256A CN102760289A CN 102760289 A CN102760289 A CN 102760289A CN 2011101092562 A CN2011101092562 A CN 2011101092562A CN 201110109256 A CN201110109256 A CN 201110109256A CN 102760289 A CN102760289 A CN 102760289A
- Authority
- CN
- China
- Prior art keywords
- connected domain
- pixel
- ltregion
- scanning
- labeled graph
- 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 relates to an embedded complex connected domain searching method. The method comprises steps of first, reading image data and performing binarization; creating a labeled graph flagmap and LT (language translation) Region information sheet; scanning an image from top to bottom and from left to right; storing information of left and right end points of a searched connected domain in the LTRegion information sheet of the connected domain; marking on the flagmap; combining redundant connected domains by utilizing the LTRegion information sheet and the LTRegion labeled graph; and finally outputting a result. The embedded complex connected domain searching method has the advantages that the method is non-recursive, the amount of codes is low, occupied resources of an operation system and hardware of a processor are fewer, and scanning speed is fast. Besides, multistage forky complex connected domain searching can be finished excellently, robust performance is good, searching speed almost is hardly affected by the shape and the number of objects, and the method has an excellent application value, and is suitable for various embedded hardware platforms.
Description
Technical field
The invention belongs to computer vision technique and machine vision technique field, relate to a kind of embedded complicated connected domain searching method, specifically relate to a kind of new complicated connected domain onrecurrent searching method that can in embedded system, use.
Background technology
It is a kind of important technology in the computer vision that Blob detects (also being called as spot detection), and it provides topological structure, size, shape, position and the directional information of spot in the image for vision system.Many computation vision applications such as the Blob detection has been applied in OCR identification, the banknote printing, circuit board detecting, robot location, chip pin detection, interaction multimedia amusement.Complicated connected domain search is the treatment scheme of a key in Blob detects, and Blob detects to handle and must carry out the connected domain search, so that come out all isolated area marks in the image, for follow-up judgement identification provides necessary information.Therefore the speed of connected domain search and program complexity are extremely important for the handling property that improves the computation vision system.
Along with the development of microelectronics, sensor and computer science has driven the development of EMBEDDED AVIONICS, the volume of embedded electronic device is little, and it is redundant to have eliminated software and hardware, low in energy consumption, flexible and convenient to use, applied range.With the embedded electronic product of consumer electronics real-time, speed and the system resource of visual processes software are had higher requirement in the industry, therefore must develop the visual processing method that is suitable for the embedded platform use.Therefore current jumbo cost of memory is very cheap, can usage space on embedded platform changes the connected domain disposal route of time.
Connected domain searching method commonly used at present is to use recurrence or chained list to realize that it is more to take operating-system resources, and program complexity is big, and mark merging method is complicated, and dealing with complicated object connected domain speed is slow.Document " a kind of new quick labeling algorithm in image connectivity territory " (Song Bin. a kind of new quick labeling algorithm in image connectivity territory. electronic measurement technique; 2009.9; 32 (9): 67-73), " a kind of label based on line is propagated binary picture connected component method for quick " (Zhang Shusheng. a kind of label based on line is propagated binary picture connected component method for quick. computer research and development; 1994; 31 (10): the quick connected domain searching method to simple objects has been proposed 51-54), but lower to the complex object connected domain search efficiency of multistage bifurcated; Patent " point target connected domain real-time mark and recognition methods that a kind of parallel pipeline the is realized " (patent No.: 200910090766.2; Applicant: Zhou Ping; Liu Yue has proposed the quick connected domain searching method to simple wisp in Wang Yongtian), and document " connected domain analytical algorithm and realization thereof fast " (hole is refined. connected domain analytical algorithm and realization thereof fast. and pattern-recognition and artificial intelligence; 2003; 16 (1): 1102115.) the middle chain type mechanism of using solves connected domain search and computational problem, but when a plurality of connected domains were searched for, these two kinds of method efficient can significantly reduce.
Summary of the invention
To take operating-system resources in the existing connected domain searching method more in order to solve, and program complexity is big, and mark merging method is complicated, and the slow-footed deficiency of dealing with complicated object connected domain the present invention proposes a kind of embedded complicated connected domain searching method.The connected domain detection method that the present invention realizes have a kind of non-recursive, size of code is little, take operating system and the processor hardware resource is few, sweep velocity is fast advantage; Can well accomplish the complicated connected domain search of multistage bifurcated; And robust performance is good; Search speed receives the influence of body form and number hardly, is suitable for various embedded hardware platforms and uses.
The technical solution adopted for the present invention to solve the technical problems is: a kind of embedded complicated connected domain searching method comprises following steps:
(1) reads in view data and binaryzation;
(2) create flagmap labeled graph and LTRegion information table
Set up a labeled graph two-dimensional array flagmap consistent and a connected domain information table LTRegion with area size to be searched; Labeled graph two-dimensional array flagmap is used for the different connected domain of mark; Connected domain information table LTRegion is used to store border, left and right sides endpoint location, validity flag and the encirclement frame information of all connected domains, and the type of each list item of connected domain information table LTRegion is Region;
(3) from top to bottom, from a left side and image of right scanning is stored in the connected domain left and right sides terminal point information that searches among the connected domain information table LTRegion, and in the flagmap labeled graph mark
1. from bottom to top, from a left side and image of right scanning is counted scan line through linage-counter i, every scanning finish delegation, and linage-counter adds 1, if linage-counter withdraws from the line scanning processing greater than picturedeep;
2. in each line scanning process; Read earlier current pixel point t, left pixel td and with the lower left tdl of current pixel point; Under the gray-scale value of pixel tdc and lower right pixel tdr, through the finite states machine control scanning process, whole line scanning process comprises initialization " init "; Line scanning " linescan "; Merge " Uniting ", four processes of background scans " Backscan ", in scanning through judge current pixel td with below the relation of neighbor accomplish establishment, mark and the union operation of connected domain.
3. each line scanning begins, and state machine is introduced into " init " state, if gray scale becomes white at the current pixel point place; Just td and t gray scale difference are greater than setting value thresh; Create a new connected domain, be numbered ID, with the left margin point of current point as new connected domain; And store in the connected domain information table left end point variable, state machine jumps to " linescan " state then; If td and t gray scale difference are less than setting value thresh, linage-counter n adds 1 state machine and jumps to " Backscan " state.At " linescan " state; If current pixel t gray scale greater than thresh and lower pixel less than thresh; Use the row-coordinate of t to upgrade the right margin end points counter of regional ID, and on the labeled graph picture the corresponding new connected domain sequence number ID of location of pixels mark, linage-counter adds 1 then; In " linescan " state; If current pixel t with under pixel tdown all greater than thresh; Numerical value in the labeled graph of adjacent position, current t below is composed to variable ID1; Just use ID1 to preserve current pixel below three neighbors (tdownr, tdownc, connected component labelings townl).At merging phase " Uniting "; If current pixel is greater than thresh; Just the right endpoint of connected domain ID1 is set to current pixel, if current pixel less than thresh, all white pixel positions on the current pixel and the left side are set to ID1 on labeled graph; Just these continuous white pixel of current line are merged among the connected domain ID1, state switches to " Backscan " then; At " Backscan " if in current pixel be white (gray scale is greater than thresh), then state switches to " linescan " again, does not blame column counter n and adds 1, and 2. each row of image is carried out and scan operation 3., finishes up to scanning last column;
(4) after step (3) is accomplished,, calculate the encirclement frame of the connected domain that has searched, be stored among the connected domain information table LTRegion according to the left and right sides end points coordinate of record;
(5) use LTRegion information table and the merging of LTRegion labeled graph more than connected region
(3) scanning processes of step can only be handled the connection search of solid body, after pertusate connected domain scanning, also need carry out union operation; To any connected domain K; Variable combine of elder generation's initialization is zero, judges then whether connected domain validity flag Regiondeleted equals zero, if Regiondeleted is zero; Then scan right-hand member from the connected domain left end; Whether neighbor belongs to the another one connected domain above each pixel of inspection connected domain coboundary on the labeled graph flagmap, if belong to another one connected region ObjectID, it is 1 that combine is set simultaneously; After connected domain has from left to right scanned; Judge again whether combine equals 1,, then connected region K and connected region ObjectID are merged if equal 1; Upgrade the corresponding information table of connected region ObjectID (about and up-and-down boundary); Be that the unit of K all is revised as ObjectID with all index value on the labeled graph flagmap position then, it is invalid that current connected region K is set at last, and the connected domain validity flag Regiondeleted that K unit of connected domain information table promptly is set is 1;
Repeat this step 2 times, accomplish the merging of all adjacent connected regions;
(6) finish the output result.
Further, in the step (2), use Region class data structure to preserve connected domain information, each connected domain information stores is in a Region variable; Its member comprises left and right sides terminal point information, label, effective character state variable; Surround frame, sense of rotation, filling rate and labeled graph pointer etc.
Further; The deletion action of unnecessary connected domain all realizes through the Regiondeleted state variable of Region class in the step (5); When certain connected domain because merge or filter operation is made as 1 with the Regiondeleted variable when being judged as illegal (undesirable); The expression deletion, otherwise remain 0, expression is preserved.
Compared with prior art, the invention has the beneficial effects as follows:
(1) search speed is fast, uses single pass and one to two sub-region to merge to image and can obtain all connected regions in the image;
(2) applied widely, can accomplish multiple hole and various concavo-convex polygonal connected domain search, can be used for landmark identification, action recognition, feature point detection, robot guiding, many field of machine vision such as character recognition;
(3) it is few to take processor resource, uses non-recursive labeled graph and connected domain information table mark Programming Strategy, has avoided a large amount of time-consuming stack manipulations, is applicable to embedded system;
(4) be convenient to transplant, use the standard C language exploitation to form, program and hardware platform are irrelevant, can be advantageously used in hardware platforms such as ARM, DSP and x86.
Description of drawings
Fig. 1 is a labeled graph flagmap synoptic diagram of the present invention.
Fig. 2 is a flow chart of steps of the present invention.
Fig. 3 scans process flow diagram the first time of the present invention.
Fig. 4 is an Init state processing process flow diagram of the present invention.
Fig. 5 is a Linescan state processing process flow diagram of the present invention.
Fig. 6 is a Uniting state processing process flow diagram of the present invention.
Fig. 7 is a Backscan state processing process flow diagram of the present invention.
Fig. 8 is that adjacent connected domain of the present invention merges processing flow chart at last.
Fig. 9 is to use this method to handle the result of bianry image.
Figure 10 is to use Blob sreen analysis method locating circuit board reference point example.
Figure 11 is to use Blob sreen analysis method location to have a part center of complex background.
Figure 12 is to use Blob sreen analysis method to detect the hole in piece part manufacturing deficiency.
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is further specified.
Referring to Fig. 1; The labeled graph flagmap of the use that the present invention uses is used for the different connected domain of mark; Among the figure " 1 " and " 2 " be two connected regions that flock together, the LTRegion table is used to store the information such as border, left and right sides endpoint location, validity flag and encirclement frame of all connected domains.The type of each list item of LTRegion table is Region, uses Region class data structure to preserve connected domain information, and connected domain information is carried out closed management, makes this method be convenient to platform transplantation.Each connected domain information stores is in a Region variable, and its member comprises left and right sides terminal point information, label, effective character state variable; Surround frame, sense of rotation, filling rate and labeled graph pointer etc.; So that accomplish the connected domain search fast, area filters, and filling rate filters; Operations such as connected domain direction and shape facility calculating for follow-up fast B lob analyzes, provide essential parameter information.The Region class definition is following:
Before getting into the first pass image, at first labeled graph and connected domain information table are carried out initialization, with the flagmap zero clearing, left leftPOINT [.] and right-hand member point value rightPOINT [.] that each Region categorical variable in the LTRegion table is set are-1.
Referring to Fig. 2, the present invention comprises following steps:
(1) reads in view data and binaryzation;
(2) create flagmap labeled graph and LTRegion information table
(3) from top to bottom, from a left side and image of right scanning is stored in the connected domain left and right sides terminal point information that searches among the connected domain information table LTRegion, and in the flagmap marked.Scan process flow diagram for the first time referring to Fig. 3, the C among Fig. 3, D, E, F are four kinds of state processing, and i representes the row-coordinate variable; N representes the row coordinate variable, Height presentation video height, Width presentation video width; Lstate representes state variable, Img presentation video matrix array, and Img coordinate representation is (n; I), in the line scanning initialization
t＝Img(n，i)； td＝Img(n-1，i)；
tdownc＝Img(n，i-1)；tdownr＝Img(n-1，i-1)；
tdownl＝Img(n+1，i-1)；
Calculate lower pixel value summation tdown=tdownr+tdownc+tdownl;
Lower right pixel index pindexr=(m-1) * Width+n+1;
Lower left pixel index pindexl=(m-1) * Width+n-1;
In Fig. 3-7, the implementation of this step is following:
1. from bottom to top, from a left side and image of right scanning is counted scan line through linage-counter i, every scanning finish delegation, and linage-counter adds 1, if linage-counter withdraws from the line scanning processing greater than picturedeep;
2. in each line scanning process; Read earlier current pixel point t, left pixel td and with the lower left tdl of current pixel point; Under the gray-scale value of pixel tdc and lower right pixel tdr, through the finite states machine control scanning process, whole line scanning process comprises initialization " init "; Line scanning " linescan "; Merge " Uniting ", four processes of background scans " Backscan ", in scanning through judge current pixel td with below the relation of neighbor accomplish establishment, mark and the union operation of connected domain.
3. each line scanning begins, and state machine is introduced into " init " state, if gray scale becomes white at the current pixel point place; Just td and t gray scale difference are greater than setting value thresh; Create a new connected domain, be numbered ID, with the left margin point of current point as new connected domain; And store in the connected domain information table left end point variable, state machine jumps to " linescan " state then; If td and t gray scale difference are less than setting value thresh, linage-counter n adds 1 state machine and jumps to " Backscan " state.At " linescan " state; If current pixel t gray scale greater than thresh and lower pixel less than thresh; Use the row-coordinate of t to upgrade the right margin end points counter of regional ID, and on the labeled graph picture the corresponding new connected domain sequence number ID of location of pixels mark, linage-counter adds 1 then; In " linescan " state; If current pixel t with under pixel tdown all greater than thresh; Numerical value in the labeled graph of adjacent position, current t below is composed to variable ID1; Just use ID1 to preserve current pixel below three neighbors (tdownr, tdownc, connected component labelings townl).At merging phase " Uniting "; If current pixel is greater than thresh; Just the right endpoint of connected domain ID1 is set to current pixel, if current pixel less than thresh, all white pixel positions on the current pixel and the left side are set to ID1 on labeled graph; Just these continuous white pixel of current line are merged among the connected domain ID1, state switches to " Backscan " then; At " Backscan " if in current pixel be white (gray scale is greater than thresh), then state switches to " linescan " again, does not blame column counter n and adds 1, and 2. each row of image is carried out and scan operation 3., finishes up to scanning last column;
(4) after step (3) is accomplished,, calculate the encirclement frame of the connected domain that has searched, be stored among the connected domain information table LTRegion according to the left and right sides end points coordinate of record;
(5) use LTRegion information table and the merging of LTRegion labeled graph more than connected region
Referring to Fig. 8, (3) scanning processes of step can only be handled the connection search of solid body, after pertusate connected domain scanning; Also need carry out union operation, to any connected domain K, variable combine of first initialization is zero; Judge then whether connected domain validity flag Regiondeleted equals zero; If Regiondeleted is zero, then scan right-hand member from the connected domain left end, whether neighbor belongs to the another one connected domain above each pixel of inspection connected domain coboundary on the labeled graph flagmap; If belong to another one connected region ObjectID, it is 1 that combine is set simultaneously; After connected domain has from left to right scanned; Judge again whether combine equals 1,, then connected region K and connected region ObjectID are merged if equal 1; Upgrade the corresponding information table of connected region ObjectID (about and up-and-down boundary); Be that the unit of K all is revised as ObjectID with all index value on the labeled graph flagmap position then, it is invalid that current connected region K is set at last, and the connected domain validity flag Regiondeleted that K unit of connected domain information table promptly is set is 1; Repeat this step 2 times, accomplish the merging of all adjacent connected regions.
During practical implementation of the present invention; Computing machine obtains piece image from harddisk memory or outside collecting device; It is put into computer-internal random access memory RAM according to the BMP bitmap format, if this width of cloth image is a coloured image, earlier with its converted 8bits gray level image form; Then this image is carried out binaryzation, then directly carry out binaryzation for the 8bits gray level image.Preferably once corrode and expansive working after the binaryzation, so that the assorted point on the removal of images.Create following size and handle identical labeled graph flagmap and the LTRegion connected domain information table of image by application program then.
Algorithm of the present invention is encapsulated in RegionSearch (void*proc_img, int Width, int Height; Int*flagmap, Region*LTRegions) in the function, proc_img is an image cursor; Width is a picture traverse; Height is a picture altitude, and flagmap is the labeled graph pointer, and the LTRegions parameter is returned the connected domain information that searches.System software is at the internal memory that calls the good labeled graph flagmap of static allocation and connected domain information table LTRegions before the RegionSearch.The clauses and subclauses sum of LTRegions table is maxNum.When using the RegionSearch function, import processing image cursor parameter p roc_img, labeled graph internal memory pointer flagmap and connected domain pointer LTRegions into to function.After function executes; Flagmap will deposit the unique identifying number of connected domain at image respective pixel coordinate position; LTRegions returns border, encirclement frame, the validity information of all connected domains; They are used by follow-up Blob analytic function, so that be used for calculating filling rate, and characteristics such as object size and object center of gravity.
When algorithm of the present invention moves in the x86 of cpu frequency 2.4GHz platform validation, for 640*480 image connectivity search time be 8ms, 1280*1024 sized images search time is 32ms.When this algorithm moves at the DSP embedded platform; Be applied in the vision systems such as the online detection of Motor Components, workpiece size vision measurement; The frequency of DSP is 600MHz; Figure 10 and Figure 11 are to use this algorithm location workpiece and characteristic picture, image size 720*576, and locating speed is less than 0.1 second.This algorithm also is applied in the ARM9 flush bonding processor, running frequency 400MHz, and image size 640*480 is about object identification speed 0.2s.
Fig. 9 is to use this method to handle the result of bianry image; Red encirclement frame is used for sign and searches object; After detecting object object is shown as green, can finds out, adopt this method; Can detect all connected domains in the same width of cloth image simultaneously, also can detect for the complicated line image in the upper right corner.
Figure 10 is to use Blob sreen analysis method locating circuit board reference point example; Use image segmentation algorithm that Figure 10 left image is cut apart earlier; Use this paper method to search for out the white portion of all connections then; The connected region of the reference point that comes out according to the area and the size identification of connected region then, the center of calculating the reference point zone at last are as the initial alignment point, and the right figure of Figure 10 is the result of Blob location.Use Blob positioning circuit plate technique, be applied in circuit board automatic punching equipment and the circuit board optical check instrument and equipment.
Figure 11 is to use Blob sreen analysis method location to have a part center of complex background; Use image segmentation algorithm that Figure 11 left image is cut apart earlier; Use this paper method to search for out the white portion of all connections then; The center of part is calculated in the part zone of coming out according to the size identification of connected region then at last, and the right figure of Figure 11 is the result of Blob location.
Figure 12 is to use Blob sreen analysis method to detect the hole in piece part manufacturing deficiency, earlier the right-hand image of Figure 12 is cut apart and inverse, uses this paper method to search for out bore region then; Whether meet the requirements according to the area of bore region then and judge the product quality quality; Because each boring, the shape in hole is all inconsistent, and the part angle is at random; So adopt fixed form matching process false determination ratio high, and processing speed is slow.
Claims (3)
1. embedded complicated connected domain searching method is characterized in that: comprise following steps:
(1) reads in view data and binaryzation;
(2) create flagmap labeled graph and LTRegion information table
Set up a labeled graph two-dimensional array flagmap consistent and a connected domain information table LTRegion with area size to be searched; Labeled graph two-dimensional array flagmap is used for the different connected domain of mark; Connected domain information table LTRegion is used to store border, left and right sides endpoint location, validity flag and the encirclement frame information of all connected domains, and the type of each list item of connected domain information table LTRegion is Region;
(3) from top to bottom, from a left side and image of right scanning is stored in the connected domain left and right sides terminal point information that searches among the connected domain information table LTRegion, and in the flagmap labeled graph mark
1. from bottom to top, from a left side and image of right scanning is counted scan line through linage-counter i, every scanning finish delegation, and linage-counter adds 1, if linage-counter withdraws from the line scanning processing greater than picturedeep;
2. in each line scanning process; Read earlier current pixel point t, left pixel td and with the lower left tdl of current pixel point; Under the gray-scale value of pixel tdc and lower right pixel tdr, through the finite states machine control scanning process, whole line scanning process comprises initialization " init "; Line scanning " linescan "; Merge " Uniting ", four processes of background scans " Backscan ", in scanning through judge current pixel td with below the relation of neighbor accomplish establishment, mark and the union operation of connected domain;
3. each line scanning begins, and state machine is introduced into " init " state, if gray scale becomes white at the current pixel point place; Just td and t gray scale difference are greater than setting value thresh; Create a new connected domain, be numbered ID, with the left margin point of current point as new connected domain; And store in the connected domain information table left end point variable, state machine jumps to " linescan " state then; If td and t gray scale difference are less than setting value thresh; Linage-counter n adds 1 state machine and jumps to " Backscan " state; At " linescan " state, if current pixel t gray scale greater than thresh and lower pixel less than thresh, use the row-coordinate of t to upgrade the right margin end points counter of regional ID; And on the labeled graph picture the corresponding new connected domain sequence number ID of location of pixels mark, linage-counter adds 1 then; In " linescan " state, if current pixel t with under pixel tdown all greater than thresh, numerical value in the labeled graph of adjacent position, current t below is composed to variable ID1; Just use ID1 to preserve three the neighbor (tdownr in current pixel below; Tdownc, connected component labeling townl) is at merging phase " Uniting "; If current pixel is greater than thresh; Just the right endpoint of connected domain ID1 is set to current pixel, if current pixel less than thresh, all white pixel positions on the current pixel and the left side are set to ID1 on labeled graph; Just these continuous white pixel of current line are merged among the connected domain ID1, state switches to " Backscan " then; At " Backscan " if in current pixel be white (gray scale is greater than thresh), then state switches to " linescan " again, does not blame column counter n and adds 1, and 2. each row of image is carried out and scan operation 3., finishes up to scanning last column;
(4) after step (3) is accomplished,, calculate the encirclement frame of the connected domain that has searched, be stored among the connected domain information table LTRegion according to the left and right sides end points coordinate of record;
(5) use LTRegion information table and the merging of LTRegion labeled graph more than connected region
(3) scanning processes of step can only be handled the connection search of solid body, after pertusate connected domain scanning, also need carry out union operation; To any connected domain K; Variable combine of elder generation's initialization is zero, judges then whether connected domain validity flag Regiondeleted equals zero, if Regiondeleted is zero; Then scan right-hand member from the connected domain left end; Whether neighbor belongs to the another one connected domain above each pixel of inspection connected domain coboundary on the labeled graph flagmap, if belong to another one connected region ObjectID, it is 1 that combine is set simultaneously; After connected domain has from left to right scanned; Judge again whether combine equals 1,, then connected region K and connected region ObjectID are merged if equal 1; Upgrade the corresponding information table of connected region ObjectID (about and up-and-down boundary); Be that the unit of K all is revised as ObjectID with all index value on the labeled graph flagmap position then, it is invalid that current connected region K is set at last, and the connected domain validity flag Regiondeleted that K unit of connected domain information table promptly is set is 1; Repeat this step 2 times, accomplish the merging of all adjacent connected regions;
(6) finish the output result.
2. a kind of embedded complicated connected domain searching method according to claim 1 is characterized in that: in the step (2), use Region class data structure to preserve connected domain information; Each connected domain information stores is in a Region variable, and its member comprises left and right sides terminal point information, label; The effective character state variable; Surround frame, sense of rotation, filling rate and labeled graph pointer.
3. a kind of embedded complicated connected domain searching method according to claim 1; It is characterized in that: the deletion action of unnecessary connected domain all realizes through the Regiondeleted state variable of Region class in the step (5); When certain connected domain because merge or filter operation is made as 1 with the Regiondeleted variable when being judged as illegal (undesirable); The expression deletion, otherwise remain 0, expression is preserved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101092562A CN102760289A (en) | 2011-04-29 | 2011-04-29 | Embedded complex connected domain searching method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2011101092562A CN102760289A (en) | 2011-04-29 | 2011-04-29 | Embedded complex connected domain searching method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102760289A true CN102760289A (en) | 2012-10-31 |
Family
ID=47054742
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2011101092562A Pending CN102760289A (en) | 2011-04-29 | 2011-04-29 | Embedded complex connected domain searching method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102760289A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127786A (en) * | 2016-07-04 | 2016-11-16 | 大连理工大学 | The Fast Calibration of a kind of complicated connected region feature and extracting method |
| WO2017113422A1 (en) * | 2015-12-31 | 2017-07-06 | 深圳配天智能技术研究院有限公司 | Image speckle processing method |
| CN107045727A (en) * | 2017-03-27 | 2017-08-15 | 中国科学院深圳先进技术研究院 | A kind of texture synthesis method and its device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101499169A (en) * | 2009-02-13 | 2009-08-05 | 北京中星微电子有限公司 | Method and apparatus for scanning connected region in combined image |
| WO2009150236A1 (en) * | 2008-06-12 | 2009-12-17 | Renaudie, Franck Sébastien | Method and device for image processing, particularly for medical image processing |
-
2011
- 2011-04-29 CN CN2011101092562A patent/CN102760289A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009150236A1 (en) * | 2008-06-12 | 2009-12-17 | Renaudie, Franck Sébastien | Method and device for image processing, particularly for medical image processing |
| CN101499169A (en) * | 2009-02-13 | 2009-08-05 | 北京中星微电子有限公司 | Method and apparatus for scanning connected region in combined image |
Non-Patent Citations (2)
| Title |
|---|
| KUFFNER, J.J.等: "RRT-connect: An efficient approach to single-query path planning", 《ROBOTICS AND AUTOMATION, 2000. PROCEEDINGS. ICRA "00. IEEE INTERNATIONAL CONFERENCE ON》 * |
| 王晓东等: "嵌入式DSP激光雷达距离图像处理系", 《仪器仪表学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017113422A1 (en) * | 2015-12-31 | 2017-07-06 | 深圳配天智能技术研究院有限公司 | Image speckle processing method |
| CN107615333A (en) * | 2015-12-31 | 2018-01-19 | 深圳配天智能技术研究院有限公司 | The spot processing method of image |
| CN106127786A (en) * | 2016-07-04 | 2016-11-16 | 大连理工大学 | The Fast Calibration of a kind of complicated connected region feature and extracting method |
| CN106127786B (en) * | 2016-07-04 | 2018-12-18 | 大连理工大学 | A kind of Fast Calibration and extracting method of complexity connected region feature |
| CN107045727A (en) * | 2017-03-27 | 2017-08-15 | 中国科学院深圳先进技术研究院 | A kind of texture synthesis method and its device |
| CN107045727B (en) * | 2017-03-27 | 2020-07-24 | 中国科学院深圳先进技术研究院 | Texture synthesis method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20140003723A1 (en) | Text Detection Devices and Text Detection Methods | |
| CN108305243A (en) | A kind of magnetic tile surface defect detection method based on deep learning | |
| CN111368852A (en) | Article identification and pre-sorting system and method based on deep learning and robot | |
| Wu et al. | Pcbnet: A lightweight convolutional neural network for defect inspection in surface mount technology | |
| US11100303B2 (en) | Method, system and apparatus for auxiliary label detection and association | |
| CN109784250A (en) | The localization method and device of automatically guiding trolley | |
| CN109271842A (en) | A kind of generic object detection method, system, terminal and storage medium returned based on key point | |
| CN109190662A (en) | A kind of three-dimensional vehicle detection method, system, terminal and storage medium returned based on key point | |
| CN102760289A (en) | Embedded complex connected domain searching method | |
| CN113505808A (en) | Detection and identification algorithm for power distribution facility switch based on deep learning | |
| CN113239227A (en) | Image data structuring method and device, electronic equipment and computer readable medium | |
| CN111167731B (en) | Product sorting method, product sorting system and intelligent sorting robot | |
| Xiao et al. | A real-time system for lane detection based on FPGA and DSP | |
| CN114444997A (en) | Intelligent warehousing system based on supply chain management | |
| Wong et al. | Tracking indoor construction progress by deep-learning-based analysis of site surveillance video | |
| CN111027601B (en) | Plane detection method and device based on laser sensor | |
| Drouilly et al. | Fast hybrid relocation in large scale metric-topologic-semantic map | |
| Morales Rosales et al. | On-road obstacle detection video system for traffic accident prevention | |
| Karrach et al. | Options to use data matrix codes in production engineering | |
| Yang et al. | Explorations on visual localization from active to passive | |
| N Kamal et al. | License plate tilt correction: a review | |
| Han et al. | Dynahead-YOLO-Otsu: an efficient DCNN-based landslide semantic segmentation method using remote sensing images | |
| El Amrani Abouelassad et al. | Vehicle instance segmentation with rotated bounding boxes in uav images using cnn | |
| CN112288038B (en) | Object recognition or object registration method based on image classification and computing system | |
| Sun et al. | Research on Multi-target Recognition and Classification Strategy Based on Yolo v5 Framework |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| DD01 | Delivery of document by public notice |
Addressee: Wang Xiaodong Document name: the First Notification of an Office Action |
|
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20121031 |