CN108805896A - A kind of range image segmentation method applied to urban environment - Google Patents

A kind of range image segmentation method applied to urban environment Download PDF

Info

Publication number
CN108805896A
CN108805896A CN201810424191.2A CN201810424191A CN108805896A CN 108805896 A CN108805896 A CN 108805896A CN 201810424191 A CN201810424191 A CN 201810424191A CN 108805896 A CN108805896 A CN 108805896A
Authority
CN
China
Prior art keywords
pixel
range image
row
center pixel
distance
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.)
Granted
Application number
CN201810424191.2A
Other languages
Chinese (zh)
Other versions
CN108805896B (en
Inventor
袁夏
毛阳宇坤
赵春霞
陆建峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Science and Technology
Original Assignee
Nanjing University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN201810424191.2A priority Critical patent/CN108805896B/en
Publication of CN108805896A publication Critical patent/CN108805896A/en
Application granted granted Critical
Publication of CN108805896B publication Critical patent/CN108805896B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30181—Earth observation
    • G06T2207/30184—Infrastructure

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of range image segmentation methods applied to urban environment.This method is：Adjusting the distance first in image, often the different size of distance value of row pixel carries out quantity statistics, generates histogram；Then fitting a straight line is carried out using random sampling unification algorism according to histogram, obtain the corresponding distance value of road sections in every row, pixel in range image often row pixel in this distance value and its top and the bottom section is split, the segmentation result of road sections is obtained；Finally centered on the pixel of first, the image upper left corner pixel queue initial center pixel, carry out 8- neighborhoods breadth first traversal, residual pixel part is split.The present invention improves the accuracy of whole segmentation efficiency and segmentation, to improve the accuracy of subsequent processing.

Description

A kind of range image segmentation method applied to urban environment
Technical field
The present invention relates to technical field of image segmentation, especially a kind of range image segmentation side applied to urban environment Method.
Background technology
With the progress of sensor technology, the acquiring way of range image is more and more, how to understand with range image table The three-dimensional scenic shown is the fields major issues to be solved such as vehicle navigation, accident processing.Three-dimensional point is utilized merely The method complexity height of cloud progress environment understanding, time efficiency are low, the method for using range image to carry out urban environment segmentation, energy Enough is fast and effeciently various objects by scene cut, and target detection for next step and tracking provide pretreatment well and tie Fruit.It is to use more modes that road sections and rest part, which are separately handled, is suitable for several scenes.
At present be commonly used to segmentation range image in road sections mode be calculate the normal vector of each pixel corresponding points, or Range image is projected into three-dimensional point cloud to handle, such mode is difficult to reach the requirement of unmanned vehicle real-time processing, and Segmentation to non-rice habitats part, But most of algorithms are too strong to the sensibility of threshold value.
Invention content
The purpose of the present invention is to provide a kind of range image segmentation methods applied to urban environment, to preferably will Range image segmentation is blocking.
Realize that the technical solution of the object of the invention is：A kind of range image segmentation method applied to urban environment, Include the following steps：
Step 1, the range image P for setting urban environment are made of h rows, w row pixels, and the value of each pixel corresponds to distance Value, distance value are up to d；Quantity of the distance value of often row pixel on each section is counted, obtains statistic histogram P1；
Step 2, using random sampling unification algorism in statistic histogram P1Middle fitting a straight line y=kx+b；
Step 3 takes top and the bottom distance value section according to a beeline y=kx+b, and range image P is corresponded to and meets this area in row Between distance value pixel be labeled as road sections；
Step 4, the pixel centered on the 1st row the 1st row of range image P, it is excellent that image of adjusting the distance carries out 8- neighborhood ranges It first traverses, ergodic condition is：Neighborhood pixels are judged as connected region with center pixel；The sequence direction of traversal be from a left side to Right, from top to bottom, until h rows w row, traversal terminates to complete the segmentation of range image.
Further, quantity of the distance value for counting often row pixel on each section described in step 1, obtains statistics Histogram P1, specific as follows：
Step 1-1, initial setting P1The image being made of h rows, w_1 row pixels for one, the numerical values recited that each column indicates It is 1, the number of scanning lines of setpoint distance image P is i=1, columns j；
Step 1-2, by the distance value round in the i-th rows of range image P, the number of each distance value is counted, By this value filling histogram P1In the i-th row jth arrange, wherein
Step 1-3, number of scanning lines i=i+1 repeats step 1-2, until i>Until h, complete statistic histogram is obtained P1。
Further, utilize random sampling unification algorism in statistic histogram P described in step 21Middle fitting a straight line y= Kx+b, it is specific as follows：
Step 2-1, by statistic histogram P1In it is all be less than threshold value T1Element set to 0, remaining element is set to 1, and by P1In The line information for the element that all values are 1 counts on and is used as coordinate information in set Q；
Step 2-2, linear equation y=kx+b is set, take different two coordinate (y in set Q at random1,x1),(y2, x2), calculate the straight line parameter that two coordinates are linked to be：
bi=y1-kx1
Step 2-3, setting counts tiEach coordinate (y in=0, set of computations Qi,xi) arrive straight line y=kix+biDistance di：
IfThen count tiIt is constant；IfThen count ti=ti+1；
Step 2-4, step 2-2 and 2-3 are repeated 50~5000 times, terminates to compare counting t every timeiIf ti≥ti+1, then protect Stay former straight line y=kix+bi；If ti<ti+1, then retain this straight line y=ki+1x+bi+1, until repeating to terminate, obtain best straight line y =kx+b.
Further, top and the bottom distance value section is taken according to a beeline y=kx+b described in step 3, by range image P The distance value pixel for meeting this section in corresponding row is labeled as road sections, specific as follows：
Step 3-1, in statistic histogram P1In, the bound formula of best straight line y=kx+b is：
Step 3-2, initial setting up line number i=1 substitutes into i as y in bound formula, obtains road sections in the row Corresponding distance value rangePixel in the i-th rows of artwork P within the scope of respective distances is labeled as road portion Point；
Step 3-3, line number i=i+1 is set, step 3-2 is repeated, until i>H, whole image road sections label terminate.
Further, the 8- neighborhood breadth first traversals described in step 4, it is specific as follows：
Step 4-1, scanning element position coordinates are set as m rows, the n-th row, are denoted as (m, n), m=1, n=under original state 1, setting center pixel queue q；
Step 4-2, center pixel queue q is added in (m, n) and marked, respectively to the neighborhood pixels of this center pixel, i.e., (m-1, n-1), (m-1, n), (m-1, n+1), (m, n-1), (m, n+1), (m+1, n-1), (m+1, n), (m+1, n+1) 8 are not The neighborhood pixels of label carry out connected domain judgement and are jumped if center pixel (m, n) is marked in image border or neighborhood pixels Cross determination step；If neighborhood pixels are judged as being connected to center pixel (m, n), center pixel is added in this adjacent pixels Queue, and it is labeled as center pixel；
Step 4-3, after 8 neighborhood pixels of center pixel (m, n) judge, this center pixel (m, n) is removed Center pixel queue q；If center pixel queue q is not sky, centered on next pixel in center pixel queue q Pixel, then repeatedly step 4-2；If center pixel queue q is sky, label is replaced, and by (m+w, n+z) as in newly Imago element, wherein w are the distance of (m, n) to the region right side boundary of connection, and z is boundary on the downside of the region of (m, n) distance connection Distance, then repeatedly step 4-2；Until each pixel of image is scanned, traversal terminates.
Further, the connected domain judgement described in step 4-2, it is specific as follows：
Step 4-2-1, setting A and B is the central pixel point and its neighbor pixel in range image P, O respectively To generate the point of observation position of this range image P, A, B, O are located in the same rectangular coordinate system, and OA and OB respectively represent observation O'clock to 2 points of straight line, then the length d of OA and OB1And d2The distance value for representing two pixels, the line AB distances between A and B For d3；
Step 4-2-2, select point B ' in the positive direction on straight line OA from O to A as reference, | | OB ' | |=| | OB | |, By B and B ' line BB ' length d4With d3It is compared, calculates d3With d4Ratio r：
Wherein, θuIt is the angle ∠ AOB of OA and OB, is fixed value, depends on the hardware of acquisition range image；
Step 4-2-3, given threshold r0And 1≤r0≤ 5, if r≤r0, then judge central pixel point A and neighbor pixel B can be connected to, and can form connected region.
Compared with prior art, the present invention its remarkable advantage is：(1) it when dividing the road sections of range image, utilizes The statistics with histogram algorithm consistent with random sampling, keeps calculating simpler, shortens and calculates the time；(2) in segmentation distance map When the non-rice habitats part of picture, the mode that is traversed using 8- neighborhoods, it is ensured that the region that can be connected to is connected to completely, is avoided in details Missing；(3) geometry length of side ratio is used to reduce the sensibility to threshold value as the decision condition of connection in ergodic condition, It can operate with Various Complex scene.
The present invention will be further described below in conjunction with the accompanying drawings.
Description of the drawings
Fig. 1 is the flow chart for the range image segmentation method that the present invention is applied to urban environment.
Fig. 2 is the schematic diagram of initial distance image in the embodiment of the present invention.
Fig. 3 is the schematic diagram of image after dividing in the embodiment of the present invention.
Specific implementation mode
The present invention is applied to the range image segmentation method of urban environment, includes the following steps：
Step 1, the range image P for setting urban environment are made of h rows, w row pixels, and the value of each pixel corresponds to distance Value, distance value are up to d；Quantity of the distance value of often row pixel on each section is counted, obtains statistic histogram P1, specifically It is as follows：
Step 1-1, initial setting P1The image being made of h rows, w_1 row pixels for one, w_1=100, what each column indicated Numerical values recited is 1, and the number of scanning lines of setpoint distance image P is i=1, columns j；
Step 1-2, by the distance value round in the i-th rows of range image P, the number of each distance value is counted, By this value filling histogram P1In the i-th row jth arrange, wherein
Step 1-3, number of scanning lines i=i+1 repeats step 1-2, until i>Until h, complete statistic histogram is obtained P1。
Step 2, using random sampling unification algorism in statistic histogram P1Middle fitting a straight line y=kx+b, it is specific as follows：
Step 2-1, by statistic histogram P1In all 200 elements of being less than set to 0, remaining element is set to 1, and by P1Middle institute It is the line information of 1 element to have value, counts on and is used as coordinate information in set Q；
Step 2-2, linear equation y=kx+b is set, take different two coordinate (y in set Q at random1,x1),(y2, x2), calculate the straight line parameter that two coordinates are linked to be：
Step 2-3, design number tiEach coordinate (y in=0, set of computations Qi,xi) arrive straight line y=kix+biDistance：
IfThen count tiIt is constant；IfThen count ti=ti+1；
Step 2-4, step 2-2 and 2-3 are repeated 500 times, terminates to compare counting t every timeiIf ti≥ti+1, then retain original Straight line y=kix+bi；If ti<ti+1, then retain this straight line y=ki+1x+bi+1, until 500 repetitions terminate, obtain best straight line y =kx+b.
Step 3 takes its top and the bottom distance value section according to a beeline y=kx+b, and former range image P is corresponded to the symbol in row The distance value pixel for closing this section is labeled as road sections, specific as follows：
Step 3-1, in statistic histogram P1In, the bound formula of best straight line y=kx+b is：
Step 3-2, line number i=1 initially is set, is substituted into i as y in bound formula, obtains road sections pair in the row The distance value range answeredPixel in the i-th rows of artwork P within the scope of respective distances is labeled as road sections；
Step 3-3, line number i=i+1 is set, step 3-2 is repeated, until i>H, whole image road sections label terminate.
Step 4, the pixel p centered on the 1st row the 1st row of range image P1, image of adjusting the distance progress 8- neighborhood ranges First traversal, ergodic condition are：Neighborhood pixels are judged as connected region with center pixel；The sequence direction of traversal be from a left side to It is right, from top to bottom, until h rows w row, traversal terminate.It is specific as follows：
Step 4-1, scanning element position coordinates are set as m rows, the n-th row, are denoted as (m, n), m=1, n=under original state 1, setting center pixel queue q；
Step 4-2, center pixel queue q is added in (m, n) and marked, respectively to the neighborhood pixels of this center pixel, i.e., (m-1, n-1), (m-1, n), (m-1, n+1), (m, n-1), (m, n+1), (m+1, n-1), (m+1, n), (m+1, n+1) 8 are not The neighborhood pixels of label carry out connected domain judgement and are jumped if center pixel (m, n) is marked in image border or neighborhood pixels Cross determination step；If neighborhood pixels are judged as being connected to center pixel (m, n), center pixel is added in this adjacent pixels Queue, and it is labeled as center pixel；Judge that connected region step is specific as follows：
Step 4-2-1, setting A and B is the central pixel point and its neighbor pixel in range image P, O respectively To generate the point of observation position of this range image P, A, B, O are located in the same rectangular coordinate system, and OA and OB respectively represent observation O'clock to 2 points of straight line, then the length d of OA and OB1And d2The distance value for representing two pixels, the line AB distances between A and B For d3；
Step 4-2-2, select point B ' in the positive direction on straight line OA from O to A as reference, | | OB ' | |=| | OB | |, By B and B ' line BB ' length d4With d3It is compared, calculates d3With d4Ratio r：
Wherein, θuIt is the angle ∠ AOB of OA and OB, is fixed value, depends on the hardware of acquisition range image；
Step 4-2-3, given threshold r0=2.8, if r≤r0, then judge that central pixel point A can connect with neighbor pixel B It is logical, connected region can be formed.
Step 4-3, after 8 neighborhood pixels of center pixel (m, n) judge, this center pixel (m, n) is removed Center pixel queue q；If center pixel queue q is not sky, by picture centered on next pixel of center pixel queue q Element, then repeatedly step 4-2；If center pixel queue q is sky, label is replaced, and by (m+w, n+z) as new center Pixel, wherein w are the distance of (m, n) to the region right side boundary of connection, and z is boundary on the downside of the region of (m, n) distance connection Distance, then repeatedly step 4-2；Until each pixel of image is scanned, traversal terminates.
With reference to specific embodiment, the present invention will be further described.
Embodiment 1
In conjunction with Fig. 1, the present invention is applied to the range image segmentation method of urban environment, includes the following steps：
Step 1, a range image P for describing urban environment, are made of, the value pair of each pixel 32 rows, 600 row pixels Distance value is should be, distance value is up to 70；Quantity of the distance value of often row pixel on each section is counted, obtains statistics histogram Scheme P1；Specifically include following steps：
Step 1-1, P is initially set1The image being made of 32 row w_1 row pixels for one, w_1=100, the number that each column indicates It is 1 to be worth size, and the number of scanning lines of setpoint distance image P is i=1, columns j；
Step 1-2, by the distance value round in the i-th rows of range image P, the number of each distance value is counted, By this value filling histogram P1In the i-th row jth arrange, wherein
Step 1-3, number of scanning lines i=i+1 repeats step 1-2, until i>Until 32, complete statistic histogram is obtained P1。
Step 2, using random sampling unification algorism in statistic histogram P1Middle fitting a straight line y=kx+b；It is specific as follows：
Step 2-1, by statistic histogram P1In all 200 elements of being less than set to 0, remaining element is set to 1, and by P1Middle institute It is the line information of 1 element to have value, counts on and is used as coordinate information in set Q；
Step 2-2, linear equation y=kx+b is set, take different two coordinate (y in set Q at random1,x1),(y2, x2), calculate the straight line parameter that two coordinates are linked to be：
Step 2-3, design number tiEach coordinate (y in=0, set of computations Qi,xi) arrive straight line y=kix+biDistance, i.e.,：
IfThen count tiIt is constant；IfThen count ti=ti+1；
Step 2-4, step 2-2 and 2-3 are repeated 500 times, terminates to compare counting t every timeiIf ti≥ti+1, then retain original Straight line y=kix+bi；If ti<ti+1, then retain this straight line y=ki+1x+bi+1, until 500 repetitions terminate, obtain best straight Line y=kx+b.
Step 3 takes its top and the bottom distance value section according to a beeline y=kx+b, and former range image P is corresponded to the symbol in row The distance value pixel for closing this section is labeled as road sections；It is specific as follows：
Step 3-1, in statistic histogram P1In, the bound formula of best straight line y=kx+b is：
Step 3-2, line number i=1 initially is set, is substituted into i as y in bound formula, obtains road sections pair in the row The distance value range answeredPixel in the i-th rows of artwork P within the scope of respective distances is labeled as road portion Point；
Step 3-3, line number i=i+1 repeats step 3-2, until i>H, whole image road sections label terminate.
Step 4, the pixel p centered on the 1st row the 1st row of range image P1, image of adjusting the distance progress 8- neighborhood ranges First traversal, ergodic condition are：Neighborhood pixels are judged as connected region with center pixel；The sequence direction of traversal be from a left side to It is right, from top to bottom, until h rows w row, traversal terminate.It is specific as follows：
Step 4-1, scanning element position coordinates are set as m rows, the n-th row, are denoted as (m, n), m=1, n=under original state 1, setting center pixel queue q；
Step 4-2, center pixel queue q is added in (m, n) and marked, respectively to the neighborhood pixels of this center pixel, i.e., (m-1, n-1), (m-1, n), (m-1, n+1), (m, n-1), (m, n+1), (m+1, n-1), (m+1, n), (m+1, n+1) 8 are not The neighborhood pixels of label carry out connected domain judgement and are jumped if center pixel (m, n) is marked in image border or neighborhood pixels Cross determination step；If neighborhood pixels are judged as being connected to center pixel (m, n), center pixel is added in this adjacent pixels Queue, and it is labeled as center pixel；The step of judging connected region is as follows：
Step 4-2-1, setting A and B is the central pixel point and its neighbor pixel in range image, O respectively To generate the point of observation position of this range image, A, B, O are located in the same rectangular coordinate system, and OA and OB respectively represent point of observation To 2 points of straight line, then the length d of OA and OB1And d2The distance value for representing two pixels, the line AB distances between A and B are d3；
Step 4-2-2, select point B ' in the positive direction on straight line OA from O to A as reference, | | OB ' | |=| | OB | |, By B and B ' line BB ' length d4With d3It is compared, calculates d3With d4Ratio r：
Wherein, θuIt is the angle ∠ AOB of OA and OB, is fixed value, depends on the hardware of acquisition range image；
Step 4-2-3, given threshold r0=2.8, if r≤r0, then judge that central pixel point A can connect with neighbor pixel B It is logical, connected region can be formed.
Step 4-3, after 8 neighborhood pixels of center pixel (m, n) judge, this center pixel (m, n) is removed Center pixel queue q；If center pixel queue q is not sky, by picture centered on next pixel of center pixel queue q Element, then repeatedly step 4-2；If center pixel queue q is sky, label is replaced, and by (m+w, n+z) as new center Pixel, wherein w are the distance of (m, n) to the region right side boundary of connection, and z is boundary on the downside of the region of (m, n) distance connection Distance, then repeatedly step 4-2；Until each pixel of image is scanned, traversal terminates.
In conjunction with Fig. 2, the gray scale of each pixel indicates its distance in range image, and the smaller expression distance of gray scale is bigger, can To see, gray scale tapers into the pixel of road sections from the near to the remote, if whole be split, can not be divided into road sections The relationship of adjacent pixel preferably judges in object of the same race in a whole part rather than road sections.
In conjunction with Fig. 3, using urban environment dividing method by the result after range image segmentation, it can be seen that road plane portion Divide and be clearly labeled as black region, rest part is also respectively labeled as various different gray scales, segmentation result clear and definite.
The present invention is when dividing the road sections of range image, using the statistics with histogram method consistent with random sampling, Keep calculating simpler, the calculating time greatly shortens；When dividing the non-rice habitats part of range image, traversed using 8- neighborhoods Mode, it is ensured that the region that can be connected to is connected to completely, avoids the missing in details；Geometry length of side ratio is used in ergodic condition As the decision condition of connection, the sensibility to threshold value is reduced, can operate with Various Complex scene.

Claims (6)

1. a kind of range image segmentation method applied to urban environment, which is characterized in that include the following steps：
Step 1, the range image P for setting urban environment are made of h rows, w row pixels, and the value of each pixel corresponds to distance value, Distance value is up to d；Quantity of the distance value of often row pixel on each section is counted, obtains statistic histogram P1；
Step 2, using random sampling unification algorism in statistic histogram P1Middle fitting a straight line y=kx+b；
Step 3 takes top and the bottom distance value section according to a beeline y=kx+b, and range image P is corresponded to and meets this section in row Distance value pixel is labeled as road sections；
Step 4, the pixel centered on the 1st row the 1st row of range image P, image of adjusting the distance carry out 8- neighborhood breadth Firsts time It goes through, ergodic condition is：Neighborhood pixels are judged as connected region with center pixel；The sequence direction of traversal be from left to right, from Under, until h rows w row, traversal terminates to complete the segmentation of range image.
2. the range image segmentation method according to claim 1 applied to urban environment, which is characterized in that in step 1 Quantity of the distance value for counting often row pixel on each section, obtains statistic histogram P1, specific as follows：
Step 1-1, initial setting P1The image being made of h rows, w_1 row pixels for one, the numerical values recited that each column indicates are 1, if The number of scanning lines of set a distance image P is i=1, columns j；
Step 1-2, by the distance value round in the i-th rows of range image P, the number of each distance value is counted, by this Value filling histogram P1In the i-th row jth arrange, wherein
Step 1-3, number of scanning lines i=i+1 repeats step 1-2, until i>Until h, complete statistic histogram P is obtained1。
3. the range image segmentation method according to claim 1 or 2 applied to urban environment, which is characterized in that
Utilize random sampling unification algorism in statistic histogram P described in step 21Middle fitting a straight line y=kx+b, it is specific as follows：
Step 2-1, by statistic histogram P1In it is all be less than threshold value T1Element set to 0, remaining element is set to 1, and by P1In own The line information for the element that value is 1 counts on and is used as coordinate information in set Q；
Step 2-2, linear equation y=kx+b is set, take different two coordinate (y in set Q at random1,x1),(y2,x2), meter Calculate the straight line parameter that two coordinates are linked to be：
bi=y1-kx1
Step 2-3, setting counts tiEach coordinate (y in=0, set of computations Qi,xi) arrive straight line y=kix+biDistance di：
IfThen count tiIt is constant；IfThen count ti=ti+1；
Step 2-4, step 2-2 and 2-3 are repeated 50~5000 times, terminates to compare counting t every timeiIf ti≥ti+1, then retain original Straight line y=kix+bi；If ti<ti+1, then retain this straight line y=ki+1x+bi+1, until repeating to terminate, obtain best straight line y=kx+ b。
4. the range image segmentation method according to claim 3 applied to urban environment, which is characterized in that in step 3 Described takes top and the bottom distance value section according to a beeline y=kx+b, and range image P is corresponded to the distance for meeting this section in row It is worth pixel and is labeled as road sections, it is specific as follows：
Step 3-1, in statistic histogram P1In, the bound formula of best straight line y=kx+b is：
Step 3-2, initial setting up line number i=1 is substituted into i as y in bound formula, is obtained road sections in the row and is corresponded to Distance value rangePixel in the i-th rows of artwork P within the scope of respective distances is labeled as road sections；
Step 3-3, line number i=i+1 is set, step 3-2 is repeated, until i>H, whole image road sections label terminate.
5. the range image segmentation method according to claim 1 or 4 applied to urban environment, which is characterized in that step 4 Described in 8- neighborhood breadth first traversals, it is specific as follows：
Step 4-1, scanning element position coordinates are set as m rows, the n-th row, are denoted as (m, n), m=1, n=1 under original state, if Center pixel queue q；
Step 4-2, center pixel queue q is added in (m, n) and marked, respectively to the neighborhood pixels of this center pixel, i.e., (m-1, N-1), (m-1, n), (m-1, n+1), (m, n-1), (m, n+1), (m+1, n-1), (m+1, n), (m+1, n+1) 8 are unlabelled Neighborhood pixels carry out connected domain judgement and skip judgement if center pixel (m, n) is marked in image border or neighborhood pixels Step；If neighborhood pixels are judged as being connected to center pixel (m, n), center pixel queue is added in this adjacent pixels, And it is labeled as center pixel；
Step 4-3, after 8 neighborhood pixels of center pixel (m, n) judge, this center pixel (m, n) is removed into center Pixel queue q；If center pixel queue q is not sky, by picture centered on next pixel in center pixel queue q Element, then repeatedly step 4-2；If center pixel queue q is sky, label is replaced, and by (m+w, n+z) as new center Pixel, wherein w are the distance of (m, n) to the region right side boundary of connection, and z is boundary on the downside of the region of (m, n) distance connection Distance, then repeatedly step 4-2；Until each pixel of image is scanned, traversal terminates.
6. the range image segmentation method according to claim 5 applied to urban environment, which is characterized in that step 4-2 Described in connected domain judgement, it is specific as follows：
Step 4-2-1, setting A and B is that central pixel point in range image P and its neighbor pixel, O make a living respectively At the point of observation position of this range image P, A, B, O are located in the same rectangular coordinate system, and OA and OB respectively represent point of observation and arrive 2 points of straight line, then the length d of OA and OB1And d2The distance value for representing two pixels, the line AB distances between A and B are d3；
Step 4-2-2, select point B ' in the positive direction on straight line OA from O to A as reference, | | OB ' | |=| | OB | |, by B With the length d of the line BB ' of B '4With d3It is compared, calculates d3With d4Ratio r：
Wherein, θuIt is the angle ∠ AOB of OA and OB, is fixed value, depends on the hardware of acquisition range image；
Step 4-2-3, given threshold r0And 1≤r0≤ 5, if r≤r0, then judge that central pixel point A and neighbor pixel B can Connection, can form connected region.
CN201810424191.2A 2018-05-07 2018-05-07 Distance image segmentation method applied to urban environment Active CN108805896B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810424191.2A CN108805896B (en) 2018-05-07 2018-05-07 Distance image segmentation method applied to urban environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810424191.2A CN108805896B (en) 2018-05-07 2018-05-07 Distance image segmentation method applied to urban environment

Publications (2)

Publication Number Publication Date
CN108805896A true CN108805896A (en) 2018-11-13
CN108805896B CN108805896B (en) 2022-02-18

Family

ID=64091006

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810424191.2A Active CN108805896B (en) 2018-05-07 2018-05-07 Distance image segmentation method applied to urban environment

Country Status (1)

Country Link
CN (1) CN108805896B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110008837A (en) * 2019-03-06 2019-07-12 东南大学 A kind of Fast implementation of the statistics of the road network closed polygon according to belonging to POI
CN112766338A (en) * 2021-01-11 2021-05-07 明峰医疗系统股份有限公司 Method, system and computer readable storage medium for calculating distance image

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105137412A (en) * 2015-08-19 2015-12-09 重庆大学 Accurate fitting method of line segment features in 2D laser radar distance image
CN105989334A (en) * 2015-02-12 2016-10-05 中国科学院西安光学精密机械研究所 Monocular vision-based road detection method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105989334A (en) * 2015-02-12 2016-10-05 中国科学院西安光学精密机械研究所 Monocular vision-based road detection method
CN105137412A (en) * 2015-08-19 2015-12-09 重庆大学 Accurate fitting method of line segment features in 2D laser radar distance image

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王钤等: "基于激光雷达距离图像的道路边缘检测与跟踪", 《计算机科学》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110008837A (en) * 2019-03-06 2019-07-12 东南大学 A kind of Fast implementation of the statistics of the road network closed polygon according to belonging to POI
CN110008837B (en) * 2019-03-06 2022-04-29 东南大学 Quick realization method for calculating POI according to closed polygon of road network
CN112766338A (en) * 2021-01-11 2021-05-07 明峰医疗系统股份有限公司 Method, system and computer readable storage medium for calculating distance image
CN112766338B (en) * 2021-01-11 2023-06-16 明峰医疗系统股份有限公司 Method, system and computer readable storage medium for calculating distance image

Also Published As

Publication number Publication date
CN108805896B (en) 2022-02-18

Similar Documents

Publication Publication Date Title
CN106780612B (en) Object detecting method and device in a kind of image
CN104282020B (en) A kind of vehicle speed detection method based on target trajectory
CN107248159A (en) A kind of metal works defect inspection method based on binocular vision
CN109903331B (en) Convolutional neural network target detection method based on RGB-D camera
CN110807496B (en) Dense target detection method
CN110084241B (en) Automatic ammeter reading method based on image recognition
CN110096920B (en) High-precision high-speed positioning label and positioning method for visual servo
CN109712153A (en) A kind of remote sensing images city superpixel segmentation method
CN107038416B (en) Pedestrian detection method based on binary image improved HOG characteristics
CN110349260B (en) Automatic pavement marking extraction method and device
CN107392929B (en) Intelligent target detection and size measurement method based on human eye vision model
CN105335952B (en) Matching power flow computational methods and device and parallax value calculating method and equipment
CN109615653A (en) Percolating water area detecting and recognition methods based on deep learning and visual field projection model
Lu et al. A new region growing-based method for road network extraction and its application on different resolution SAR images
CN102646343A (en) Vehicle detection apparatus
CN103295239A (en) Laser-point cloud data automatic registration method based on plane base images
CN109829476B (en) End-to-end three-dimensional object detection method based on YOLO
CN112766184B (en) Remote sensing target detection method based on multi-level feature selection convolutional neural network
CN110414385A (en) A kind of method for detecting lane lines and system based on homography conversion and characteristic window
CN106096497B (en) A kind of house vectorization method for polynary remotely-sensed data
CN110427933A (en) A kind of water gauge recognition methods based on deep learning
CN104966348B (en) A kind of bill images key element integrality detection method and system
CN109448127A (en) A kind of farmland high-precision navigation map generation method based on unmanned aerial vehicle remote sensing
CN105405138A (en) Water surface target tracking method based on saliency detection
CN106169173A (en) A kind of image interpolation method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant