CN104915959A - Aerial photography image quality evaluation method and system - Google Patents

Abstract

The invention provides an aerial photography image quality evaluation method. The method includes the following steps that: a power line inspection corridor channel image library is constructed; the spatial domain features of each natural image in the image library are extracted; MVG model fitting is performed on each spatial domain feature, so that standard fitting parameters characterizing image quality can be obtained; the standard fitting parameters are compared with fitting parameters of an aerial photography image to be tested, so that the difference value of the standard fitting parameters and the fitting parameters of the aerial photography image to be tested can be obtained; and quality scoring is performed on the aerial photography image to be tested according to the difference value, so that quality evaluation on the aerial photography image to be tested can be realized. According to the aerial photography image quality evaluation method, the power line inspection corridor channel image library contains a plurality of original natural power line inspection images; the spatial domain features of the natural images are natural attributes of the images, and the attributes are unrelated to the content of the images, and when the images are degraded, the natural attributes will be destroyed, and fitting quantization is performed on the spatial domain features, and quality evaluation on the aerial photography images can be realized through comparing the degradation degree of the quantized natural attributes.

Description

A kind of quality evaluating method of Aerial Images and system

Technical field

The application relates to technical field of image processing, particularly relates to a kind of quality evaluating method and system of Aerial Images.

Background technology

Photography of taking photo by plane is a kind of remote, non-contacting target detection techniques and methods, by detecting target, obtains target information, then to the information processing obtained, thus realizes the location of target, qualitative or quantitative description.

Camera work of taking photo by plane more application, in power-line patrolling field, to be taken photo by plane power-line patrolling image by unmanned plane, to determine whether power-line patrolling is in safe running status.That applies in power-line patrolling field along with unmanned plane technology of taking photo by plane constantly improves and development, and the quality assessment problem of transmission line of electricity Aerial Images is also more and more subject to the attention of the worker of taking photo by plane.Inventor finds through practical studies repeatedly, and Aerial Images owing to not having reference picture, thus does not have reference frame to the quality assessment of Aerial Images, and the quality of quality evaluating method to Aerial Images therefore needing a kind of Aerial Images badly is evaluated.

Summary of the invention

Technical problems to be solved in this application are to provide a kind of quality evaluating method and system of Aerial Images, for evaluating the quality of Aerial Images.

A quality evaluating method for Aerial Images, comprising:

Build power-line patrolling corridor image library;

Extract the space domain characteristic of each natural image in described power-line patrolling corridor image library;

MVG models fitting is carried out to each space domain characteristic extracted, obtains the standard fit parameter of token image quality;

The fitting parameter of described standard fit parameter and acquired Aerial Images to be tested is compared, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested;

According to described difference value, quality score is carried out to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.

Above-mentioned method, preferably, described structure power-line patrolling corridor image library comprises:

According to the Selecting All Parameters preset, choose the power-line patrolling corridor image comprising power-line patrolling target; Described Selecting All Parameters comprises image background, image angle and access time; Described power-line patrolling corridor image is natural image;

Carry out sharpness computation to each natural image chosen respectively, screening sharpness meets the natural image of preset requirement;

Natural image sharpness being met to preset requirement goes mean normalization MSCN coefficient Gaussian distribution to add up, and screening meets the natural image of Gaussian distribution;

To the natural image meeting Gaussian distribution, carry out acutance piecemeal and carefully select, composition power-line patrolling corridor image library.

Above-mentioned method, preferably, described extraction space domain characteristic comprises:

Matching is carried out in the MSCN coefficients statistics distribution of generalized Gaussian distribution to each natural image of application zero-mean, applies asymmetrical generalized Gaussian distribution simultaneously and carries out parameter fitting to point to parameter, set up the characteristic parameter model of natural image;

Using the space domain characteristic of the characteristic parameter of each natural image in characteristic parameter model as natural image.

Above-mentioned method, preferably, describedly carries out MVG models fitting to each space domain characteristic, and the standard fit parameter obtaining token image quality comprises:

Using multivariate Gauss model carries out matching to the characteristic parameter of each natural image;

Using the estimated parameter mean vector of application model and the covariance matrix standard fit parameter as token image quality.

Above-mentioned method, preferably, described according to difference value, quality score is carried out to described Aerial Images to be tested and comprises:

Determine the correspondence position of described difference value in default scoring interval;

According to the quality score of described corresponding position, described Aerial Images to be tested is evaluated.

Above-mentioned method, preferably, described each natural image to choosing carries out sharpness computation, and the natural image that screening sharpness meets preset requirement comprises:

Respectively the reference picture of the described natural image that the natural image chosen and employing low-pass filter construct is converted by default transformation rule, obtain described natural image and described reference picture subimage separately;

Structural similarity SSIM index is adopted to measure corresponding subimage; The SSIM weighted sum of each subimage is obtained the sharpness of described natural image.

Above-mentioned method, preferably, describedly go mean normalization MSCN coefficient Gaussian distribution to add up to the natural image that sharpness meets preset requirement, the natural image that screening meets Gaussian distribution comprises:

Natural image sharpness being met to preset requirement carries out image operation, obtains MSCN coefficient;

Judge whether described MSCN coefficient meets Gaussian distribution;

Screening MSCN coefficient meets the natural image of Gaussian distribution.

Above-mentioned method, preferably, to the natural image meeting Gaussian distribution, carries out acutance piecemeal and carefully selects, and composition power-line patrolling corridor image library comprises:

Acutance piecemeal is carried out to the natural image meeting Gaussian distribution;

Calculate the acutance of each image block, image block acutance being greater than default acutance threshold value adds power-line patrolling corridor image library.

A QA system for Aerial Images, comprising:

Construction unit, for building power-line patrolling corridor image library;

Extraction unit, for extracting the space domain characteristic of each natural image in described power-line patrolling corridor image library;

Fitting unit, for carrying out MVG models fitting to each space domain characteristic extracted, obtains the standard fit parameter of token image quality;

Comparing unit, for being compared by the fitting parameter of described standard fit parameter and acquired Aerial Images to be tested, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested;

Evaluation unit, for according to described difference value, carries out quality score to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.

Above-mentioned system, preferably, described construction unit comprises:

Choose subelement, for according to the Selecting All Parameters preset, choose the power-line patrolling corridor image comprising power-line patrolling target; Described Selecting All Parameters comprises image background, image angle and access time; Described power-line patrolling corridor image is natural image;

Sharpness computation subelement, for carrying out sharpness computation to each natural image chosen respectively, screening sharpness meets the natural image of preset requirement;

Gaussian distribution statistics subelement, the natural image for meeting preset requirement to sharpness goes mean normalization MSCN coefficient Gaussian distribution to add up, and screening meets the natural image of Gaussian distribution;

Carefully select subelement, for the natural image meeting Gaussian distribution, carry out acutance piecemeal and carefully select, composition power-line patrolling corridor image library.

There is provided a kind of quality evaluating method of Aerial Images in the embodiment of the present application, described method comprises: build power-line patrolling corridor image library; Extract the space domain characteristic of each natural image in described power-line patrolling corridor image library; MVG models fitting is carried out to each space domain characteristic extracted, obtains the standard fit parameter of token image quality; The fitting parameter of described standard fit parameter and acquired Aerial Images to be tested is compared, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested; According to described difference value, quality score is carried out to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.The quality evaluating method that the embodiment of the present application provides, first builds power-line patrolling corridor image library, comprises multiple primitive nature power-line patrolling image in described power-line patrolling corridor image library; The space domain characteristic of described natural image is the natural quality of image, these attributes and picture material have nothing to do, during image degradation, natural quality can destroy, matching quantification is carried out to space domain characteristic, by comparing the degree of degeneration of the natural quality after quantification, to reach the object to Aerial Images quality assessment.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present application or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only some embodiments of the application, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.

The method flow diagram of the quality evaluating method of a kind of Aerial Images that Fig. 1 provides for the application;

The another method flow diagram of the quality evaluating method of a kind of Aerial Images that Fig. 2 provides for the application;

The another method flow diagram of the quality evaluating method of a kind of Aerial Images that Fig. 3 provides for the application;

The structural representation of the QA system of a kind of Aerial Images that Fig. 4 provides for the application;

The another structural representation of the QA system of a kind of Aerial Images that Fig. 5 provides for the application.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present application, be clearly and completely described the technical scheme in the embodiment of the present application, obviously, described embodiment is only some embodiments of the present application, instead of whole embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not making the every other embodiment obtained under creative work prerequisite, all belong to the scope of the application's protection.

This application provides a kind of quality evaluating method of Aerial Images, for evaluating the shooting quality of Aerial Images, this method is according to natural scene statistical theory, and the executive agent of the method can be processor, and its method flow diagram as shown in Figure 1, comprising:

Step S101: build power-line patrolling corridor image library;

Build power-line patrolling corridor image library, described image library comprises the primitive nature image of multiple power-line patrolling; Each natural image has original natural quality, and these natural qualities are specifically as follows the space domain characteristic of natural image.

Step S102: the space domain characteristic extracting each natural image in described power-line patrolling corridor image library;

Extract the space domain characteristic of each natural image respectively, the particular content of these space domain characteristic and natural image has nothing to do, and only can change when natural image is subject to degenerating or lose.

Step S103: carry out MVG models fitting to each space domain characteristic extracted, obtains the standard fit parameter of token image quality;

Models fitting is carried out to each space domain characteristic extracted, sets up space domain characteristic model, the quantification of implementation model coefficient, namely obtain the standard fit parameter of token image quality.

Step S104: the fitting parameter of described standard fit parameter and acquired Aerial Images to be tested is compared, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested;

Aerial Images can be degenerated relative to primitive nature image, and the fitting parameter of Aerial Images can there are differences relative to standard fit parameter, obtains the difference value between the fitting parameter of Aerial Images and standard fit parameter;

Step S105: according to described difference value, carries out quality score to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.

The corresponding quality score of described difference value, using this quality score as the Appreciation gist to Aerial Images.

The quality evaluating method of the Aerial Images that the embodiment of the present application provides, according to natural scene statistical theory, natural scene statistical nature catches the bottom characteristic of natural image, image some natural qualities before degeneration are provided, these attributes and picture material have nothing to do, by measuring the destruction of image natural quality of degenerating and causing, can the degree of degeneration of effectively evaluating image.

The quality evaluating method that the embodiment of the present application provides, in essence, catch the statistical law of natural image, and quantize when image is subject to degenerating, these rules how to change or lose.In the application, on spatial domain, go that mean normalization coefficient (MSCN) is approximate meets Gaussian distribution from primitive nature image, and the angle that this distribution does not affect this characteristic by picture material is set out, in conjunction with this specific industry of unmanned plane power-line patrolling, set up power-line patrolling corridor image library targetedly.The aspect of model parameter of power-line patrolling corridor image library is added up by generalized Gaussian distribution matching and asymmetric generalized Gaussian distribution matching, and the characteristic parameter of the testing image obtained of taking photo by plane, adopt multivariate Gaussian model carry out the quantification of two stack features parameters and carry out comparison in difference, realize unmanned plane and to take photo by plane power-line patrolling image quality evaluation.

In the embodiment of the present application, build the method flow diagram of power-line patrolling corridor image library, as shown in Figure 2, comprising:

Step S201: according to the Selecting All Parameters preset, choose the power-line patrolling corridor image comprising power-line patrolling target; Described Selecting All Parameters comprises image background, image angle and access time; Described power-line patrolling corridor image is natural image;

In the embodiment of the present application, according to the feature of unmanned plane power-line patrolling, tentatively choose different background, different angles, different time and comprise the power-line patrolling corridor image of electric power target.

Step S202: respectively sharpness computation is carried out to each natural image chosen, screening sharpness meets the natural image of preset requirement;

In the embodiment of the present application, adopt the sharpness evaluation method based on CSF, articulation index calculating is carried out to the electric power target image of primary election, chooses the high quality graphic that sharpness is high.

Step S203: natural image sharpness being met to preset requirement goes mean normalization MSCN coefficient Gaussian distribution to add up, screening meets the natural image of Gaussian distribution;

Step S204: to the natural image meeting Gaussian distribution, carries out acutance piecemeal and carefully selects, composition power-line patrolling corridor image library.

In the embodiment of the present application, in conjunction with human eye area-of-interest, and facilitate data to store, choose the region unit of sharp keen conversion sensitivity in image stored in image library.

In the embodiment of the present application, high quality power line walking corridor image library effectively can react some common feature of unmanned plane power-line patrolling field Aerial Images; And for Aerial Images data volume large problem, by carrying out the extraction of sharp image block to human eye area-of-interest, can effectively in conjunction with people subjective factor and facilitate data to store.

Further, adopt the non-mean normalization coefficient of zero-mean generalized Gaussian distribution fitted figure picture, adopt non-generalized Gaussian distribution matching can react the point of adjacent image information to parametric data, make characteristic parameter can react image attributes comprehensively.

In the embodiment of the present application, whole algorithm flow designs for unmanned plane power-line patrolling field, the data that can obtain taking photo by plane when non-reference picture and other prior imformations carry out quality assessment real-time, provide good basis to the target localization in following needs line walking, shaft tower detection, defect recognition etc.

The evaluation method that the embodiment of the present application provides, utilize natural scene statistical property, without the need to carrying out artificial rating database study, also without the need to obtaining image degradation type in advance, directly simulate fixing characteristic parameter model, for the characteristic parameter to Aerial Images to be tested by power-line patrolling corridor image library.

In the embodiment of the present application, the process extracting space domain characteristic comprises the steps:

Matching is carried out in the MSCN coefficients statistics distribution of generalized Gaussian distribution to each natural image of application zero-mean, applies asymmetrical generalized Gaussian distribution simultaneously and carries out parameter fitting to point to parameter, set up the characteristic parameter model of natural image;

Using the space domain characteristic of the characteristic parameter of each natural image in characteristic parameter model as natural image.

Above-mentioned steps is in concrete implementation procedure, according to spatial domain natural scene statistical property, the MSCN coefficients statistics distribution of primitive nature image and quality degradation image has obvious difference, in order to this difference difference be quantized, matching is carried out in the MSCN coefficients statistics distribution of generalized Gaussian distribution to image of application zero-mean.

Zero-mean generalized Gaussian distribution is defined as follows:

$f(x,\mathrm{\α},{\mathrm{\σ}}^2)=\frac{\mathrm{\α}}{2\mathrm{\β\Γ}(1/\mathrm{\α})}\exp (-{\left(\frac{\left|x\right|}{\mathrm{\β}}\right)}^{\mathrm{\α}})$

In formula, wherein Γ () operation definition is as follows

$\mathrm{\Γ}\left(\mathrm{\α}\right)={\∫}_0^{\∞}{t}^{\mathrm{\α}-1}{e}^{-t}\mathrm{dt\α}>0$

In order to response diagram picture MSCN coefficient adjacent coefficient between point to direction relations, utilize the directive point of tool to the qualitative character of the more comprehensive Description Image of parameter, the variation relation between adjacent coefficient also brought in the qualitative character parameter of image.Use asymmetrical generalized Gaussian distribution to carry out parameter fitting to point to parameter, asymmetrical generalized Gaussian distribution not only has the parameter controlling distribution global shape, also has the controling parameters controlling left and right tail shape.Concrete matching is as follows:

$f(x,v,{\mathrm{\&sigma;}}_l^2,{\mathrm{\&sigma;}}_r^2)=\left\{\begin{array}{c}\frac{v}{({\mathrm{\&beta;}}_l+{\mathrm{\&beta;}}_r)\mathrm{\&Gamma;}(1/v)}\exp (-{\left(\frac{-x}{\mathrm{\&beta;}}\right)}^v),x<0\\ \frac{v}{({\mathrm{\&beta;}}_1+{\mathrm{\&beta;}}_r)\mathrm{\&Gamma;}(1/v)}\exp (-{\left(\frac{x}{\mathrm{\&beta;}}\right)}^v),x\&GreaterEqual;0\end{array}\right.$

In the embodiment of the present application, carry out MVG models fitting to each space domain characteristic, the standard fit parameter obtaining token image quality comprises the steps:

Using multivariate Gauss model carries out matching to the characteristic parameter of each natural image;

Using the estimated parameter mean vector of application model and the covariance matrix standard fit parameter as token image quality.

In the embodiment of the present application, in order to quantize the difference degree between Aerial Images characteristic parameter to be tested and the physical feature parameter of image library, the Image quality measures parameter that the matching of using multivariate Gauss model is numerous, then the estimated parameter mean vector of application model and covariance matrix are as the parameter participating in picture quality quantization operations.

Multivariate Gaussian model defines:

$f_X(x_1,...,x_k)=\frac{1}{{\left(2\mathrm{\&pi;}\right)}^{k/2}{\left|\mathrm{\&Sigma;}\right|}^{1/2}}\exp (-\frac{1}{2}{(x-v)}^T{\mathrm{\&Sigma;}}^{-1}(x-v))$

By above-mentioned MVG models fitting, the qualitative character parameter of image is just converted to estimated parameter v and Σ of MVG, and these two parameters represent the quality information of image.

The quality quantification formula of final image is:

$D(v_1,v_2,{\mathrm{\&Sigma;}}_1,{\mathrm{\&Sigma;}}_2)=\sqrt{\left({(v_1-v_2)}^T{\left(\frac{{\mathrm{\&Sigma;}}_1+{\mathrm{\&Sigma;}}_2}{2}\right)}^{-1}(v_1-v_2)\right)}$

In the embodiment of the present application, described according to difference value, quality score is carried out to described Aerial Images to be tested and comprises:

Determine the correspondence position of described difference value in default scoring interval;

According to the quality score of described corresponding position, described Aerial Images to be tested is evaluated.

In the embodiment of the present application, described each natural image to choosing carries out sharpness computation, and the natural image that screening sharpness meets preset requirement comprises:

Respectively the reference picture of the described natural image that the natural image chosen and employing low-pass filter construct is converted by default transformation rule, obtain described natural image and described reference picture subimage separately;

Structural similarity SSIM index is adopted to measure corresponding subimage; The SSIM weighted sum of each subimage is obtained the sharpness of described natural image.

In the embodiment of the present application, respectively dct transform is carried out to the reference picture of the original image that the raw power line walking corridor image selected and employing low-pass filter construct; By conversion after high frequency and intermediate frequency coefficient be respectively divided into low high two grades carry out inverse DCT conversion obtain respective subimage; Adopt structural similarity (SSIM) index to measure corresponding subimage again, the SSIM weighted sum of each subimage is obtained the sharpness of image, filter out the high power-line patrolling corridor image of sharpness and carry out next step screening.

In the embodiment of the present application, describedly go mean normalization MSCN coefficient Gaussian distribution to add up to the natural image that sharpness meets preset requirement, the natural image that screening meets Gaussian distribution comprises:

Natural image sharpness being met to preset requirement carries out image operation, obtains MSCN coefficient;

Judge whether described MSCN coefficient meets Gaussian distribution;

Screening MSCN coefficient meets the natural image of Gaussian distribution.

In the embodiment of the present application, according to natural scene statistical property, mean normalization (MSCN) coefficient that goes not by the primitive nature image of degeneration roughly meets Gaussian distribution, and the content of this distribution and image has nothing to do, and the quality degradation of image will change this distribution.Therefore, image operation being carried out to the image tentatively chosen and obtains MSCN coefficient, by judging whether roughly to meet the further screening that Gaussian distribution carries out image, picking out the high quality power line walking corridor image of access expansion image.

In the embodiment of the present application, to the natural image meeting Gaussian distribution, carry out acutance piecemeal and carefully select, composition power-line patrolling corridor image library comprises:

Acutance piecemeal is carried out to the natural image meeting Gaussian distribution;

Calculate the acutance of each image block, image block acutance being greater than default acutance threshold value adds power-line patrolling corridor image library.

In the embodiment of the present application, because the data volume of Aerial Images is very large, choose when building storehouse and the image block of Efficient Characterization picture quality can effectively can save storage space and characteristic parameter modeling time.Because human eye is responsive to sharp keen image-region, therefore in order to make image quality evaluation and subjective assessment have better consistance, mainly choosing the sharp keen image block of change and carrying out characteristic parameter modeling.

The Another reason that piecemeal is chosen is the reason such as out of focus, shake due to imaging device, the image chosen in image library all inevitably has some quality degradations, choose the degeneration composition of image can be reduced by piecemeal, make the qualitative character of image closer to natural image, thus the foundation of model is more accurate.

The acutance of each image block is defined as:

$\mathrm{\&delta;}\left(b\right)=E\left(\mathrm{\&Sigma;}{\mathrm{\&Sigma;}}_{(i,j)\mathrm{epatch\; b}}\mathrm{\&sigma;}(i,j)\right)$

After calculating the acutance of each image block, compare with acutance threshold T, setting threshold T=75%, the image block being greater than threshold value chooses warehouse-in.

In power-line patrolling corridor image library, amount of images is determined by image quality evaluation performance.Adopt different amount of images computation model characteristic parameters, image quality evaluation is carried out to the degraded image that known DMOS marks, by comparing the correlativity that quality evaluation result and DMOS mark, can know when the amount of images in image library exceedes some, the model parameter of algorithm just can be made more stable.

The present invention, on the basis of natural statistical property, first sets up high quality power line walking corridor image library, by extracting its spatial domain characteristic, obtains characteristic parameter model; Then, characteristic parameter matching is carried out to Aerial Images to be measured, by multivariate Gaussian model, matching is carried out to characteristic parameter again, compare based on the difference between the characteristic parameter model of image library and Aerial Images characteristic parameter, complete the quality assessment of unmanned plane Aerial Images.

As shown in Figure 3, be a kind of specific implementation of the embodiment of the present application evaluation method, the process identical with the image in image library is performed to Aerial Images to be tested, after extracting space domain characteristic, carry out matching, by comparing the mode of difference, obtain the evaluation score finally treating test pattern.

In the embodiment of the present application, evaluation method in the process of implementation, all can set up a new power-line patrolling corridor image library each time, also can set up a general power-line patrolling corridor image library, in each implementation, directly call.Concrete application process, chooses according to concrete picture appraisal process.

Corresponding with the quality evaluating method of Aerial Images described in Fig. 1, the embodiment of the present application additionally provides a kind of QA system of Aerial Images, and this system is the execution carrier of described method, and the structural representation of described system as shown in Figure 4, comprising:

Construction unit 301, for building power-line patrolling corridor image library;

Extraction unit 302, for extracting the space domain characteristic of each natural image in described power-line patrolling corridor image library;

Fitting unit 303, for carrying out MVG models fitting to each space domain characteristic extracted, obtains the standard fit parameter of token image quality;

Comparing unit 304, for being compared by the fitting parameter of described standard fit parameter and acquired Aerial Images to be tested, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested;

Evaluation unit 305, for according to described difference value, carries out quality score to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.

On the basis of Fig. 4, the embodiment of the present application additionally provides a detailed construction schematic diagram of described evaluation system, and specifically as shown in Figure 5, on the basis of Fig. 4, described construction unit 301 comprises:

Choose subelement 306, for according to the Selecting All Parameters preset, choose the power-line patrolling corridor image comprising power-line patrolling target; Described Selecting All Parameters comprises image background, image angle and access time; Described power-line patrolling corridor image is natural image;

Sharpness computation subelement 307, for carrying out sharpness computation to each natural image chosen respectively, screening sharpness meets the natural image of preset requirement;

Gaussian distribution statistics subelement 308, the natural image for meeting preset requirement to sharpness goes mean normalization MSCN coefficient Gaussian distribution to add up, and screening meets the natural image of Gaussian distribution;

Carefully select subelement 309, for the natural image meeting Gaussian distribution, carry out acutance piecemeal and carefully select, composition power-line patrolling corridor image library.

In this instructions, each embodiment adopts the mode of going forward one by one to describe, and what each embodiment stressed is the difference with other embodiment, between each embodiment same or similar part mutually see.

To the above-mentioned explanation of the disclosed embodiments, professional and technical personnel in the field are realized or uses the application.To be apparent for those skilled in the art to the multiple amendment of these embodiments, General Principle as defined herein when not departing from the spirit or scope of the application, can realize in other embodiments.Therefore, the application can not be restricted to these embodiments shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.

Claims (10)

1. a quality evaluating method for Aerial Images, is characterized in that, comprising:

Build power-line patrolling corridor image library;

Extract the space domain characteristic of each natural image in described power-line patrolling corridor image library;

MVG models fitting is carried out to each space domain characteristic extracted, obtains the standard fit parameter of token image quality;

The fitting parameter of described standard fit parameter and acquired Aerial Images to be tested is compared, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested;

According to described difference value, quality score is carried out to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.

2. method according to claim 1, is characterized in that, described structure power-line patrolling corridor image library comprises:

According to the Selecting All Parameters preset, choose the power-line patrolling corridor image comprising power-line patrolling target; Described Selecting All Parameters comprises image background, image angle and access time; Described power-line patrolling corridor image is natural image;

Carry out sharpness computation to each natural image chosen respectively, screening sharpness meets the natural image of preset requirement;

Natural image sharpness being met to preset requirement goes mean normalization MSCN coefficient Gaussian distribution to add up, and screening meets the natural image of Gaussian distribution;

To the natural image meeting Gaussian distribution, carry out acutance piecemeal and carefully select, composition power-line patrolling corridor image library.

3. method according to claim 1, is characterized in that, described extraction space domain characteristic comprises:

Matching is carried out in the MSCN coefficients statistics distribution of generalized Gaussian distribution to each natural image of application zero-mean, applies asymmetrical generalized Gaussian distribution simultaneously and carries out parameter fitting to point to parameter, set up the characteristic parameter model of natural image;

Using the space domain characteristic of the characteristic parameter of each natural image in characteristic parameter model as natural image.

4. method according to claim 1, is characterized in that, describedly carries out MVG models fitting to each space domain characteristic, and the standard fit parameter obtaining token image quality comprises:

Using multivariate Gauss model carries out matching to the characteristic parameter of each natural image;

Using the estimated parameter mean vector of application model and the covariance matrix standard fit parameter as token image quality.

5. method according to claim 1, is characterized in that, described according to difference value, carries out quality score comprise described Aerial Images to be tested:

Determine the correspondence position of described difference value in default scoring interval;

According to the quality score of described corresponding position, described Aerial Images to be tested is evaluated.

6. method according to claim 2, is characterized in that, described each natural image to choosing carries out sharpness computation, and the natural image that screening sharpness meets preset requirement comprises:

Respectively the reference picture of the described natural image that the natural image chosen and employing low-pass filter construct is converted by default transformation rule, obtain described natural image and described reference picture subimage separately;

Structural similarity SSIM index is adopted to measure corresponding subimage; The SSIM weighted sum of each subimage is obtained the sharpness of described natural image.

7. method according to claim 2, is characterized in that, describedly goes mean normalization MSCN coefficient Gaussian distribution to add up to the natural image that sharpness meets preset requirement, and the natural image that screening meets Gaussian distribution comprises:

Natural image sharpness being met to preset requirement carries out image operation, obtains MSCN coefficient;

Judge whether described MSCN coefficient meets Gaussian distribution;

Screening MSCN coefficient meets the natural image of Gaussian distribution.

8. method according to claim 2, is characterized in that, to the natural image meeting Gaussian distribution, carries out acutance piecemeal and carefully selects, and composition power-line patrolling corridor image library comprises:

Acutance piecemeal is carried out to the natural image meeting Gaussian distribution;

Calculate the acutance of each image block, image block acutance being greater than default acutance threshold value adds power-line patrolling corridor image library.

9. a QA system for Aerial Images, is characterized in that, comprising:

Construction unit, for building power-line patrolling corridor image library;

Extraction unit, for extracting the space domain characteristic of each natural image in described power-line patrolling corridor image library;

Fitting unit, for carrying out MVG models fitting to each space domain characteristic extracted, obtains the standard fit parameter of token image quality;

Comparing unit, for being compared by the fitting parameter of described standard fit parameter and acquired Aerial Images to be tested, obtains the difference value of the fitting parameter of described standard fit parameter and described Aerial Images to be tested;

Evaluation unit, for according to described difference value, carries out quality score to described Aerial Images to be tested, to realize the quality assessment to described Aerial Images to be tested.

10. system according to claim 9, is characterized in that, described construction unit comprises:

Choose subelement, for according to the Selecting All Parameters preset, choose the power-line patrolling corridor image comprising power-line patrolling target; Described Selecting All Parameters comprises image background, image angle and access time; Described power-line patrolling corridor image is natural image;

Sharpness computation subelement, for carrying out sharpness computation to each natural image chosen respectively, screening sharpness meets the natural image of preset requirement;

Gaussian distribution statistics subelement, the natural image for meeting preset requirement to sharpness goes mean normalization MSCN coefficient Gaussian distribution to add up, and screening meets the natural image of Gaussian distribution;

Carefully select subelement, for the natural image meeting Gaussian distribution, carry out acutance piecemeal and carefully select, composition power-line patrolling corridor image library.