CN113516759A - Automatic modeling method of vehicle body side view parameterized model - Google Patents

Automatic modeling method of vehicle body side view parameterized model Download PDF

Info

Publication number
CN113516759A
CN113516759A CN202110843172.5A CN202110843172A CN113516759A CN 113516759 A CN113516759 A CN 113516759A CN 202110843172 A CN202110843172 A CN 202110843172A CN 113516759 A CN113516759 A CN 113516759A
Authority
CN
China
Prior art keywords
vehicle body
model
points
boundary
view
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
CN202110843172.5A
Other languages
Chinese (zh)
Other versions
CN113516759B (en
Inventor
王博
杨治赢
李宝军
胡灿
刘元斗
陈满
李勇
陈燕玲
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.)
Wuhan University of Science and Engineering WUSE
Original Assignee
Wuhan University of Science and Engineering WUSE
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 Wuhan University of Science and Engineering WUSE filed Critical Wuhan University of Science and Engineering WUSE
Priority to CN202110843172.5A priority Critical patent/CN113516759B/en
Publication of CN113516759A publication Critical patent/CN113516759A/en
Application granted granted Critical
Publication of CN113516759B publication Critical patent/CN113516759B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
    • 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
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G06T2207/10012—Stereo images

Abstract

The invention relates to an automatic modeling method of a vehicle body side view parameterized model. Firstly, selecting key points of a side view of a vehicle body, extracting boundary information of the side view by using an improved Canny operator and a watershed algorithm, clearly defining a noise edge, optimally extracting the boundary information by using a middle boundary point generation algorithm, solving a functional relation between key points of the vehicle body and a template database by using an improved reconstruction algorithm Car-OMP after calibrating a template database to finish rough extraction of the vehicle body model, introducing boundary constraint conditions, obtaining optimal control points by an optimization iteration method, and finishing accurate extraction of the model. The method provided by the invention is not only suitable for rendering the graph, but also can realize accurate model reconstruction on the dimension graph and the hand-drawn graph.

Description

Automatic modeling method of vehicle body side view parameterized model
Technical Field
The invention belongs to the technical field of vehicle body reconstruction, and particularly relates to an automatic modeling method of a vehicle body side view parameterized model.
Background
In the present society, the three-dimensional reconstruction of objects from images is a hot topic. The extraction of the two-dimensional parameterized model is not necessarily required to reconstruct the three-dimensional model of the vehicle body by using the vehicle body image. Whether the three-dimensional model is accurate or not is directly determined by the extraction precision of the two-dimensional model, and the extracted two-dimensional parameterized model of the vehicle body can be arranged into a database, so that the reuse rate of pictures and models is improved, and data support is provided for three-dimensional reconstruction of the vehicle body.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides an automatic modeling method of a vehicle body side view parameterized model. Firstly, selecting key points of a side view of a vehicle body, extracting boundary information of the side view by using an improved Canny operator and a watershed algorithm, clearly defining a noise edge, optimally extracting the boundary information by using a middle boundary point generation algorithm, solving a functional relation between key points of the vehicle body and a template database by using an improved reconstruction algorithm Car-OMP after calibrating a template database to finish rough extraction of the vehicle body model, introducing boundary constraint conditions, obtaining optimal control points by an optimization iteration method, and finishing accurate extraction of the model. The method provided by the invention is not only suitable for rendering the graph, but also can realize accurate model reconstruction on the dimension graph and the hand-drawn graph.
In order to achieve the aim, the technical scheme provided by the invention is an automatic modeling method of a vehicle body side view parameterized model, which comprises the following steps of:
step 1, selecting key points of a side view of a vehicle body;
step 2, extracting a valley line by using an improved Canny operator, and performing image segmentation by using the valley line to replace water injection discrete points in a watershed algorithm, so as to extract boundary information of a side view of the vehicle body;
step 3, deleting the noise boundary, and optimizing the boundary extracted in the step 2;
step 3.1, defining a noise boundary;
step 3.2, optimizing a boundary extraction result by using an intermediate boundary point generation algorithm;
and 4, extracting the vehicle body model, which comprises the following steps:
step 4.1, calibrating a template database;
step 4.2, performing rough extraction on the vehicle body model based on the template database and the key points selected in the step 1;
step 4.2.1, defining a vehicle body side view key point vector ZmIndividual body model shape vector siA template database matrix S and a template database key point sub-matrix Z.
Step 4.2.2, establishing a vehicle body side view key point vector ZmFunctional relation Z with template database key point submatrix ZmAnd solving the model contribution quantity by using an improved reconstruction algorithm Car-OMP, wherein the model contribution quantity is Z multiplied by theta and theta is used as the model contribution quantity.
Step 4.2.3, obtaining the shape vector S of the vehicle body model by using the model contribution amount obtained in the step 4.2.2 and the model library matrix SiThis allows the rough extraction of the body model to be completed, resulting in an initial model.
And 4.3, accurately extracting the vehicle body model by combining the boundary information in the step 3.
And in the step 1, the key points are selected by referring to the arrangement scheme of the face key points and the method of the related thesis, the total number of the key points is 25, the key points comprise the vehicle body outer contour reference points, the vehicle door and vehicle window contour characteristic points and the front and rear end points of the waist line skirt line, and by selecting reasonable key points, the vehicle body information can be completely expressed, the calculated amount can be simplified, the calculation efficiency can be improved, and the subsequent point set matching with the template is facilitated.
And 2, improving the Canny algorithm, selecting a part with small image gray intensity change to obtain low-value pixel points, namely valley points, labeling the region of 8 pixels around the valley points by using a bwleabel function in Matlab, deleting the valley points smaller than a given threshold label to obtain a valley line extraction result, and performing image segmentation on the side view image of the vehicle body by using the valley lines to replace water injection discrete points in the watershed algorithm, so that the influence of a detailed structure is eliminated to a certain extent under the condition of keeping the global structure characteristics.
Moreover, the noise edge defined in said step 3.1 includes two cases: based on Canny edge feature points, if the number of points on an edge far away from the Canny edge feature points by a given threshold lambda (lambda is 5) is more than half of all the points on the edge, the edge can be considered as a noise edge; the edge that intersects the image edge is also a noise edge.
And in the step 3.2, after the noise edges are deleted, cleaner main characteristic information of the side view of the vehicle body can be extracted, but the main boundary has a 'burr' phenomenon, and by utilizing the condition that the gray value of the middle point of the boundary is lower than the gray values of the boundary points at the two sides and is basically equal to the distance between the boundary points at the two sides, a middle boundary point generation (MEP) algorithm is designed, and the middle boundary point is calculated to obtain an ideal boundary result of the single pixel sequence.
And in the step 4.1, the wheels of the model are all formed by connecting 4 cubic Bezier curves end to end, each Bezier curve comprises 4 control points, an intersection point of connecting lines formed by two pairs of end control points is defined as a wheel center, and the wheel center of the front wheel is used as a coordinate origin to calibrate the template database, so that subsequent point set matching is facilitated. The template database is the scientific research result of the team and comprises templates of 215 types of vehicles in total, wherein the vehicles are different in brands and models.
Furthermore, step 4.2.1 is to spread the 25 keypoints of the side view into a vector Zm＝(x1，y1，x2，y2...x25，y25) T, wherein (x)i，yi) Coordinates of key points on the side view; defining a single body model as a shape vector si＝(x′1，y′1，x′2，y′2...x′n，y′n)TWherein (x'i，y′i) The coordinates of the control points on a single vehicle body model are represented by n, which is 97 multiplied by 4, that is, each vehicle type consists of 97 Bezier curves, and each Bezier curve contains 4 control points; the entire template database can be converted into a matrix S, which is equivalent to a matrix consisting of m shape vectors SiThe method comprises the following steps of (1) forming, wherein m is the number of vehicle body models in a template database; and extracting a submatrix Z with each vehicle body model only containing 25 vehicle body key points from the S.
Furthermore, said step 4.2.2 of calculating the model contribution using the Car-OMP reconstruction algorithm comprises the following steps:
step 4.2.2.1, initialize the parameter, i.e. residual r0＝ZmIndex set
Figure BDA0003179790660000031
Storing a matrix of column vectors corresponding to the index λ
Figure BDA0003179790660000032
The iteration time t is 1;
step 4.2.2.2, finding out residual error r and each column Z of the key point submatrix Z of the template databaseiThe index λ corresponding to the maximum inner product is:
λt＝argmaxi＝1...m|〈rt-1，Zi>| (1)
in the formula, ZiThe ith column of the template database key point submatrix Z is shown, m is the number of the automobile body models in the template database,<rt-1，Zi>representing the residual r and each column Z of the submatrixiInner product of, argmaxi＝1...m| · | denotes the index λ corresponding to the maximum absolute value of the inner productt；
Step 4.2.2.3, the step of the method,update index set At＝At-1∪{λtAnd updating a matrix for storing column vectors corresponding to the index lambda
Figure BDA0003179790660000033
Step 4.2.2.4, find Zm＝ZtθtThe least squares solution of (a), namely:
θt＝argmin||Zm-Ztθt||＝(Zt TZt)-1Zt TZm (2)
in the formula, argmaxi＝1...nThe corresponding model contribution theta when the maximum value is taken by the modelt；
Step 4.2.2.5, update residual rt，rtThe calculation method of (c) is as follows:
rt＝Zm-Ztθt＝Zm-Zt(Zt TZt)-1Zt TZm (3)
step 4.2.2.6, setting the total iteration times as K, and stopping iteration if the current iteration times t is more than K; otherwise, let t be t +1, perform step 4.2.2.1;
in step 4.2.2.7, the above loop is executed when K ≠ 0, and when K ≠ 0, θ ═ (Z) is employed-1×Zm。
And 4.3, calculating the distance from each point on the model curve to all pixel points on the nearest boundary and the index of the corresponding boundary, defining the two nearest points as a group, reducing the distance of each group of points to a given threshold value through iteration, considering that the model curve is consistent with the boundary extracted in the segmentation algorithm at the moment, simultaneously searching the optimal control point of the Bezier curve in the iteration process, and replacing the four control points on the original curve with the optimal control points to realize the accurate extraction of the vehicle body side view parameterized model.
Compared with the prior art, the invention has the following advantages:
(1) the traditional Canny operator and the watershed algorithm are combined and improved, a new boundary extraction algorithm is realized, and the influence of a detailed structure is eliminated to a certain extent under the condition of keeping the global structure characteristics.
(2) And (3) clearly defining the noise edge, and optimizing a boundary extraction result by using a middle boundary point generation algorithm.
(3) A novel method for extracting the car body model is provided, the rough extraction of the car body model is realized by establishing a functional relation between key points of the car body and a template database, and boundary information constraint conditions are introduced to further accurately extract the car body model.
(4) An Orthogonal Matching Pursuit (OMP) algorithm is introduced into the field of vehicle body modeling for the first time, and an improved reconstruction algorithm Car-OMP is used for solving the functional relation between key points of a vehicle body and a template database.
Drawings
FIG. 1 is a technical flow chart of an embodiment of the present invention.
FIG. 2 is a schematic diagram of the definition of key points of the side view of the vehicle body according to the embodiment of the invention.
FIG. 3 is a diagram illustrating a valley line extraction result of a side view of a vehicle body according to an embodiment of the present invention.
FIG. 4 is a side view image segmentation result diagram of a vehicle body according to an embodiment of the present invention.
FIG. 5 is a diagram of the boundary result of removing the noise edge according to the embodiment of the present invention.
FIG. 6 is a diagram of ideal single pixel sequence boundary results according to an embodiment of the present invention.
FIG. 7 is a diagram of the results of the crude template-based model extraction in the example of the present invention.
Detailed Description
The invention provides an automatic modeling method of a vehicle body side view parameterized model, which comprises the steps of firstly selecting key points of a vehicle body side view, then extracting boundary information of the side view by using an improved Canny operator and a watershed algorithm, clearly defining a noise edge, optimally extracting the boundary information by using a middle boundary point generation algorithm, after calibrating a template database, solving a functional relation between the key points of the vehicle body and the template database by using an improved reconstruction algorithm Car-OMP to finish the rough extraction of the vehicle body model, introducing boundary constraint conditions, obtaining optimal control points by using an optimization iteration method, and finishing the accurate extraction of the model.
The technical solution of the present invention is further explained with reference to the drawings and the embodiments.
As shown in fig. 1, the process of the embodiment of the present invention includes the following steps:
step 1, selecting key points of a side view of a vehicle body.
In order to facilitate the subsequent point set matching between the template and the human face, the arrangement scheme of the key points of the human face and the method of the related thesis are referred, 25 key points of the side view of the vehicle body are selected, and the number of the key points includes the outer contour reference point P of the vehicle body1-P14Vehicle window contour feature point P15-P22Door profile feature point P15、P18、P21、P23、P24、P25Front and rear end points P of skirt line23-P25As shown in fig. 2. By selecting reasonable key points, the vehicle body information can be completely expressed, the calculated amount can be simplified, and the calculation efficiency can be improved.
And 2, extracting a valley line by using an improved Canny operator, and performing image segmentation by using the valley line to replace water injection discrete points in a watershed algorithm, so as to extract boundary information of the side view of the vehicle body.
The Canny algorithm is improved, a part with small gray intensity change is selected to obtain low-value pixel points, namely valley points, the areas of 8 pixels around the valley points are labeled by utilizing a bwleabel function in Matlab, the valley points smaller than a given threshold label are deleted, and valley line extraction results are obtained, as shown in FIG. 3. The valley lines are used for replacing water injection discrete points in the watershed algorithm to perform image segmentation on the side view image of the car body, and the influence of a detailed structure is eliminated to a certain extent under the condition of keeping global structure characteristics, as shown in fig. 4.
And 3, deleting the noise boundary and optimizing the boundary extracted in the step 2.
And 3.1, defining a noise boundary.
The noise margin includes two cases: based on Canny edge feature points, if the number of points on an edge far away from the Canny edge feature points by a given threshold lambda (lambda is 5) is more than half of all the points on the edge, the edge can be considered as a noise edge; the edge that intersects the image edge is also a noise edge.
And 3.2, optimizing a boundary extraction result by using an intermediate boundary point generation algorithm.
The main characteristic information of the cleaner side view of the vehicle body can be extracted by deleting the noise edge, but the phenomenon of 'burr' appears on the main boundary at the moment, as shown in fig. 5. By using the condition that the gray value of the middle point of the boundary is lower than the gray values of the boundary points at the two sides and the distances between the middle point of the boundary and the boundary points at the two sides are basically equal, a middle boundary point generation (MEP) algorithm is designed, and the middle boundary point is calculated, so that an ideal single pixel sequence boundary result can be obtained, as shown in fig. 6.
And 4, extracting the vehicle body model, which comprises the following steps:
and 4.1, calibrating the template database.
The wheels of the model are formed by connecting 4 cubic Bezier curves end to end, each Bezier curve comprises 4 control points, an intersection point of connecting lines formed by two pairs of the control points end to end is defined as a wheel center, and the wheel center of the front wheel is used as a coordinate origin to calibrate the template database, so that subsequent point set matching is facilitated. The template database is the scientific research result of the team and comprises templates of 215 types of vehicles in total, wherein the vehicles are different in brands and models.
And 4.2, roughly extracting the vehicle body model based on the template database and the key points selected in the step 1.
Step 4.2.1, defining a vehicle body side view key point vector ZmIndividual body model shape vector siA template database matrix S and a template database key point sub-matrix Z.
Unfolding 25 keypoints of the side view into a vector Zm＝(x1，y1，x2，y2...x25，y25)TWherein (x)i，yi) Coordinates of key points on the side view; defining a single body model as a shape vector si＝(x′1，y′1，x′2，y′2...x′n，y′n)TWherein (x'i，y′i) The coordinates of the control points on a single vehicle body model are n ═ 97 × 4, that is, each vehicle type consists of 97 Bezier curves, and each B6zier curve contains 4 control points; the entire template database can be converted into a matrix S, which is equivalent to a matrix consisting of m shape vectors SiThe method comprises the following steps of (1) forming, wherein m is the number of vehicle body models in a template database; and extracting a submatrix Z with each vehicle body model only containing 25 vehicle body key points from the S.
Step 4.2.2, establishing a vehicle body side view key point vector ZmFunctional relation Z with template database key point submatrix ZmAnd solving the model contribution quantity by using an improved reconstruction algorithm Car-OMP, wherein the model contribution quantity is Z multiplied by theta and theta is used as the model contribution quantity.
The calculation of the model contribution amount by the Car-OMP reconstruction algorithm comprises the following steps:
step 4.2.2.1, initialize the parameter, i.e. residual r0＝ZmIndex set
Figure BDA0003179790660000061
Storing a matrix of column vectors corresponding to the index λ
Figure BDA0003179790660000062
The iteration time t is 1;
step 4.2.2.2, finding out residual error r and each column Z of the key point submatrix Z of the template databaseiThe index λ corresponding to the maximum inner product is:
λt＝argmaxi＝1...m|<rt-1，Zi>| (1)
in the formula, ZiThe ith column of the template database key point submatrix Z is shown, m is the number of the automobile body models in the template database,<rt-1，Zi>representing the residual r and each column Z of the submatrixiInner product of, argmaxi＝1...m| · | denotes the index λ corresponding to the maximum absolute value of the inner productt。
Step 4.2.2.3, update index set At＝At-1∪{λtAnd updating a matrix for storing column vectors corresponding to the index lambda
Figure BDA0003179790660000063
Step 4.2.2.4, find Zm＝ZtθtThe least squares solution of (a), namely:
θt＝argmin||Zm-Ztθt||＝(Zt TZt)-1Zt TZm (2)
in the formula, argmaxi＝1...nThe corresponding model contribution theta when the maximum value is taken by the modelt。
Step 4.2.2.5, update residual rt，rtThe calculation method of (c) is as follows:
rt＝Zm-Ztθt＝Zm-Zt(Zt TZt)-1Zt TZm (3)
step 4.2.2.6, setting the total iteration times as K, and stopping iteration if the current iteration times t is more than K; otherwise, let t be t +1, perform step 4.2.2.1;
in step 4.2.2.7, the above loop is executed when K ≠ 0, and when K ≠ 0, θ ═ (Z) is employed-1×Zm。
Step 4.2.3, obtaining the shape vector S of the vehicle body model by using the model contribution amount obtained in the step 4.2.2 and the model library matrix SiThis allows the rough extraction of the body model to be completed, resulting in an initial model.
And 4.3, accurately extracting the vehicle body model by combining the boundary information in the step 3.
And combining the boundary information and the initial model, and realizing automatic optimization of all cubic Bezier curves according to an optimal line matching algorithm to finish accurate extraction of the model. Specifically, the distance from each point on a model curve to all pixel points of the nearest boundary and the index of the corresponding boundary are calculated, the two nearest points are defined as one group, the distance between each group of points is reduced to a given threshold value through iteration, at the moment, the model curve is considered to be matched with the boundary extracted in the segmentation algorithm, meanwhile, in the iteration process, the optimal control point of a Bezier curve is searched, and the four control points on the original curve are replaced by the optimal control points, so that the accurate extraction of the vehicle body side view parameterized model is realized.
In specific implementation, the above process can adopt computer software technology to realize automatic operation process.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.

Claims (10)

1. An automatic modeling method for a side view parameterized model of a vehicle body is characterized by comprising the following steps:
step 1, selecting N key points of a side view of a vehicle body;
step 2, extracting a valley line by using an improved Canny operator, and performing image segmentation by using the valley line to replace water injection discrete points in a watershed algorithm, so as to extract boundary information of a side view of the vehicle body;
step 3, deleting the noise boundary, and optimizing the boundary extracted in the step 2;
and 4, extracting the vehicle body model, which comprises the following steps:
step 4.1, calibrating a template database;
step 4.2, performing rough extraction on the vehicle body model based on the template database and the key points selected in the step 1;
and 4.3, accurately extracting the vehicle body model by combining the boundary information in the step 3.
2. The method for automatically modeling the parametric model of the side view of the vehicle body as set forth in claim 1, wherein: and 2, improving a Canny algorithm, selecting a part with small image gray intensity change to obtain low-value pixel points, namely valley points, labeling the region of 8 pixels around the valley points by using a bwleabel function in Matlab, deleting the valley points smaller than a given threshold label to obtain a valley line extraction result, and performing image segmentation on the side view image of the vehicle body by using the valley lines to replace water injection discrete points in the watershed algorithm, so that the influence of a detailed structure is eliminated to a certain extent under the condition of keeping the characteristics of the global structure.
3. A method for automatically modelling a parametric model of a side view of a vehicle body according to claim 2, wherein: the step 3 of deleting the noise boundary and optimizing the boundary extracted in the step 2 comprises the following steps:
step 3.1, defining a noise boundary;
and 3.2, optimizing a boundary extraction result by using an intermediate boundary point generation algorithm.
4. A method for automatically modelling a parametric model of a side view of a vehicle body according to claim 3, wherein: the noise edge in step 3.1 includes two cases: based on the Canny edge feature points, if the number of points on a certain edge far away from the Canny edge feature points by a given threshold lambda is more than half of all the points on the edge, the edge can be regarded as a noise edge; the edge that intersects the image edge is also a noise edge.
5. The method for automatically modeling the parametric model of the side view of the vehicle body as set forth in claim 4, wherein: and 3.2, after the noise edges are deleted, the clean main characteristic information of the side view of the vehicle body can be extracted, but the phenomenon of 'burr' appears on the main boundary at the moment, and by utilizing the condition that the gray value of the middle point of the boundary is lower than the gray values of the boundary points at the two sides and the distances between the middle point of the boundary and the boundary points at the two sides are basically equal, a middle boundary point generation algorithm is designed, and the middle boundary point is calculated to obtain an ideal single pixel sequence boundary result.
6. The method for automatically modeling the parametric model of the side view of the vehicle body as set forth in claim 1, wherein: and 4.1, connecting the wheels of the model in the step 4.1 end to end by 4 cubic Bezier curves, wherein each Bezier curve comprises 4 control points, defining the intersection point of connecting lines formed by two pairs of end to end control points as the wheel center of the wheel, and calibrating the template database by taking the wheel center of the front wheel as a coordinate origin, so that the subsequent point set matching is facilitated.
7. The method for automatically modeling a parametric model of a side view of a vehicle body as set forth in claim 6, wherein: the step 4.2 of roughly extracting the vehicle body model based on the template database and the key points comprises the following steps:
step 4.2.1, defining a vehicle body side view key point vector ZmIndividual body model shape vector siA template database matrix S and a template database key point sub-matrix Z;
step 4.2.2, establishing a vehicle body side view key point vector ZmFunctional relation Z with template database key point submatrix ZmSolving the model contribution quantity by using an improved reconstruction algorithm Car-OMP, wherein theta is Z multiplied by theta and is used as the model contribution quantity;
step 4.2.3, obtaining the shape vector S of the vehicle body model by using the model contribution amount obtained in the step 4.2.2 and the model library matrix SiThis allows the rough extraction of the body model to be completed, resulting in an initial model.
8. The method for automatically modeling a parametric model of a side view of a vehicle body as set forth in claim 7, wherein: in the step 4.2.1, the N key points of the side view are unfolded into a vector Zm＝(x1,y1,x2,y2...xN,yN)TWherein (x)i,yi) Coordinates of key points on the side view; defining a single body model as a shape vector si＝(x′1,y′1,x′2,y′2...x′n,y′n)TWherein (x'i,y′i) For the coordinates of the control points on the individual body models, n is 97 × 4, i.e. 97 Be zie for each vehicle typer curves, each Bezier curve comprises 4 control points; the entire template database can be converted into a matrix S, which is equivalent to a matrix consisting of m shape vectors SiThe method comprises the following steps of (1) forming, wherein m is the number of vehicle body models in a template database; and extracting a submatrix Z with each vehicle body model only containing 25 vehicle body key points from the S.
9. The method for automatically modeling a parametric model of a side view of a vehicle body as set forth in claim 8, wherein: said step 4.2.2 calculating the model contribution using the Car-OMP reconstruction algorithm comprises the following steps:
step 4.2.2.1, initialize the parameter, i.e. residual r0＝ZmIndex set
Figure FDA0003179790650000021
Storing a matrix of column vectors corresponding to the index λ
Figure FDA0003179790650000022
The iteration time t is 1;
step 4.2.2.2, finding out residual error r and each column Z of the key point submatrix Z of the template databaseiThe index λ corresponding to the maximum inner product is:
λt＝argmaxi＝1…m|<rt-1,Zi>| (1)
in the formula, ZiThe ith column of the template database key point submatrix Z is shown, m is the number of the automobile body models in the template database,<rt-1,Zi>representing the residual r and each column Z of the submatrixiInner product of, arg maxi＝1…m| mean the index λ corresponding to the maximum absolute value of the inner productt；
Step 4.2.2.3, update index set At＝At-1∪{λtAnd updating a matrix for storing column vectors corresponding to the index lambda
Figure FDA0003179790650000031
At the step 4.2.2.4, the method further comprises the steps of,z is obtainedm＝ZtθtThe least squares solution of (a), namely:
θt＝argmin||Zm-Ztθt||＝(Zt TZt)-1Zt TZm (2)
in the formula, argmaxi＝1…nThe corresponding model contribution theta when the maximum value is taken by the modelt；
Step 4.2.2.5, update residual rt，rtThe calculation method of (c) is as follows:
rt＝Zm-Ztθt＝Zm-Zt(Zt TZt)-1Zt TZm (3)
step 4.2.2.6, setting the total iteration times as K, and stopping iteration if the current iteration times t is more than K; otherwise, let t be t +1, perform step 4.2.2.1;
in step 4.2.2.7, the above loop is executed when K ≠ 0, and when K ≠ 0, θ ═ (Z) is employed-1×Zm。
10. A method for automatically modelling a parametric model of a side view of a vehicle body according to claim 3, wherein: and 4.3, combining the boundary information in the step 3 to accurately extract the vehicle body model, namely calculating the distance from each point on the model curve to all pixel points of the nearest boundary and the index of the corresponding boundary, defining the two nearest points as a group, reducing the distance of each group of points to a given threshold value through iteration, considering that the model curve at the moment is consistent with the boundary extracted in the segmentation algorithm, simultaneously searching for the optimal control point of the Bezier curve in the iteration process, and replacing the four control points on the original curve with the optimal control points to realize the accurate extraction of the vehicle body side view parameterized model.
CN202110843172.5A 2021-07-26 2021-07-26 Automatic modeling method of vehicle body side view parameterized model Active CN113516759B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110843172.5A CN113516759B (en) 2021-07-26 2021-07-26 Automatic modeling method of vehicle body side view parameterized model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110843172.5A CN113516759B (en) 2021-07-26 2021-07-26 Automatic modeling method of vehicle body side view parameterized model

Publications (2)

Publication Number Publication Date
CN113516759A true CN113516759A (en) 2021-10-19
CN113516759B CN113516759B (en) 2022-04-26

Family

ID=78068649

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110843172.5A Active CN113516759B (en) 2021-07-26 2021-07-26 Automatic modeling method of vehicle body side view parameterized model

Country Status (1)

Country Link
CN (1) CN113516759B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104881513A (en) * 2015-04-17 2015-09-02 大连理工大学 3D (three-dimensional) printing based processing technique of automobile styling concept model
US20170030716A1 (en) * 2015-07-29 2017-02-02 Invensense, Inc. Method and apparatus for user and moving vehicle detection
CN107346437A (en) * 2017-07-03 2017-11-14 大连理工大学 The extraction method of body side view parameter model
CN107357984A (en) * 2017-07-03 2017-11-17 大连理工大学 The method for full automatic extraction of face modeling parameters model before vehicle body
US20200111264A1 (en) * 2018-10-05 2020-04-09 The Toronto-Dominion Bank System and method for remotely indicating vehicular damage

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104881513A (en) * 2015-04-17 2015-09-02 大连理工大学 3D (three-dimensional) printing based processing technique of automobile styling concept model
US20170030716A1 (en) * 2015-07-29 2017-02-02 Invensense, Inc. Method and apparatus for user and moving vehicle detection
CN107346437A (en) * 2017-07-03 2017-11-14 大连理工大学 The extraction method of body side view parameter model
CN107357984A (en) * 2017-07-03 2017-11-17 大连理工大学 The method for full automatic extraction of face modeling parameters model before vehicle body
US20200111264A1 (en) * 2018-10-05 2020-04-09 The Toronto-Dominion Bank System and method for remotely indicating vehicular damage

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
张学芳: "基于单视图的汽车造型数字化建模研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 *
王博 等: "一种自动提取车身前视图参数化模型的方法", 《汽车工程学报》 *
王博: "汽车外造型数字化与参数化关键技术研究", 《中国优秀博硕士学位论文全文数据库(博士)工程科技Ⅱ辑》 *

Also Published As

Publication number Publication date
CN113516759B (en) 2022-04-26

Similar Documents

Publication Publication Date Title
Zhou et al. Cocosnet v2: Full-resolution correspondence learning for image translation
CN114782691B (en) Robot target identification and motion detection method based on deep learning, storage medium and equipment
KR20200070409A (en) Human hairstyle creation method based on multiple feature search and transformation
CN109829353B (en) Face image stylizing method based on space constraint
CN109740537B (en) Method and system for accurately marking attributes of pedestrian images in crowd video images
CN104992143B (en) A kind of Chinese-character stroke extraction method towards vector font
CN111368759B (en) Monocular vision-based mobile robot semantic map construction system
CN108765434B (en) Contour extraction method based on additive remanufacturing point cloud model
CN110032952B (en) Road boundary point detection method based on deep learning
CN113393503A (en) Classification-driven shape prior deformation category-level object 6D pose estimation method
CN107886471A (en) A kind of unnecessary object minimizing technology of photo based on super-pixel Voting Model
CN114359642A (en) Multi-modal medical image multi-organ positioning method based on one-to-one target query Transformer
CN110909778B (en) Image semantic feature matching method based on geometric consistency
CN116310095A (en) Multi-view three-dimensional reconstruction method based on deep learning
Li et al. Multi-attribute regression network for face reconstruction
CN105869153B (en) The non-rigid Facial Image Alignment method of the related block message of fusion
WO2022222091A1 (en) Method for generating character bas-relief model on basis of single photo
CN113516759B (en) Automatic modeling method of vehicle body side view parameterized model
CN112200905B (en) Three-dimensional face complement method
Liu et al. Nnnet: New normal guided depth completion from sparse lidar data and single color image
Zhou et al. DR-Pose: A Two-Stage Deformation-and-Registration Pipeline for Category-Level 6D Object Pose Estimation
CN111563515A (en) Improved Faster-RCNN-based license plate recognition method
Gallego et al. Joint multi-view foreground segmentation and 3d reconstruction with tolerance loop
Shi et al. City-scale continual neural semantic mapping with three-layer sampling and panoptic representation
Alvén et al. Shape-aware label fusion for multi-atlas frameworks

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