CN109543568A - A kind of vehicle-logo location method - Google Patents

Abstract

The application belongs to technical field of image processing, more particularly to a kind of vehicle-logo location method.Existing vehicle-logo location method uses conventional method mostly, and positioning rate and robustness are all poor.The application provides a kind of vehicle-logo location method, and described method includes following steps: step 1): obtaining face image before vehicle；Step 2): image primary features are extracted；Step 3): fusion is calculated using multiple dimensioned spectral residuum method and obtains final notable figure, notable figure is divided into multiple sub- marking areas, then forms multiple focus-of-attentions；Step 4): the complexity in each region is calculated；Step 5): optimize focus transfer path using ant group algorithm is improved；Step 6): determine car mark region.The application vehicle-logo location algorithm has better positioning rate and robustness.

Description

Car logo positioning method

Technical Field

The application belongs to the technical field of image processing, and particularly relates to a car logo positioning method.

Background

The car logo refers to marks of various car brands, is mainly used for vehicle identification, and the marks are often representatives of car enterprises. The automobile mark includes a trademark or a factory mark of the automobile, a product label, an engine model and a factory number, a whole automobile model and a factory number, a vehicle identification code and the like. According to national regulations, new vehicles need to be checked during registration and annual inspection. The vehicle should have its own logo to indicate the manufacturer of the vehicle, the type of vehicle, the engine power, the load bearing quality, the engine and the factory number of the entire vehicle, etc. They function to facilitate the identification of the "identity" of the vehicle by the seller, user, maintenance personnel, traffic management.

The vehicle logo is an important attribute of the vehicle, is different from a license plate, is difficult to replace and alter, and is widely applied to the fields of vehicle theft and robbery investigation, automatic recording of illegal vehicles, unmanned management of parking lots, automatic charging at bridge intersections and the like. However, most of the existing car logo positioning methods use the traditional method, and the positioning rate and the robustness are poor.

Disclosure of Invention

1. Technical problem to be solved

The automobile logo is an important attribute of the automobile, is different from a license plate, is difficult to replace and alter, and is widely applied to the fields of vehicle stealing and robbing investigation, automatic recording of illegal vehicles, unmanned management of parking lots, automatic charging at bridge intersections and the like. However, most of the existing car logo positioning methods use traditional methods, and the positioning rate and robustness are poor.

2. Technical scheme

In order to achieve the above object, the present application provides a car logo positioning method, including the steps of:

step 1): acquiring a front face image of a vehicle;

step 2): extracting primary features of the image;

step 3): calculating and fusing by adopting a multi-scale spectrum residual method to obtain a final saliency map, and dividing the saliency map into a plurality of sub saliency areas to form a plurality of attention focuses;

step 4): calculating the complexity of each region;

step 5): optimizing a focus transfer path by adopting an improved ant colony algorithm;

step 6): and judging the car logo area.

Optionally, the primary feature in step 2) includes one or more of color, brightness, or direction.

Optionally, the color and brightness feature extraction is to perform calculation on the original image through a nonlinear anisotropic diffusion equation to extract a saliency map about the color feature and the brightness feature.

Optionally, the color and brightness feature extraction method is as follows:

the gaussian filter function G (x, y, σ) is modified to a regularized nonlinear anisotropic diffusion equation, P-M equation:

the scale σ is 1/2, 1/4 of the original image, respectively; the method comprises the steps of performing sub-sampling and low-pass filtering on an original input image I (x, y) step by step under different scale filters, simultaneously combining the filtering with the edge detection of the image by using the gradient module value of the image, and changing the diffusion coefficient according to the information of the imageAnd the diffusion coefficient reaches a minimum value at the edge of the image according to the gradient of the image obtained by each iterationThe size of the image is subjected to edge judgment, and then a regularized P-M equation is used for obtaining an image subjected to nonlinear diffusion filtering; solution I calculated in the above equation_σAnd (x, y, t) is the image after regularization filtering.

Optionally, the directional feature extraction is to extract a feature with a sensitive orientation through a Gabor filter, and convert the original image into an image with the sensitive orientation;

the function of the Gabor filter is expressed as:

wherein x ', y' are respectively:

x′＝xcosθ+ysinθ

y′＝-xsinθ+ycosθ

where the parameter θ is the orientation of the Gabor filter, f₀Is the center frequency, σ_xAnd σ_yGaussian function variances in the spatial domain x 'and y' directions, respectively; using 4 different orientations (theta e [0 deg. ], 45 deg., 90 deg., 135 deg. ]]) Filtering the original input image I (x, y) to form 4 orientation feature maps { R }_k(x, y), k is 1, 2, 3, 4}, and then normalized to an orientation feature saliency map; which is in (x)₀，y₀) The output of (c) is represented as:

R(xy)＝l(x，y)*h(x-x₀y-)y₀)

in the formula, x₀，y₀The position corresponding to the center of the field represents the convolution operation.

Optionally, the method for obtaining the final saliency map by computing and fusing the multi-scale spectrum residual error method includes:

the process of calculating the significance under multiple scales is as follows:

wherein,the input image is I (x), Fourier transform is carried out on the input image to obtain an amplitude spectrum A (f) and a phase spectrum P (f); h is a convolution kernel with 3 × 3 mean filtering, and R (f) is a frequency domain residual spectrum; s (x) is a saliency region map;

carrying out gray scale on the same feature saliency map under different scalesPerforming supplementary adjustment to the original image in the same size and scale, and performing weighted fusion to obtain final feature saliency maps; since the saliency map is a grayscale map, for a scale ofSupplementing the rest pixels of each feature saliency map into gray values of 0; the calculation formula of the color and brightness feature saliency map for each scale is as follows:

different feature saliency maps, namely color, brightness and direction feature saliency maps are normalized and merged to obtain a final global saliency map A:

wherein S₃The weight gamma is a direction feature saliency map obtained by Gabor filtering fusion₁+γ₂+γ₃1, take γ₁＝γ₂＝γ₃＝1/3。

Optionally, the complexity of each region is calculated as three complexity indexes, i.e., a comprehensive analysis quality symmetry, an image composition complexity and a shape complexity, and linear weighted fusion is performed to obtain each significant subregion a₁、A₂...A_j...A_nRegion complexity C of₁、C₂...C_j...C_n；

The region complexity algebraic representation is:

C_j＝λQ_dj+κC′_dj+μE_j

wherein λ, κ, μ are normalization coefficients.

Optionally, the improved ant colony algorithm is that through traversal of multiple focuses, the reliability of the car logo of the focus is judged according to the area complexity and the edge information, the area complexity is adopted to drive ants to preferentially access points with larger complexity, the convergence speed is accelerated, meanwhile, the ant error rate is introduced, the ants are enabled to walk to a target with less pheromones according to a certain probability, and the algorithm is prevented from falling into a local optimal solution;

m ants are arranged, the next target is selected according to a probability function taking the target distance as a heuristic factor and the quantity of pheromones on the path as a variable, and an ant taboo table tabu is established_kAdding the focus visited by the ants into a taboo list; the first ant is set as a miss ant and only goes to the target with low pheromone, so the probability that the kth ant transfers from the target i to the target j is as follows:

wherein C is_jFor the region complexity of the next target j, ω is the importance coefficient of the region complexity, allowed_k∈({1，2，...n}-tabu_k) Goal to allow ants to choose, η_ij＝1/d_0jAs a heuristic function of the path (i, j), d_0jLength of path from origin to target j, τ_ijα represents the importance coefficient of the path, β represents the relative balance coefficient of the heuristic factor, and rho is the pheromone persistence factor, namely the pheromone existence intensity;

at the initial time, the number of information elements on each path is the same, i.e. τ_ij(0) And C is a constant, and the pheromone content on each path is updated by the ants each time the ants complete the search. The pheromone increment on a target path with large complexity in a specified area is additionally enhanced, and an importance factor is introducedWhen the complexity of the next transferred target is larger than that of the previous target, the pheromone increment is larger, otherwise, the pheromone increment is inhibited to a certain extent, and the ant surrounding model is adopted to update the global informationThe pheromone increment and pheromone updating mode is as follows:

the kth ant passes through the cycle (ij)

Where Δ τ is_ijIndicating pheromone increment on the path ij in the current cycle; l is_kThe path length of the kth ant in the circumcircle is shown, and Q is the information intensity and represents the total amount of pheromone released on the path by the ant in the circumcircle.

Optionally, the optimized focus transition path is formed by numbering the attention focuses from 1 to n; initializing parameters, randomly placing m ants on n vertexes, tau_ij(t) initialization to Δ τ_ij(0) C, initialization pheromone increment Δ τ_ij(t)，N_c←0(N_cIteration times), setting an upper limit of the iteration times; tabu with ant k_kPlacing the initial starting point of each ant in a current taboo list; calculating the probability of transferring each ant to the next target j, moving the ants, and adding the target point j into a taboo list; calculating the path length L of each ant_k(k 1, 2, 3.. m) and pheromone increments on path (i, j), modifying pheromone intensities on path (i, j) according to a pheromone update equation; for each path (i, j), a [ Delta ] tau is set_ijNo. 0, nc ← nc + 1; if N is present_cWhen the preset iteration times are reached or no more optimal solution appears, the loop is exited, otherwise, the taboo table tabu of the ant k is set_kPlacing the initial starting point of each ant in a current taboo list; and outputting the best solution at present, namely the optimal path U.

Optionally, the credibility of the car logo area determined as the sub-area is defined as:

wherein the constraint condition is the focus transfer path U optimized by the ant colony, the credibility of the sub-regions is judged one by one according to the focus transfer path U, and when the CRE of a certain region_UDetermining the area as a car logo area and terminating the judgment; wherein P is_CiFor regional complexity, P_EiAs boundary ratio, unity_iThe vehicle logo in the front face image of the vehicle is unity.

3. Advantageous effects

Compared with the prior art, the beneficial effects of the car logo positioning method provided by the application are that:

according to the method for positioning the car logo by combining the improved ant colony algorithm visual attention mechanism, a new index of 'quality symmetry' is provided by utilizing the center symmetry characteristic of the center of mass of the car logo, the area complexity is measured by utilizing three indexes of the quality symmetry, the image composition complexity and the shape complexity, the improved ant colony algorithm is provided, the area complexity and the ant error rate are introduced into the ant colony algorithm, the algorithm convergence speed is accelerated, and the influence of the possibility of falling into local optimum is relieved; compared with the traditional vehicle logo positioning algorithm, the vehicle logo positioning rate of the vehicle logo positioning algorithm provided by the application can reach 98.43% at most; the method is superior to the traditional symmetry detection vehicle logo positioning, and can also meet the real-time requirement; the stability and robustness of the algorithm are superior to the other two algorithms. The car logo positioning algorithm has better positioning rate and robustness.

Drawings

Fig. 1 is a schematic flow chart of a car logo positioning method according to the present application.

Detailed Description

Hereinafter, specific embodiments of the present application will be described in detail with reference to the accompanying drawings, and it will be apparent to those skilled in the art from this detailed description that the present application can be practiced. Features from different embodiments may be combined to yield new embodiments, or certain features may be substituted for certain embodiments to yield yet further preferred embodiments, without departing from the principles of the present application.

The Gaussian pyramid model based on significance proposed by itti-koch adopts a central peripheral difference calculation method to obtain significance measurement on the basis of non-uniform sampling, but the calculation method only considers the local characteristics of a significant region and does not consider the global information of the whole image.

The object-based computational model proposed by roche et al using multi-scale analysis and grouping extracts edges using differential operators, and contour grouping derived from the perceptual organization of the lattice towers organizes the edges into perceptual objects, but the model is computationally more efficient but the main clue for grouping is closed, thus limiting the scope of application of the model.

The method for detecting the spectrum residual significant region provided by H Xiaodi based on frequency domain signal analysis adopts an information theory information forming mode to filter redundant information from a frequency domain to obtain useful information, but the effect of calculating the significance is obvious and the real-time performance is strong, but the method only calculates the significance under a single scale and cannot well describe other local characteristics of an image, such as color characteristics and the like.

Radharkrishna et al propose that a saliency value is directly defined by a color difference between a pixel and an average color of an entire image, have the advantages of good real-time performance, simplicity and quickness, but have an undesirable effect on image calculation with a complex background. In the aspect of focusing attention, the currently used hierarchical search mechanism from the coarse scale to the fine scale is not efficient and has poor real-time performance.

In machine vision, Saliency is a pattern of image partitions, and Saliency map (English: Saliency map) is an image showing the uniqueness of each pixel. The goal of the saliency map is to simplify or change the representation of the generic image into a more easily analyzable style. For example, a pixel has a higher gray level in a color map, which is displayed in a more obvious manner in the saliency map. From the viewpoint of visual stimulation, if certain features are particularly captured, the characteristic of this is referred to as significance (saliency) psychologically.

The ant colony Algorithm (AG) is a simulated optimization algorithm simulating ant foraging behavior, which was first proposed by the italian scholaro M et al in 1991 and was first used to solve TSP (traveling salesman problem).

The basic principle of the ant colony algorithm is as follows:

1. ants release pheromones on the path.

2. And randomly selecting a path to walk when the crossing which is not walked is touched. At the same time, the pheromone associated with the path length is released.

3. Pheromone concentration is inversely proportional to the path length. When the subsequent ants touch the intersection again, the path with higher pheromone concentration is selected.

4. The pheromone concentration on the optimal path is increasing.

5. And finally finding the optimal food searching path by the ant colony.

Termite round model (Ant-Cycle model)

Ant Quantity model (Ant-Quantity model)

Ant Density model (Ant-sensitivity model)

The difference is as follows:

1. the ant surrounding model utilizes global information, namely, the ants update pheromones on all paths after completing a cycle;

2. the ant quantity and density model utilizes local information, namely, the ants update pheromones on a path after completing one step.

Traversal (Traversal) refers to making one and only one access to each node in the tree in sequence along a search route. The operation of the access node depends on the particular application. Traversal is one of the most important operations in the binary tree, and is the basis for performing other operations in the binary tree. Of course the concept of traversal is also applicable to the case of multi-element sets, such as arrays.

Gabor is a linear filter for edge extraction, which has frequency and direction expression similar to the human visual system, provides good direction selection and scale selection characteristics, and is insensitive to illumination variation, thus being well suited for texture analysis.

Referring to fig. 1, the present application provides a car logo positioning method, including the steps of:

step 1): acquiring a front face image of a vehicle;

step 2): extracting primary features of the image;

step 3): calculating and fusing by adopting a multi-scale spectrum residual method to obtain a final saliency map, and dividing the saliency map into a plurality of sub saliency areas to form a plurality of attention focuses;

step 4): calculating the complexity of each region;

step 5): optimizing a focus transfer path by adopting an improved ant colony algorithm;

step 6): and judging the car logo area.

Optionally, the primary feature in step 2) includes one or more of color, brightness, or direction.

Optionally, the color and brightness feature extraction is to perform calculation on the original image through a nonlinear anisotropic diffusion equation to extract a saliency map about the color feature and the brightness feature.

Optionally, the color and brightness feature extraction method is as follows:

the gaussian filter function G (x, y, σ) is modified to a regularized nonlinear anisotropic diffusion equation, P-M equation:

the scale σ is 1/2, 1/4 of the original image, respectively; the method comprises the steps of performing sub-sampling and low-pass filtering on an original input image I (x, y) step by step under different scale filters, simultaneously combining the filtering with the edge detection of the image by using the gradient module value of the image, and changing the diffusion coefficient according to the information of the imageAnd the diffusion coefficient reaches a minimum value at the edge of the image according to the gradient of the image obtained by each iterationThe size of the image is subjected to edge judgment, and then a regularized P-M equation is used for obtaining an image subjected to nonlinear diffusion filtering; solution I calculated in the above equation_σAnd (x, y, t) is the image after regularization filtering. Through improved gaussian filtering, color features and brightness features under 3 scales σ (original image, 1/2 of original image, 1/4 of original image) are obtained, and 6 saliency maps are obtained in total.

Optionally, the directional feature extraction is to extract a feature with a sensitive orientation through a Gabor filter, and convert the original image into an image with the sensitive orientation;

the function of the Gabor filter is expressed as:

wherein x ', y' are respectively:

x′＝x cosa+y sinθ

y′＝-x sinθ+y cosθ

where the parameter θ is the orientation of the Gabor filter, f₀Is the center frequency, σ_xAnd σ_yGaussian function variances in the spatial domain x 'and y' directions, respectively; using 4 different orientations (theta e [0 deg. ], 45 deg., 90 deg., 135 deg. ]]) Filtering the original input image I (x, y) to form 4 orientation feature maps { R }_k(x, y), k is 1, 2, 3, 4}, and then normalized to an orientation feature saliency map; which is in (x)₀，y₀) The output of (c) is represented as:

R(x，y)＝I(x，y)*h(x-x₀，y-y₀)

in the formula, x₀,y₀The position corresponding to the center of the field represents the convolution operation.

Optionally, the method for obtaining the final saliency map by computing and fusing the multi-scale spectrum residual error method includes:

the process of calculating the significance under multiple scales is as follows:

wherein,the input image is I (x), Fourier transform is carried out on the input image to obtain an amplitude spectrum A (f) and a phase spectrum P (f); h is a convolution kernel with 3 × 3 mean filtering, and R (f) is a frequency domain residual spectrum; s (x) is a saliency region map; the color, brightness and direction characteristics of 7 saliency maps at each scale can be obtained by the algorithm.

Performing gray level supplement adjustment on the same characteristic saliency map under different scales to the same scale with the original image size, and performing weighted fusion to obtain final characteristic saliency maps; since the saliency map is a grayscale map, for a scale ofThe feature saliency map of (1) complements the remaining pixels thereofCharging to gray value of 0; the calculation formula of the color and brightness feature saliency map for each scale is as follows:

different feature saliency maps, namely color, brightness and direction feature saliency maps are normalized and merged to obtain a final global saliency map A:

wherein S₃The weight gamma is a direction feature saliency map obtained by Gabor filtering fusion₁+γ₂+γ₃1, take γ₁＝γ₂＝γ₃＝1/3。

Optionally, the complexity of each region is calculated as three complexity indexes, i.e., a comprehensive analysis quality symmetry, an image composition complexity and a shape complexity, and linear weighted fusion is performed to obtain each significant subregion a₁、A₂…A_j…A_nRegion complexity C of₁、C₂...C_j...C_n；

The region complexity algebraic representation is:

C_j＝λQ_dj+κC′_dj+μE_j

wherein λ, κ, μ are normalization coefficients.

Obtaining a saliency map A (gray level map) of a vehicle front face picture according to a multi-scale global saliency map calculation method, and obtaining n sub-saliency areas A after segmentation₁、A₂...A_j...A_n. In the subsequent focus transfer, the sub-salient region needs to be identified to distinguish the emblem region. The distinction degree between the car logo area complexity and the background which can be intuitively obtained by human vision is larger, so the calculation area complexity is providedAs one of the indexes for measuring the car logo area.

The method adopts three indexes of quality symmetry, image composition complexity and shape complexity to measure the region complexity.

(1) And (4) quality symmetry degree. Because the center of mass of the car logo is on the axis of the car logo area, the mass difference of the left side and the right side of the central axis of the salient sub-area is calculated, namely the gray value is (0, T)₁) The threshold value 0 < T is set₁≤255、Th_D> 0, define the quality symmetry degree (quality symmetry degree)

WhereinIs the sum of the masses on the left side of the central axis,is the sum of the masses on the right side of the central axis.

(2) Complexity of image composition. According to the information entropy theory proposed by the Zhang study, the composition complexity can be defined by the richness degree of the internal state of the generalized set, and the specific method is as follows: if there are k different flag values x in the set of N individuals₁,x₂,…x_i,…x_kAnd the number of the corresponding individuals is n₁,2_n，…n_i,…n_kThen the complexity of the generalized set can be obtainedThe image composition complexity can be defined by the generalized probability of each gray-value pixel appearing in the salient sub-region, i.e.The unit is a bit (bit). Wherein p is_iIs P (x)_i)＝n_iN, indicating the presence of a pixel having a grey value iProbability. The larger C, the greater the image composition complexity.

(3) The complexity of the shape. The car logo is special in shape and has certain regularity, and the shape complexity not only comprises external contour information, but also comprises a car logo internal topological structure. The inside topology of the car logo is complex and the boundary line is long, so the shape complexity is described by using an edge ratio (edge ratio). Boundary ration_edgeIs the boundary length, i.e. the number of boundary pixels, and N is the total number of pixels in the sub-region. E_jThe larger the shape complexity.

The global saliency map A is obtained through multi-scale saliency calculation, and n saliency sub-regions are obtained through segmentation, namely n attention focuses are obtained. The reliability of the car logo of the focus is judged by traversing the n focuses and according to the area complexity and the edge information, and the problem of path planning can be actually converted. According to the method and the device, the ant colony algorithm guided by the area complexity and the ant error rate is introduced according to the specific characteristics of the front face picture of the vehicle, the convergence speed is improved, and the situation that the ant colony algorithm falls into a local optimal solution is avoided.

The method aims at rapidly and accurately judging the car logo area, and the judgment termination condition is that a certain target meets the area complexity and the edge information. According to the conventional ant colony algorithm, the car logo area may be accessed later in the optimal path, thereby prolonging the convergence time of the whole algorithm. The method and the device utilize the characteristic of larger complexity of the vehicle logo region, drive ants to preferentially visit points with larger complexity by using the region complexity, accelerate the convergence speed, and simultaneously introduce the ant error rate, so that the ants can go to the target with less pheromones according to a certain probability, and the algorithm is prevented from falling into the local optimal solution. The focus transfer mechanism of the present application follows the following 3 principles: 1) forbidding a return principle, i.e. not repeatedly accessing the focus; 2) the complexity priority principle is that the ants are driven to visit the focus with higher complexity first; 3) on the basis of complexity priority, a near sub-optimal principle makes a planning path better follow a human eye focus transfer rule as far as possible.

Optionally, the improved ant colony algorithm is that through traversal of multiple focuses, the reliability of the car logo of the focus is judged according to the area complexity and the edge information, the area complexity is adopted to drive ants to preferentially access points with larger complexity, the convergence speed is accelerated, meanwhile, the ant error rate is introduced, the ants are enabled to walk to a target with less pheromones according to a certain probability, and the algorithm is prevented from falling into a local optimal solution;

setting m ants, selecting a next target according to a probability function taking a target distance as a heuristic factor and taking the number of pheromones on a path as a variable, establishing an ant taboo list tabuk according to an ant selection strategy related to the complexity of a next target area, and adding a focus visited by the ants into the taboo list; the first ant is set as a miss ant and only goes to the target with low pheromone, so the probability that the kth ant transfers from the target i to the target j is as follows:

wherein C is_jFor the region complexity of the next target j, ω is the importance coefficient of the region complexity, allowed_k∈({1，2，...n}-tabu_k) Goal to allow ants to choose, η_ij＝1/d_0jAs a heuristic function of the path (i, j), d_0jLength of path from origin to target j, τ_ijα represents the importance coefficient of the path, β represents the relative balance coefficient of the heuristic factor, and rho is the pheromone persistence factor, namely the pheromone existence intensity;

at the initial time, the number of information elements on each path is the same, i.e. τ_ij(0) And C is a constant, and the pheromone content on each path is updated by the ants each time the ants complete the search. The pheromone increment on a target path with large complexity in a specified area is additionally enhanced, and an importance factor is introducedI.e. the next target complexity of transferWhen the quantity of the pheromones is larger than the previous target, the pheromone increment is larger, otherwise, the pheromone increment is inhibited to a certain extent, and an Ant-cycle (Ant-cycle) model is adopted to update the global information, wherein the pheromone increment and pheromone updating mode is as follows:

the kth ant passes through the cycle (ij)

Where Δ τ is_ijIndicating pheromone increment on the path ij in the current cycle; l is_kThe path length of the kth ant in the circumcircle is shown, and Q is the information intensity and represents the total amount of pheromone released on the path by the ant in the circumcircle.

Optionally, the optimized focus transition path is formed by numbering the attention focuses from 1 to n; initializing parameters, randomly placing m ants on n vertexes, tau_ij(t) initialization to τ_ij(0) C, initialization pheromone increment Δ τ_ij(t)，N_c←0(N_cIteration times), setting an upper limit of the iteration times; tabu with ant k_kPlacing the initial starting point of each ant in a current taboo list; calculating the probability of transferring each ant to the next target j, moving the ants, and adding the target point j into a taboo list; calculating the path length L of each ant_k(k 1, 2, 3.. m) and pheromone increments on path (i, j), modifying pheromone intensities on path (i, j) according to a pheromone update equation; for each path (i, j), a [ Delta ] tau is set_ijNo. 0, nc ← nc + 1; if N is present_cWhen the preset iteration times are reached or no more optimal solution appears, the loop is exited, otherwise, the taboo table tabu of the ant k is set_kPlacing the initial starting point of each ant in a current taboo list; and outputting the best solution at present, namely the optimal path U.

The complexity of the car logo region is high, but the region with high region complexity is not necessarily the car logo region, that is, the region complexity should be taken as a necessary and insufficient condition for judging the car logo region, and on the basis, the possibility that the previously extracted target edge information, that is, the region with high boundary ratio is the car logo region is considered to be high. In addition, since the emblems in the front image of the vehicle have no repeatability, i.e., unity, and the headlight, the fog light, the turn light, and the like appear in pairs, the possibility that the region with unique complexity information is the emblem region is the greatest. Therefore, the area meeting all the three conditions is defined as the vehicle logo area, and one of the areas is not available.

Optionally, the credibility of the car logo area determined as the sub-area is defined as:

wherein the constraint condition is the focus transfer path U optimized by the ant colony, the credibility of the sub-regions is judged one by one according to the focus transfer path U, and when the CRE of a certain region_UDetermining the area as a car logo area and terminating the judgment; wherein P is_CiFor regional complexity, P_EiAs boundary ratio, unity_iThe vehicle logo in the front face image of the vehicle is unity. Therefore, the reliability of the judged vehicle logo area is greatly increased through the layer-by-layer judgment of the three indexes.

(1) Extracting primary visual features of the vehicle front face picture, calculating and fusing by adopting a multi-scale frequency spectrum residual method to obtain a final saliency map, and comprehensively considering local information and global information of the image.

(2) The center symmetry characteristic of the mass center of the car logo is utilized to provide a new index of 'quality symmetry', and three indexes of quality symmetry, image composition complexity and shape complexity are utilized to measure the area complexity.

(3) The improved ant colony algorithm is adopted to optimize a focus transfer path, ants are driven to preferentially access points with higher complexity through the regional complexity, and an adjustment factor of ant error rate is introduced, so that the path searching efficiency is effectively increased, and the situation that the ant falls into a local optimal solution is avoided.

According to the method for positioning the car logo by combining the improved ant colony algorithm visual attention mechanism, a new index of 'quality symmetry' is provided by utilizing the center symmetry characteristic of the center of mass of the car logo, the area complexity is measured by utilizing three indexes of the quality symmetry, the image composition complexity and the shape complexity, the improved ant colony algorithm is provided, the area complexity and the ant error rate are introduced into the ant colony algorithm, the algorithm convergence speed is accelerated, and the influence of the possibility of falling into local optimum is relieved; compared with the traditional vehicle logo positioning algorithm, the vehicle logo positioning rate of the vehicle logo positioning algorithm provided by the application can reach 98.43% at most; the method is superior to the traditional symmetry detection vehicle logo positioning, and can also meet the real-time requirement; the stability and robustness of the algorithm are superior to the other two algorithms. The car logo positioning algorithm has better positioning rate and robustness.

Although the present application has been described above with reference to specific embodiments, those skilled in the art will recognize that many changes may be made in the configuration and details of the present application within the principles and scope of the present application. The scope of protection of the application is determined by the appended claims, and all changes that come within the meaning and range of equivalency of the technical features are intended to be embraced therein.

Claims (10)

1. A car logo positioning method is characterized in that: the method comprises the following steps:

step 1): acquiring a front face image of a vehicle;

step 2): extracting primary features of the image;

step 3): calculating and fusing by adopting a multi-scale spectrum residual method to obtain a final saliency map, and dividing the saliency map into a plurality of sub saliency areas to form a plurality of attention focuses;

step 4): calculating the complexity of each region;

step 5): optimizing a focus transfer path by adopting an improved ant colony algorithm;

step 6): and judging the car logo area.

2. The emblem positioning method of claim 1, characterized in that: the primary features in the step 2) comprise one or more of color, brightness or direction.

3. The emblem positioning method of claim 2, characterized in that: the color and brightness feature extraction is to extract a saliency map about color features and brightness features from an original image through calculation of a nonlinear anisotropic diffusion equation.

4. The emblem positioning method of claim 3, characterized in that: the color and brightness feature extraction method comprises the following steps:

the gaussian filter function G (x, y, σ) is modified to a regularized nonlinear anisotropic diffusion equation, P-M equation:

the scale σ is 1/2, 1/4 of the original image, respectively; the method comprises the steps of performing sub-sampling and low-pass filtering on an original input image I (x, y) step by step under different scale filters, simultaneously combining the filtering with the edge detection of the image by using the gradient module value of the image, and changing the diffusion coefficient according to the information of the imageAnd the diffusion coefficient reaches a minimum value at the edge of the image according to the gradient of the image obtained by each iterationThe size of the image is subjected to edge judgment, and then a regularized P-M equation is used for obtaining an image subjected to nonlinear diffusion filtering; solution I calculated in the above equation_σ(x, y, t) is the processRegularizing the filtered image.

5. The emblem positioning method of claim 2, characterized in that: the directional feature extraction is to extract orientation-sensitive features through a Gabor filter and convert an original image into an image of the orientation features;

the function of the Gabor filter is expressed as:

wherein x ', y' are respectively:

x′＝xcosθ+ysinθ

y′＝-xsin9+ycosθ

where the parameter θ is the orientation of the Gabor filter, f₀Is the center frequency, σ_xAnd σ_yGaussian function variances in the spatial domain x 'and y' directions, respectively; using 4 different orientations (theta e [0 deg. ], 45 deg., 90 deg., 135 deg. ]]) Filtering the original input image I (x, y) to form 4 orientation feature maps { R }_k(x, y), k is 1, 2, 3, 4}, and then normalized to an orientation feature saliency map; which is in (x)₀，y₀) The output of (c) is represented as:

R(x，y)＝I(x，y)*h(x-x₀，y-y₀)

in the formula, x₀,y₀The position corresponding to the center of the field represents the convolution operation.

6. The emblem positioning method of claim 1, characterized in that: the method for obtaining the final saliency map through calculation fusion by the multi-scale spectrum residual error method comprises the following steps:

the process of calculating the significance under multiple scales is as follows:

wherein,the input image is I (x), Fourier transform is carried out on the input image to obtain an amplitude spectrum A (f) and a phase spectrum P (f); h is a convolution kernel with 3 × 3 mean filtering, and R (f) is a frequency domain residual spectrum; s (x) is a saliency region map;

performing gray level supplement adjustment on the same characteristic saliency map under different scales to the same scale with the original image size, and performing weighted fusion to obtain final characteristic saliency maps; since the saliency map is a grayscale map, for a scale ofSupplementing the rest pixels of each feature saliency map into gray values of 0; the calculation formula of the color and brightness feature saliency map for each scale is as follows:

different feature saliency maps, namely color, brightness and direction feature saliency maps are normalized and merged to obtain a final global saliency map A:

wherein S₃The weight gamma is a direction feature saliency map obtained by Gabor filtering fusion₁+γ₂+γ₃1, take γ₁＝γ₂＝γ₃＝1/3。

7. The emblem positioning method of claim 1, characterized in that: the complexity of each region is calculated into three complexity indexes of comprehensive analysis quality symmetry, image composition complexity and shape complexity, and linear weighted fusion is carried out to respectively obtain each significant subregion A₁、A₂...A_j...A_nRegion complexity C of₁、C₂...C_j...C_n；

The region complexity algebraic representation is:

C_j＝λQ_dj+κC′_dj+μE_j

wherein λ, κ, μ are normalization coefficients.

8. The emblem positioning method of claim 1, characterized in that: the improved ant colony algorithm is characterized in that through traversal of a plurality of focuses, the reliability of the car logo of the focus is judged according to the regional complexity and the edge information, the regional complexity is adopted to drive ants to preferentially access points with higher complexity, the convergence speed is accelerated, meanwhile, the ant error rate is introduced, the ants do not walk to targets with more pheromones according to a certain probability, and the algorithm is prevented from falling into a local optimal solution;

m ants are arranged, the next target is selected according to a probability function taking the target distance as a heuristic factor and the quantity of pheromones on the path as a variable, and an ant taboo table tabu is established_kAdding the focus visited by the ants into a taboo list; the first ant is set as a miss ant and only goes to the target with low pheromone, so the probability that the kth ant transfers from the target i to the target j is as follows:

wherein C is_jFor the region complexity of the next target j, ω is the importance coefficient of the region complexity, allowed_k∈({1，2，...n}-tabu_k) Goal to allow ants to choose, η_ij=1/d_0jAs a heuristic function of the path (i, j), d_0jLength of path from origin to target j, τ_ijα represents the importance coefficient of the path, β represents the relative balance coefficient of the heuristic factor, and rho is the pheromone persistence factor, namely the pheromone existence intensity;

at the initial time, the number of information elements on each path is the same, i.e. τ_ij(0) And C is a constant, and the pheromone content on each path is updated by the ants each time the ants complete the search. The pheromone increment on a target path with large complexity in a specified area is additionally enhanced, and an importance factor is introducedWhen the complexity of the next transfer target is larger than that of the previous target, the pheromone increment is larger, otherwise, the pheromone increment is inhibited to a certain extent, the global information is updated by adopting the ant surrounding model, and the pheromone increment and pheromone updating mode is as follows:

the kth ant passes through the cycle (ij)

Where Δ τ is_ijIndicating pheromone increment on the path ij in the current cycle; l is_kThe path length of the kth ant in the circumcircle is shown, and Q is the information intensity and represents the total amount of pheromone released on the path by the ant in the circumcircle.

9. The emblem positioning method of claim 8, characterized in that: the optimized focus transfer path is formed by numbering attention focuses from 1 to n; initializing parameters, randomly placing m ants on n vertexes, tau_ij(t) initialization to τ_ij(0) C, initialization pheromone increment Δ τ_ij(t)，N_c←0(N_cIteration times), setting an upper limit of the iteration times; tabu with ant k_kPlacing the initial starting point of each ant in a current taboo list; calculating the probability of transferring each ant to the next target j, moving the ants, and adding the target point j into a taboo list; calculating the path length L of each ant_k(k 1, 2, 3.. m) and pheromone increments on path (i, j), modifying pheromone intensities on path (i, j) according to a pheromone update equation; for each path (i, j), a [ Delta ] tau is set_ijNo. 0, nc ← nc + 1; if N is present_cWhen the preset iteration times are reached or no more optimal solution appears, the loop is exited, otherwise, the taboo table tabu of the ant k is set_kPlacing the initial starting point of each ant in a current taboo list; and outputting the best solution at present, namely the optimal path U.

10. The emblem positioning method of claim 1, characterized in that: the credibility of the car logo area judged as the sub-area is defined as:

wherein the constraint condition is the focus transfer path U optimized by the ant colony, the credibility of the sub-regions is judged one by one according to the focus transfer path U, and when the CRE of a certain region_UDetermining the area as a car logo area and terminating the judgment; wherein P is_CiFor regional complexity, P_EiAs boundary ratio, unity_iThe vehicle logo in the front face image of the vehicle is unity.