CN111709494B - Novel hybrid-optimized image stereo matching method - Google Patents
Novel hybrid-optimized image stereo matching method Download PDFInfo
- Publication number
- CN111709494B CN111709494B CN202010669921.2A CN202010669921A CN111709494B CN 111709494 B CN111709494 B CN 111709494B CN 202010669921 A CN202010669921 A CN 202010669921A CN 111709494 B CN111709494 B CN 111709494B
- Authority
- CN
- China
- Prior art keywords
- optimization
- parallax
- variable
- algorithm
- region
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
A novel hybrid optimized image stereo matching method belongs to the field of computer vision and image processing. The method solves the problems that the traditional linear growth algorithm is easily influenced by the selection of initial parameters, and has poor reliability and low efficiency. It determines the range of parallax areas in the binocular image pair to be matched; optimizing the objective function by utilizing an optimization algorithm to obtain an optimization result and an optimization variable; substituting the optimization result and the optimization variable into a linear growth algorithm, calculating the parallax energy of the growth area near each root point, and obtaining the parallax energy of the parallax area; filtering parallax energy of the parallax region through a box type filtering algorithm to eliminate image edge mismatching; and completing the matching of the binocular image pair to be matched. The invention is suitable for image matching.
Description
Technical Field
The invention belongs to the field of computer vision and image processing, and particularly relates to a hybrid-optimized image stereo matching method.
Background
With the rapid development of artificial intelligence, computer vision has become a research hotspot. Computer vision is a discipline that studies how to track, identify, analyze, process, etc., objects with cameras and computers instead of the human eye. Stereoscopic vision is an important branch. The stereo matching is to find out the corresponding points and points between the images in the given two-dimensional image, and calculate the parallax of the corresponding pixel points matched in the left and right binocular images, thereby obtaining the depth information. The method has wide application in the fields of topographic mapping, three-dimensional reconstruction, robots, remote sensing image analysis, target detection and recognition and the like.
There are many methods for solving the problem of stereo matching of images. The method is widely applied to methods such as a global-based stereo matching method and a local-based stereo matching method, but is constrained by an algorithm, and the algorithm is often required to be improved to be optimized in actual application so as to reduce the error matching rate and improve the reliability.
Because the initial root point and the threshold value of the traditional linear growth algorithm are often set through human experience, the matching rate of the algorithm can be greatly influenced by the fact that the initial parameters are too large or too small, and the reliability of the algorithm is not ideal.
Disclosure of Invention
The invention provides a hybrid-optimized image stereo matching method for solving the problems that a traditional linear growth algorithm is easily influenced by initial parameter selection, and has poor reliability and low efficiency.
The novel hybrid optimized image stereo matching method specifically comprises the following steps:
step a, determining a parallax region range in a binocular image pair to be matched;
step b, limiting the function value of the target function matched with the image by utilizing the parallax region range, and optimizing the target function by utilizing an optimization algorithm to obtain an optimization result and an optimization variable; the optimization variables comprise a threshold value and a parallax value of a parallax region;
the optimization variables include a threshold value of a parallax region and a parallax value:
substituting the optimization result and the optimization variable into a linear growth algorithm, calculating the parallax energy of the growth area near each root point, and obtaining the parallax energy of the parallax area;
step d, filtering parallax energy of the parallax region through a box type filtering algorithm to eliminate image edge mismatching; and completing the matching of the binocular image pair to be matched.
Further, in the invention, the optimization algorithm in the step b comprises a pigeon cluster algorithm and a secondary annealing algorithm.
Further, in the present invention, the specific method for optimizing the objective function by using the optimization algorithm in the step b to obtain the optimization result and the optimization variation is as follows:
step b1, initializing a pigeon cluster algorithm according to the parallax region range in the binocular image pair to be matched, and setting an initial value, a result threshold delta and an iteration number maximum value N2 of a variable;
step b2, taking a cost function J of a pigeon swarm algorithm as an objective function;
step b3, updating the position and speed of each pigeon and updating the pigeon group center according to the compass operator and the adaptive value; optimizing the variable to obtain an optimized variable;
step b4, introducing the optimization variable into a cost function J, and calculating an optimization result; b, judging whether the obtained optimization result is smaller than a result threshold delta, if yes, completing the variable optimization result and the optimization variable, otherwise, returning to the execution step b5;
step b5, if the iteration times is smaller than the maximum value N2 of the iteration times, returning to the step b3, otherwise, executing the step b6;
and b6, dynamically adjusting the annealing temperature by using a secondary annealing algorithm, adjusting the optimization variable, and returning to the step b4.
Further, in the present invention, the specific method for obtaining the optimization variable in the step b3 is as follows:
using the formula:
V (t) (k)=V (t-1) (k)e -Rt +rand·(X g -V (t-1) (k))
X (t) (k)=X (t-1) (k)+V (t) (k)
updating the pigeon group center, wherein V (t) (k) Representing the speed of the t-th iteration of the optimization variable, V (t-1) (k) Representing the speed of t-1 th iteration of the optimization variable, wherein t is the current iteration number, R is a map compass factor, k is the kth pigeon individual in the pigeon cluster, and X (t) (k) Representing the position of the t-th iteration of the optimization variable, X (t-1) (k) Representing the position of the t-1 th iteration of the optimization variable, xg is the global optimal solution in the current population, and rand is [0,1]Random values over the interval;
selecting a population center point C (t) as a flying reference direction of the residual pigeon group, updating the positions of the individuals, and utilizing an updating formula:
X (t) (k)=X (t-1) (k)+rand·(C(t)-X (t-1) (k))
obtaining the position X of the t-th iteration of the optimization variable (t) (k) I.e. optimizing variables, wherein, the fitness (·) is an fitness function, which is constructed by an objective function, and the NP is the current population number.
Further, in the present invention, the specific method for dynamically adjusting the annealing temperature and adjusting the optimization variables by using the secondary annealing algorithm in the step b6 is as follows:
step b61, initializing a secondary annealing algorithm; initializing parameters;
step b62, using the new point generation formula:
X new =Xo ld +r·s·(X max -Xo ld )
or (b)
X new =Xo ld +r·s·(X min -Xo ld )
Acquiring new optimization changesQuantity X new Wherein X is new The variation range of (C) is X min ≤X new ≤X max ,X old Initial value is position X of the N2 nd iteration of the optimization variable (N2) (k) The method comprises the steps of carrying out a first treatment on the surface of the r is a random scaling factor, s is an adaptive direction factor, and r is a variable change amplitude value by the scaling factor.
Further, in the present invention, in step c, substituting the optimization result and the optimization variable into the linear growth algorithm, and calculating the parallax energy of the growth area near each root point, the specific method for obtaining the parallax energy of the parallax area is as follows:
step c1, performing parallax region calculation by using a traditional linear growth algorithm, and calculating an optimization variable and time consumed by calculating the optimization variable; the optimization variables include threshold and disparity values for the disparity region:
step c2, determining constraint conditions of variable optimization problems according to variables to be optimized;
step c3, obtaining a constrained objective function according to the constraint conditions of the optimization variables and the variable optimization problem;
and c4, when the difference between the function values of the two iterations before and after the constrained objective function is within a set convergence range, obtaining the parallax energy of the parallax region.
Further, in the present invention, the specific method for calculating the optimization variable and the time consumed for calculating the optimization variable in step c1 by using the conventional linear growth algorithm to calculate the parallax region includes the following steps:
step c11, obtaining an initial root point error energy initial threshold Th d ;
Step c12, calculating the parallax energy of the root point adjacent point as the regional parallax Th d i ;
Step c13, judging Th d i <Th d If yes, allowing the region to continue growing until the region growth is completed, and returning to c11 to find a new root point; c14, executing the step until all the parallax areas are grown; otherwise, directly returning to c11 to search for a new root point;
step c14, counting the time required for the complete growth of the parallax region by using time statistics functions (tic and toc) in matlab.
Further, in the present invention, constraint conditions of the variable optimization problem described in step c2 are:
the constraint conditions are as follows: d is more than or equal to 0 and less than or equal to d max ,0≤Th d ≤Th dmax Wherein d is the parallax value, d max Is the upper limit of the disparity value; th (Th) d Is threshold value, th dmaxx Is the upper threshold; d, d max =∈[40,200],Th dmax =∈(0,60]。
Further, in the present invention, the constrained objective function described in step c3 is:
(n=1,2,…,N)
Subject to:0≤d≤d max
wherein r is 1 ·E d (di,Th d n ) As a function of the difference energy, r 1 Weights representing optimization targets, d i A parallax value, th, representing the ith root point in the current parallax region d n A threshold value of the nth parallax region, E d Representing error energy; r is (r) 2 ·t f (n) is the time taken for the parallax region; r is (r) 2 Weights representing optimization targets, t f (n) represents a calculation time consumed for calculating the n-th parallax region.
The invention can adaptively adjust the parallax threshold by combining the traditional linear growth model with the pigeon group-simulated annealing algorithm, and overcomes the defects of local optimization and global optimization. Compared with the traditional linear growth algorithm, the method is not influenced by the initial value of the parameter, has good robustness, and in addition, the box type filtering greatly improves the reliability of the parallax map, and provides a good calculation method for obtaining a more effective parallax map through stereo matching.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a schematic illustration of a specific flow of the method of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that, without conflict, the embodiments of the present invention and features of the embodiments may be combined with each other.
The first embodiment is as follows: the following describes a method for stereo matching of images with new hybrid optimization according to the present embodiment with reference to fig. 1 and 2, where the method specifically includes:
step a, determining a parallax region range in a binocular image pair to be matched;
step b, limiting the function value of the target function matched with the image by utilizing the parallax region range, and optimizing the target function by utilizing an optimization algorithm to obtain an optimization result and an optimization variable; the optimization variables comprise a threshold value and a parallax value of a parallax region;
substituting the optimization result and the optimization variable into a linear growth algorithm, calculating the parallax energy of the growth area near each root point, and obtaining the parallax energy of the parallax area;
step d, filtering parallax energy of the parallax region through a box type filtering algorithm to eliminate image edge mismatching; and completing the matching of the binocular image pair to be matched.
The range of the parallax region in the binocular image pair to be matched is usually set by experience according to two images of the binocular image pair acquired by a binocular system, and the range of the parallax value is usually [0,40].
Further, in this embodiment, the optimization algorithm described in the step b includes a pigeon cluster algorithm and a secondary annealing algorithm.
Further, in this embodiment, the specific method for optimizing the objective function by using the optimization algorithm in the step b to obtain the optimization result and the optimization variation is as follows:
step b1, initializing a pigeon cluster algorithm according to the parallax region range in the binocular image pair to be matched, and setting an initial value, a result threshold delta and an iteration number maximum value N2 of a variable;
step b2, taking a cost function J of a pigeon swarm algorithm as an objective function;
step b3, updating the position and speed of each pigeon and updating the pigeon group center according to the compass operator and the adaptive value; optimizing the variable to obtain an optimized variable;
step b4, introducing the optimization variable into a cost function J, and calculating an optimization result; b, judging whether the obtained optimization result is smaller than a result threshold delta, if yes, completing the variable optimization result and the optimization variable, otherwise, returning to the execution step b5;
step b5, if the iteration times is smaller than the maximum value N2 of the iteration times, returning to the step b3, otherwise, executing the step b6;
and b6, dynamically adjusting the annealing temperature by using a secondary annealing algorithm, adjusting the optimization variable, and returning to the step b4.
Further, in this embodiment, the specific method for obtaining the optimization variable in the step b3 is as follows:
using the formula:
V (t) (k)=V (t-1) (k)e -Rt +rand·(X g -V (t-1) (k))
X (t) (k)=X (t-1) (k)+V (t) (k)
updating the pigeon group center, wherein V (t) (k) Representing the speed of the t-th iteration of the optimization variable, V (t-1) (k) Representing the speed of t-1 th iteration of the optimization variable, wherein t is the current iteration number, R is a map compass factor, and k is the kth pigeon in the pigeon groupBody, X (t) (k) Representing the position of the t-th iteration of the optimization variable, X (t-1) (k) Representing the position of the t-1 th iteration of the optimization variable, xg is the global optimal solution in the current population, and rand is [0,1]Random values over the interval;
selecting a population center point C (t) as a flying reference direction of the residual pigeon group, updating the positions of the individuals, and utilizing an updating formula:
X (t) (k)=X (t-1) (k)+rand·(C(t)-X (t-1) (k))
obtaining the position X of the t-th iteration of the optimization variable (t) (k) I.e. optimizing variables, wherein, the fitness (·) is an fitness function, which is constructed by an objective function, and the NP is the current population number.
Further, in this embodiment, the specific method for dynamically adjusting the annealing temperature and adjusting the optimization variables in the step b6 by using the secondary annealing algorithm is as follows:
step b61, initializing a secondary annealing algorithm; initializing parameters;
initial temperature t 0 The maximum number of times of external circulation N1, the maximum number of times of internal circulation N2, the self-adaptive direction s and the scale factor r. Initial variable initial value temperature t of the algorithm 0 =1000°,N1=100,N2=100;
Step b62, using the new point generation formula:
X new =Xo ld +r·s·(X max -Xo ld )
or (b)
X new =Xo ld +r·s·(X min -Xo ld )
Obtaining new optimization variables X new Wherein X is new The variation range of (C) is X min ≤X new ≤X max ,X old Initial value is position X of the N2 nd iteration of the optimization variable (N2) (k) The method comprises the steps of carrying out a first treatment on the surface of the r is a random proportionality coefficient, s is a self-adaptive squareThe orientation factor, r, is the variable change amplitude by the scaling factor.
In this embodiment, r is a random scaling factor, r= (1-t) k )×rand/t k The value of rand E (0.1) is related to the temperature value, and the higher the temperature is, the larger the proportional coefficient is, and the more the variable changes. As the temperature decreases, the variation of the variable tends to smooth and eventually converge. s is an adaptive direction factor, s=2rand_1, to determine the direction of the new point change. After the new point is obtained, the numerical range of the new point needs to be judged, and the new point out of the range is processed. Then for the new point X generated new Judging if the corresponding cost function value J new Less than the current optimal cost function value J min Then n 1 =n 1 +1, and update J min =J new And corresponding x=x new A value; judging whether the number of external circulation times of the simulated annealing method is reached, namely n 1 >N 1 If the value is satisfied, the minimum parallax value is saved, otherwise, the temperature is updated, and the temperature is removed (the parameters in the marked yellow formula are determined) t k+1 =update(t k ) Processing, update (t) k )=C t k (temperature update formula: t) k+1 =Ct k . Updating the reason: used for adjusting the random proportionality coefficient r, the higher the temperature is, the larger the r is; c is the temperature decay coefficient) k=k+1, C e (0, 1), and let n 1 =n 1 +1; judging whether the external circulation times of the simulated annealing method are reached or not again until n is met 1 >And N, saving the minimum parallax value.
Further, in this embodiment, in the step c, the optimization result and the optimization variable are substituted into the linear growth algorithm, and the specific method for calculating the parallax energy of the growth area near each root point and obtaining the parallax energy of the parallax area is as follows:
step c1, performing parallax region calculation by using a traditional linear growth algorithm, and calculating an optimization variable and time consumed by calculating the optimization variable; the optimization variables are the threshold value of the parallax region and the parallax value:
step c2, determining constraint conditions of variable optimization problems according to variables to be optimized;
step c3, obtaining a constrained objective function according to the constraint conditions of the optimization variables and the variable optimization problem;
and c4, when the difference between the function values of the two iterations before and after the constrained objective function is within a set convergence range, obtaining the parallax energy of the parallax region.
Further, in this embodiment, the specific method for calculating the optimization variable and the time consumed for calculating the optimization variable in step c1 by using the conventional linear growth algorithm to calculate the parallax region includes the following steps:
step c11, obtaining an initial root point error energy initial threshold Th d ;
Step c12, calculating the parallax energy of the root point adjacent point as the regional parallax Th d i The method comprises the steps of carrying out a first treatment on the surface of the Step c13, judging Th d i <Th d If yes, allowing the region to continue growing until the region growth is completed, and returning to c11 to find a new root point; c14, executing the step until all the parallax areas are grown; otherwise, directly returning to d11 to find a new root point;
step c14, counting the time required for the complete growth of the parallax region by using time statistics functions (tic and toc) in matlab.
Further, in the present embodiment, constraint conditions of the variable optimization problem described in step c2 are:
the constraint conditions are as follows: d is more than or equal to 0 and less than or equal to d max ,0≤Th d ≤Th dmax Wherein d is the parallax value, d max Is the upper limit of the disparity value; th (Th) d Is threshold value, th dmaxx Is the upper threshold; d, d max =∈[40,200],Th dmax =∈(0,60]。
Further, in the present invention, the constrained objective function described in step c3 is:
(n=1,2,…,N)
Subject to:0≤d≤d max
wherein r is 1 ·E d (di,Th d n ) As a function of the difference energy, r 1 Weights representing optimization targets, d i A parallax value, th, representing the ith root point in the current parallax region d n A threshold value of the nth parallax region, E d Representing error energy; r is (r) 2 ·t f (n) when consumed by the parallax region; r is (r) 2 Weights representing optimization targets, t f (n) represents a calculation time consumed for calculating the n-th parallax region.
The method disclosed by the invention comprehensively considers the calculation efficiency and the image reliability, can adaptively adjust the threshold value in the parallax calculation process and calculate the optimal parallax value of the root point, is not influenced by the initial threshold value parameter, reduces the image mismatching again through box type filtering, and well overcomes the defects of high mismatching rate and poor real-time robustness of the traditional linear growth algorithm. The group optimization algorithm is inspired by the group activity of the natural living beings, is integrated into the group intelligent optimization algorithm on the basis of the traditional linear growth algorithm, and has the advantages of strong optimizing capability and strong robustness. Firstly, a pigeon swarm algorithm is utilized to solve the problem that the traditional linear growth algorithm is easily disturbed by an initial value. The pigeon swarm algorithm has the advantages of rapid convergence, less required adjustment parameters and easy realization; different formulas are used for optimizing variables at different stages, and satisfactory optimization results are found. However, the pigeon cluster algorithm has low convergence accuracy and is easy to fall into local optimum, so that the variable is further optimized by using a secondary simulated annealing algorithm. The simulated annealing algorithm can accept non-optimal solutions with a certain probability, and the accuracy can be further improved under the condition that the premature convergence of pigeon flocks is avoided by using the simulated annealing algorithm for two times. Finally, considering that the probability of mismatching at the edge of the image is highest when the image is matched, filtering is performed by using a box type filtering linear filtering method, and the mismatching at the edge of the image is eliminated by rapidly integrating and calculating the image, so that the matching efficiency is further improved. Compared with the traditional linear growth algorithm, the algorithm provided by the invention has the advantages that the generated parallax image pixels are higher, the outline information is more obvious, the edges are clearer, the continuity is higher, the noise is obviously reduced, more view characteristics can be captured, and the parallax image is more obvious in a single-color area.
Although the invention herein has been described with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present invention. It is therefore to be understood that numerous modifications may be made to the illustrative embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the different dependent claims and the features described herein may be combined in ways other than as described in the original claims. It is also to be understood that features described in connection with separate embodiments may be used in other described embodiments.
Claims (8)
1. The novel hybrid optimized image stereo matching method is characterized by comprising the following steps of:
step a, determining a parallax region range in a binocular image pair to be matched;
step b, limiting the function value of the target function matched with the image by utilizing the parallax region range, and optimizing the target function by utilizing an optimization algorithm to obtain an optimization result and an optimization variable; the optimization variables comprise a threshold value and a parallax value of a parallax region;
the specific method for optimizing the objective function by utilizing the optimization algorithm to obtain the optimization result and the optimization variation comprises the following steps:
step b1, initializing a pigeon cluster algorithm according to the parallax region range in the binocular image pair to be matched, and setting an initial value, a result threshold delta and an iteration number maximum value N2 of a variable;
step b2, taking a cost function J of a pigeon swarm algorithm as an objective function of image matching;
step b3, updating the position and speed of each pigeon and updating the pigeon group center according to the compass operator and the fitness function; optimizing the variable to obtain an optimized variable;
step b4, introducing the optimization variable into a cost function J, and calculating an optimization result; b, judging whether the obtained optimization result is smaller than a result threshold delta, if yes, completing the variable optimization result and the optimization variable, otherwise, returning to the execution step b5;
step b5, if the iteration times is smaller than the maximum value N2 of the iteration times, returning to the step b3, otherwise, executing the step b6;
step b6, dynamically adjusting annealing temperature by using a secondary annealing algorithm, adjusting an optimization variable, and returning to the step b;
substituting the optimization result and the optimization variable into a linear growth algorithm, calculating the parallax energy of the growth area near each root point, and obtaining the parallax energy of the parallax area;
step d, filtering parallax energy of the parallax region through a box type filtering algorithm to eliminate image edge mismatching; and completing the matching of the binocular image pair to be matched.
2. The novel hybrid optimized image stereo matching method according to claim 1, wherein the optimization algorithm in the step b comprises a pigeon cluster algorithm and a secondary annealing algorithm.
3. The novel hybrid-optimized image stereo matching method according to claim 2, wherein the specific method for obtaining the optimized variable in step b3 is as follows:
using the formula:
V (t) (k)=V (t-1) (k)e -Rt +rand·(X g -V (t-1) (k))
X (t) (k)=X (t-1) (k)+V (t) (k)
updating the pigeon group center, wherein V (t) (k) Representing the speed of the t-th iteration of the optimization variable, V (t-1) (k) Representing the speed of t-1 th iteration of the optimization variable, wherein t is the current iteration number, R is a map compass factor, k is the kth pigeon individual in the pigeon cluster, and X (t) (k) Representing the t-th time of the optimization variableIterative position, X (t-1) (k) Representing the position of the t-1 th iteration of the optimization variable, xg is the global optimal solution in the current population, and rand is [0,1]Random values over the interval;
selecting a population center point C (t) as a flying reference direction of the residual pigeon group, updating the positions of the individuals, and utilizing an updating formula:
X (t) (k)=X (t-1) (k)+rand·(C(t)-X (t-1) (k))
obtaining the position X of the t-th iteration of the optimization variable (t) (k) I.e. optimizing variables, wherein, the fitness (·) is an fitness function, which is constructed by an objective function, and the NP is the current population number.
4. The novel hybrid-optimized image stereo matching method according to claim 2, wherein the specific method for dynamically adjusting the annealing temperature and adjusting the optimization variables in the step b6 by using a secondary annealing algorithm is as follows:
step b61, initializing a secondary annealing algorithm; initializing parameters;
step b62, using the new point generation formula:
X new =X old +r·s·(X max -X old )
or (b)
X new =X old +r·s·(X min -X old )
Obtaining new optimization variables X new Wherein X is new The variation range of (C) is X min ≤X new ≤X max ,X old Initial value is position X of the N2 nd iteration of the optimization variable (T) (k) The method comprises the steps of carrying out a first treatment on the surface of the r is a random scaling factor, s is an adaptive direction factor, and r is a variable change amplitude value by the scaling factor.
5. The novel hybrid-optimized image stereo matching method according to claim 2, wherein in the step c, the optimization result and the optimization variable are substituted into the linear growth algorithm, and the parallax energy of the growth area near each root point is calculated, and the specific method for obtaining the parallax energy of the parallax area is as follows:
step c1, performing parallax region calculation by using a traditional linear growth algorithm, and calculating an optimization variable and time consumed by calculating the optimization variable; the optimization variables are the threshold value of the parallax region and the parallax value:
step c2, determining constraint conditions of variable optimization problems according to variables to be optimized;
step c3, obtaining a constrained objective function according to the constraint conditions of the optimization variables and the variable optimization problem;
and c4, when the difference between the function values of the two iterations before and after the constrained objective function is within a set convergence range, obtaining the parallax energy of the parallax region.
6. The method for stereoscopic matching of images in accordance with claim 5, wherein the method for computing the optimized variables and the time consumed for computing the optimized variables in step c1 by computing the parallax region using the conventional linear growth algorithm comprises the steps of:
step c11, obtaining an initial root point error energy initial threshold Th d ;
Step c12, calculating the parallax energy of the root point adjacent point as the regional parallax Th d i ;
Step c13, judging Th d i <Th d If yes, allowing the region to continue growing until the region growth is completed, and returning to c11 to find a new root point; c14, executing the step until all the parallax areas are grown; otherwise, directly returning to c11 to search for a new root point;
and step c14, counting the time required for completing all growth of the parallax region by utilizing a time counting function in matlab.
7. The novel hybrid-optimized image stereo matching method according to claim 5, wherein the constraint condition of the variable optimization problem in step c2 is:
the constraint conditions are as follows: d is more than or equal to 0 and less than or equal to d max ,0≤Th d ≤Th dmax Wherein d is the parallax value, d max Is the upper limit of the disparity value; th (Th) d Is threshold value, th dmaxx Is the upper threshold; d, d max =∈[40,200],Th dmax =∈(0,60]。
8. The method for stereoscopic matching of images according to claim 5, wherein,
the constrained objective function described in step c3 is:
Subject to:0≤d≤d max
wherein r is 1 ·E d (di,Th d n ) As a function of the difference energy, r 1 Weights representing optimization targets, d i A parallax value, th, representing the ith root point in the current parallax region d n A threshold value of the nth parallax region, E d Representing error energy; r is (r) 2 ·t f (n) is the time taken for the parallax region; r is (r) 2 Weights representing optimization targets, t f (n) represents a calculation time consumed for calculating the n-th parallax region.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010669921.2A CN111709494B (en) | 2020-07-13 | 2020-07-13 | Novel hybrid-optimized image stereo matching method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010669921.2A CN111709494B (en) | 2020-07-13 | 2020-07-13 | Novel hybrid-optimized image stereo matching method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111709494A CN111709494A (en) | 2020-09-25 |
CN111709494B true CN111709494B (en) | 2023-05-26 |
Family
ID=72545756
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010669921.2A Active CN111709494B (en) | 2020-07-13 | 2020-07-13 | Novel hybrid-optimized image stereo matching method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111709494B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106355570A (en) * | 2016-10-21 | 2017-01-25 | 昆明理工大学 | Binocular stereoscopic vision matching method combining depth characteristics |
WO2018072351A1 (en) * | 2016-10-20 | 2018-04-26 | 北京工业大学 | Method for optimizing support vector machine on basis of particle swarm optimization algorithm |
CN111191694A (en) * | 2019-12-19 | 2020-05-22 | 浙江科技学院 | Image stereo matching method |
-
2020
- 2020-07-13 CN CN202010669921.2A patent/CN111709494B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018072351A1 (en) * | 2016-10-20 | 2018-04-26 | 北京工业大学 | Method for optimizing support vector machine on basis of particle swarm optimization algorithm |
CN106355570A (en) * | 2016-10-21 | 2017-01-25 | 昆明理工大学 | Binocular stereoscopic vision matching method combining depth characteristics |
CN111191694A (en) * | 2019-12-19 | 2020-05-22 | 浙江科技学院 | Image stereo matching method |
Non-Patent Citations (7)
Title |
---|
吴立刚 等."一种稳健的角点匹配方法".《南京信息工程大学学报(自然科学版)》.2012,(第6期),550-554. * |
宋雯君."基于双目立体匹配算法的雾霾图像清晰化处理研究".《中国优秀硕士学位论文全文数据库 信息科技辑》.2019,(第12期),I138-571. * |
彭彬彬 等."基于区域快速生长的双目视觉立体匹配算法研究".《系统仿真技术》.2015,(第7期),213-218. * |
杨育三."基于LED显示屏的集成立体裸眼显示机理研究".《中国优秀硕士学位论文全文数据库 信息科技辑》.2020,(第1期),I136-178. * |
胡步发 等."基于极线距离变换的人脸立体匹配算法".《仪器仪表学报》.2015,(第2期),360-367. * |
霍智勇 等.基于区域的立体匹配算法研究.《南京邮电大学学报(自然科学版)》.2011,(第03期),44-49. * |
霍智勇 等.基于线性生长的区域立体匹配算法研究.《仪器仪表学报》.2011,(第4期),819-825. * |
Also Published As
Publication number | Publication date |
---|---|
CN111709494A (en) | 2020-09-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109949375B (en) | Mobile robot target tracking method based on depth map region of interest | |
CN111798462B (en) | Automatic delineation method of nasopharyngeal carcinoma radiotherapy target area based on CT image | |
WO2018000752A1 (en) | Monocular image depth estimation method based on multi-scale cnn and continuous crf | |
CN109146948A (en) | The quantization of crop growing state phenotypic parameter and the correlation with yield analysis method of view-based access control model | |
CN110246151B (en) | Underwater robot target tracking method based on deep learning and monocular vision | |
CN108235725A (en) | Track based on high in the clouds ground drawing generating method, device, equipment and application program | |
CN111998862B (en) | BNN-based dense binocular SLAM method | |
CN111402311A (en) | Knowledge distillation-based lightweight stereo parallax estimation method | |
CN106778767B (en) | Visual image feature extraction and matching method based on ORB and active vision | |
CN112614163B (en) | Target tracking method and system integrating Bayesian track reasoning | |
CN105631899A (en) | Ultrasonic image motion object tracking method based on gray-scale texture feature | |
CN110796691A (en) | Heterogeneous image registration method based on shape context and HOG characteristics | |
CN110706270A (en) | Self-adaptive scene binocular stereo matching method based on convolutional neural network | |
CN112581543A (en) | Visual navigation method for moving robot | |
CN110349176B (en) | Target tracking method and system based on triple convolutional network and perceptual interference learning | |
CN107093182B (en) | A kind of human height's estimation method based on feature corners | |
CN102446357A (en) | Level set SAR (Synthetic Aperture Radar) image segmentation method based on self-adaptive finite element | |
CN111582198A (en) | Automatic sea-land segmentation method for remote sensing image | |
CN115511759A (en) | Point cloud image depth completion method based on cascade feature interaction | |
CN111709494B (en) | Novel hybrid-optimized image stereo matching method | |
Hamid et al. | Stereo matching algorithm based on hybrid convolutional neural network and directional intensity difference | |
CN112857379A (en) | Improved Gmapping-SLAM map updating method and system | |
CN108564564A (en) | Based on the medical image cutting method for improving fuzzy connectedness and more seed points | |
CN115932913B (en) | Satellite positioning pseudo-range correction method and device | |
CN117036689A (en) | Left ventricle indoor membrane image segmentation method and system based on deep reinforcement learning |
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 |