CN113689067B - Image guarantee information evaluation system - Google Patents

Abstract

The invention provides an image guarantee information evaluation system, which is based on an image processing theory and a deep learning theory, aims at a plurality of links including image acquisition, image transmission, image processing and image analysis in specific fields and links of image guarantee, evaluates the independent guarantee capability, and divides the interior of each link into sub-links, evaluates the guarantee capability of the sub-links, establishes an evaluation index system based on the function subdivision protection, and can embody indexes such as time characteristics, accuracy characteristics, reliability characteristics and the like of the processing process on the whole, thereby providing support for the construction of the information processing evaluation capability of the image.

Description

Image guarantee information evaluation system

Technical Field

The present invention relates to the field of information evaluation, and more particularly, to an image assurance information evaluation system.

Background

Along with the development of information technology, an integrated combined combat system becomes a development trend of a future battlefield, and a combined information guarantee system is taken as an important component of the combined combat system and is a basis for combat decision and command control command generation.

Among them, scout images and videos acquired based on various scout means such as satellites, radars, etc. have become important campaign and tactical information, which is an important information source for commanding decisions and performing combat actions. In order to improve the working efficiency of image information interpretation personnel, the image information guarantee system can automatically score and intelligently analyze the obtained image quality. Therefore, the construction of the assessment tool for the guarantee capability of the image information is important to realize scoring of the image transmission flow and intelligent processing of the image.

Disclosure of Invention

Aiming at the technical problems existing in the prior art, the invention provides an image guarantee information evaluation system which comprises an image acquisition module, an image transmission module, an image processing module, an image analysis module, an image evaluation module and a comprehensive analysis module; the image acquisition module is used for acquiring pictures from a remote server or locally and recording a picture acquisition mode; the image transmission module is used for transmitting the acquired picture and recording network delay and coding format during picture transmission; the image processing module is used for performing image processing on the acquired picture; the image analysis module is used for analyzing and detecting targets in the pictures to obtain analysis and detection results; the image evaluation module is used for calculating scores of related image guarantee information indexes in the process of acquiring and transmitting the images according to the image acquisition mode, the network delay and the encoding mode; after the image processing, calculating the score of the image processing related guarantee information index; after the image analysis, calculating the score of the image analysis related guarantee information index according to the analysis detection result; the comprehensive analysis module is used for calculating a final image guarantee information evaluation comprehensive score according to the score of the related image guarantee information index, the score of the related image processing guarantee information index and the score of the related image analysis guarantee information index in the process of acquiring and transmitting the picture.

On the basis of the technical scheme, the invention can also make the following improvements.

Optionally, the image obtaining mode includes a remote server obtaining mode or a local obtaining mode, and the image evaluation module is configured to calculate, according to the image obtaining mode, the network delay and the encoding mode, a score of a relevant image guarantee information index in the process of obtaining and transmitting the image, where the score includes: the image evaluation module calculates a real-time index score and a fidelity index score of the picture in the acquisition process according to the picture acquisition mode, and calculates a confidentiality index score, a transmission delay index score and a compression degree index score of the picture in the transmission process according to the network delay and the coding mode in the picture transmission process.

Optionally, the image analysis module is configured to perform analysis and detection on a target in the picture, to obtain an analysis and detection result, and includes: detecting and analyzing targets in the pictures based on a remote sensing image detection and analysis mode or an SAR image detection and analysis mode to obtain target detection and analysis results; the remote sensing image detection analysis mode and the SAR image detection analysis mode are both based on a target detection network to identify targets in the picture.

Optionally, the target detection network comprises a dense feature pyramid module, an inclined candidate region generation module and a classification positioning module; the intensive feature pyramid module is used for extracting feature graphs of the pictures;

the inclination candidate region generation module is used for generating a plurality of inclination candidate regions representing each target position in the picture according to the feature map of the picture; the classifying and positioning module is used for correcting the positions of the plurality of inclined candidate areas based on the feature map of the picture, and identifying the types, the number and the positions of the targets in the picture and the identification confidence of each target.

Optionally, the dense feature pyramid module is a structure with tight connection added on the basis of an FPN structure, the dense feature pyramid module includes C2, C3, C4 and C5 convolution layers and P2, P3, P4 and P5 convolution layers, the C2, C3, C4 and C5 convolution layers are sequentially connected in a convolution manner, the P2, P3, P4 and P5 convolution layers are connected in a convolution dense manner, the C2 convolution layer is connected with the P2 convolution layer, the C3 convolution layer is connected with the P3 convolution layer, the C4 convolution layer is connected with the P4 convolution layer, and the C5 convolution layer is connected with the P5 convolution layer; and the output characteristic diagram of the C2 convolution layer is subjected to characteristic diagram splicing on the P2 convolution layer to obtain a spliced characteristic diagram, and 3X 3 convolution is carried out on the spliced characteristic diagram to obtain a characteristic diagram of the picture.

Optionally, the tilt candidate region generating module is configured to generate, according to a feature map of a picture, a plurality of tilt candidate regions that characterize respective target positions in the picture, including: the inclination candidate region generation module is used for generating inclination anchor point frames representing different angles, different scales and different proportions of each target position in the picture according to the feature map of the picture; correspondingly, the classifying and positioning module is configured to correct positions of a plurality of tilt candidate areas based on a feature map of a picture, and includes: extracting a region frame with fixed length for each inclined candidate region based on a rotation region pooling method according to the feature map of the picture; correspondingly, the image evaluation module is configured to calculate, after image analysis, a score of an image analysis related guarantee information index according to an analysis detection result, and includes: after the image analysis, calculating a remote sensing image detection analysis index score and a SAR image detection analysis index score according to the types, the number and the positions of the targets and the identification confidence of each target in the images identified by the remote sensing image detection analysis mode and the SAR image detection analysis mode.

Optionally, the image processing module is configured to perform image processing on the acquired picture, and includes: respectively carrying out image denoising processing, image deblurring processing and image super processing on the picture; correspondingly, the image evaluation module is configured to calculate a score of the image processing related guarantee information index after the image processing, and includes: after image processing, an image denoising index score, an image deblurring index score and an image superscore index score are calculated respectively.

Optionally, after the image processing, calculating an image denoising index score, an image deblurring index score and an image oversubscription index score respectively includes: according to the processed image, calculating an MSCN parameter feature vector of the processed image based on a local normalized luminance coefficient method MSCN; and obtaining a corresponding index score based on the mapping relation between the MSCN parameter feature vector and the index score.

Optionally, the comprehensive analysis module is configured to calculate a final image security information evaluation comprehensive score according to a score of a relevant image security information index, a score of an image processing relevant security information index, and a score of an image analysis relevant security information index in the process of acquiring and transmitting the image, and includes: establishing an index hierarchical structure module, wherein related image guarantee information indexes, image processing related guarantee information indexes and image analysis related guarantee information indexes in the image acquisition and transmission process are used as three-level indexes, an image acquisition module, an image transmission module, an image processing module and an image analysis module are used as two-level indexes, and an image guarantee information comprehensive index is used as a first-level index; calculating the score of each secondary index by weighted summation according to the score of the tertiary index included by each secondary index; the score of the image assurance information comprehensive index is calculated by weighted summation based on the score of each secondary index.

Optionally, the weight of each tertiary index or the weight of each secondary index is obtained by the following method: and converting qualitative descriptions of relative importance degrees between every two single indexes into quantitative descriptions in a comparison matrix for the multiple three-level indexes or the multiple two-level indexes under the same two-level index, solving the maximum eigenvector of the comparison matrix, normalizing the comparison matrix to obtain a weight vector, and corresponding to the weight of each three-level index or each two-level index in the image guarantee information.

The image guarantee information evaluation system provided by the invention is based on an image processing theory and a deep learning theory, aims at a plurality of specific fields and links of image guarantee, including image acquisition, image transmission, image processing and image analysis, evaluates the independent guarantee capability, and each link is internally divided into sub-links, and then evaluates the guarantee capability aiming at the sub-links, and establishes an evaluation index system on the basis of guaranteeing functional subdivision, so that the index system can embody indexes such as time characteristics, accuracy characteristics, reliability characteristics and the like of the processing process and support the construction of the image information processing evaluation capability.

Drawings

FIG. 1 is a schematic diagram of an image assurance information evaluation system according to the present invention;

FIG. 2 is a schematic diagram of a structure of a target detection network;

FIG. 3 is a schematic diagram of a dense feature pyramid module in a target detection network;

FIG. 4 is a schematic diagram of a feature map of a dense feature pyramid module extraction picture;

FIG. 5 is a schematic diagram of a plurality of tilt candidate regions generated by the tilt candidate region generation module;

FIG. 6 is a schematic diagram of a classification and positioning module correcting a tilt candidate region;

FIG. 7 is a schematic diagram of calculating an image processing-related assurance information indicator score;

FIG. 8 is a diagram of an established index hierarchy;

FIG. 9 is a schematic diagram of the relationship between different levels of indicators.

Detailed Description

The following describes in further detail the embodiments of the present invention with reference to the drawings and examples. The following examples are illustrative of the invention and are not intended to limit the scope of the invention.

Based on defects and requirements in the background art, fig. 1 shows an image guarantee information evaluation system provided by the invention, and as shown in fig. 1, the evaluation system comprises an image acquisition module, an image transmission module, an image processing module, an image analysis module, an image evaluation module and a comprehensive analysis module.

It will be appreciated that the image assurance information includes various index information during image acquisition, image transmission, image processing, and image analysis. It should be noted that, in the embodiment of the present invention, the image is an image taken by a radar or a satellite, and the image includes some small objects, for example, objects such as an airplane, a ship, and the like.

The image acquisition module is used for acquiring pictures from a remote server or locally and recording a picture acquisition mode; the image transmission module is used for transmitting the acquired picture and recording network delay and coding format during picture transmission; the image processing module is used for performing image processing on the acquired picture; the image analysis module is used for analyzing and detecting targets in the pictures to obtain analysis and detection results; the image evaluation module is used for calculating scores of related image guarantee information indexes in the process of acquiring and transmitting the pictures according to the picture acquisition mode, the network delay and the coding mode; after the image processing, calculating the score of the image processing related guarantee information index; after the image analysis, calculating the score of the image analysis related guarantee information index according to the analysis detection result; and the comprehensive analysis module is used for calculating a final image guarantee information evaluation comprehensive score according to the score of the related image guarantee information index, the score of the related image processing guarantee information index and the score of the related image analysis guarantee information index in the process of acquiring and transmitting the picture.

The invention is based on an image processing theory and a deep learning theory, aims at a plurality of links including image acquisition, image transmission, image processing and image analysis in specific fields of image guarantee, independently guarantees the capability assessment, and divides the interior of each link into each sub-link, then carries out the guarantee capability assessment aiming at the sub-links, establishes an assessment index system on the basis of guaranteeing functional subdivision, and can embody indexes such as time characteristics, accuracy characteristics, reliability characteristics and the like of the processing process on the whole, thereby providing support for the construction of the image information processing assessment capability.

In a possible embodiment, the image acquisition mode includes an acquisition mode from a remote server or an acquisition mode from a local server, and the image evaluation module is configured to calculate, according to the image acquisition mode, the network delay and the encoding mode, a score of a relevant image guarantee information index in an image acquisition and transmission process, where the score includes: the image evaluation module calculates a real-time index score and a fidelity index score of the picture in the acquisition process according to the picture acquisition mode, and calculates a confidentiality index score, a transmission delay index score and a compression degree index score of the picture in the transmission process according to the network delay and the coding mode in the picture transmission process.

It can be understood that the image transmission and acquisition indexes in the image guarantee information include: real-time, fidelity, confidentiality, transmission delay, compression degree and other indexes. When the related image guarantee information indexes in the image acquisition process are evaluated, calculating the real-time index score and the fidelity index score of the image in the acquisition process according to the image acquisition mode. When relevant image guarantee information indexes in the image transmission process are evaluated, a confidentiality index score, a transmission delay index score and a compression degree index score of the image in the transmission process are calculated according to network delay and a coding mode in the image transmission process.

In a possible embodiment, the image analysis module is configured to perform analysis and detection on a target in a picture, to obtain an analysis and detection result, and includes: detecting and analyzing targets in the pictures based on a remote sensing image detection and analysis mode or an SAR image detection and analysis mode to obtain target detection and analysis results; the remote sensing image detection analysis mode and the SAR image detection analysis mode are used for identifying targets in the pictures based on the target detection network.

It will be appreciated that the image analysis and image processing process includes: remote sensing image detection analysis, SAR image detection analysis, image denoising, image deblurring, image super-resolution and other image processing. The system needs to record various performance indexes in the image transmission and acquisition process according to specific conditions, and also needs to provide corresponding image analysis and image processing functions. In image analysis, corresponding processing algorithms are required to be provided for different image data types, and the image analysis method based on deep learning is researched for different data sources, so that the analysis result is more accurate. For a remote sensing image, the target detection difficulty is that the detection targets are dense, the size is smaller, and the detection difficulty is high. The target detection of the SAR image is relatively easy, but the problem of target blurring also exists.

In the embodiment of the invention, when the image analysis module analyzes and detects the target in the picture, corresponding detection and analysis methods are adopted to identify the target in the picture according to different picture types. Specifically, if the picture is a remote sensing image, adopting a remote sensing image detection analysis mode to detect and analyze the picture; if the picture is the SAR image, adopting an SAR image detection analysis mode to detect and analyze the picture.

The method is characterized in that whether a remote sensing image detection analysis method or an SAR image detection analysis method is adopted to detect and analyze the picture, targets in the picture are identified based on a target identification network.

In one possible embodiment, the object detection network includes a dense feature pyramid module, an inclined candidate region generation module, and a classification location module; the intensive feature pyramid module is used for extracting feature graphs of the pictures; the inclination candidate region generation module is used for generating a plurality of inclination candidate regions representing each target position in the picture according to the feature map of the picture; and the classification positioning module is used for correcting the positions of the plurality of inclined candidate areas based on the feature map of the picture, and identifying the types, the number and the positions of the targets in the picture and the identification confidence of each target.

It can be understood that, as shown in fig. 2, the shallow feature map has more position information, the deep feature map has more semantic information, but the position information is fuzzy, and the position information in the shallow feature map needs to be kept for the problem of inaccurate positioning when small targets are detected, so that the Dense feature pyramid module provided by the invention designs a tightly connected structure on the basis of an FPN structure, adds Dense connection similar to Dense-Net on the basis of the FPN structure, namely, all outputs of the front layer are used as inputs of the rear layer, and adopts Concat to carry out channel connection instead of superposition according to the position, so that the position information in the shallow features is fully reused.

Referring to fig. 2, the object recognition network in the embodiment of the invention mainly includes a dense feature pyramid module, an inclined candidate region generation module and a classification positioning module, wherein the image is input into the dense feature pyramid module, the dense feature pyramid module outputs a feature image of the image, the inclined candidate region generation module generates inclined candidate regions for representing each object position in the image according to the feature image of the image, and the classification positioning module analyzes the types, the numbers and the object recognition confidence of the objects in the image according to the feature image and the inclined candidate regions of the image.

Referring to fig. 3, the dense feature pyramid module is a structure with a tight connection added on the basis of an FPN structure, and includes C2, C3, C4 and C5 convolution layers and P2, P3, P4 and P5 convolution layers, the C2, C3, C4 and C5 convolution layers are sequentially connected in a convolution manner, the P2, P3, P4 and P5 convolution layers are connected in a convolution dense manner, the C2 convolution layer is connected with the P2 convolution layer, the C3 convolution layer is connected with the P3 convolution layer, the C4 convolution layer is connected with the P4 convolution layer, and the C5 convolution layer is connected with the P5 convolution layer.

Referring to fig. 4, the output feature map of the C2 convolution layer passes through a 1×1 convolution layer, and is respectively connected with the feature map obtained by double up-sampling output by the P3 convolution layer, the feature map obtained by quadrupling up-sampling output by the P4 convolution layer, and the feature map obtained by eight times up-sampling output by the P5 convolution layer, so as to obtain a spliced feature map, and the feature map of the final picture is obtained by performing 3×3 convolutions on the spliced feature map.

In a possible embodiment, the tilt candidate region generating module is configured to generate, according to a feature map of a picture, a plurality of tilt candidate regions that characterize respective target positions in the picture, including: the inclination candidate region generation module is used for generating inclination anchor point frames representing different angles, different scales and different proportions of each target position in the picture according to the feature map of the picture; correspondingly, the classifying and positioning module is used for correcting the positions of a plurality of inclined candidate areas based on the feature map of the picture, and comprises the following steps: extracting a region frame with fixed length for each inclined candidate region based on a rotation region pooling method according to the feature map of the picture; correspondingly, the image evaluation module is used for calculating the score of the image analysis related guarantee information index according to the analysis detection result after the image analysis, and comprises the following steps: after the image analysis, calculating a remote sensing image detection analysis index score and a SAR image detection analysis index score according to the types, the number and the positions of the targets and the identification confidence of each target in the images identified by the remote sensing image detection analysis mode and the SAR image detection analysis mode.

It can be understood that referring to fig. 5, relevant setting parameters in the tilt candidate region generating module in the embodiment of the present invention are used to generate tilt anchor blocks with different angles, different scales and different proportions, and specifically 3 scales, 3 aspect ratios and 6 tilt angles are set. The generated candidate regions include numerous candidate frames with various angles and various scales, and in order to complete subsequent classification and regression, a rotation region pooling method is used in a classification positioning module, as shown in fig. 6, to extract a feature vector with a fixed length for each candidate region, so as to classify regression, and the type of each target, the number of targets and the confidence of each identified target in the picture.

After processing analysis, evaluating indexes of a remote sensing image detection analysis method and an SAR image detection analysis method according to the type of each target, the number of targets and the confidence coefficient of each target in the picture identified by the target identification network, and respectively obtaining a remote sensing image detection index score and an SAR image detection index score.

In a possible embodiment, the image processing module is configured to perform image processing on an acquired picture, and includes: respectively carrying out image denoising processing, image deblurring processing and image super processing on the picture; correspondingly, the image evaluation module is configured to calculate a score of the image processing related guarantee information index after the image processing, and includes: after image processing, an image denoising index score, an image deblurring index score and an image superscore index score are calculated respectively.

It can be understood that the image processing module mainly performs image denoising processing, image deblurring processing and image super-processing on the image respectively to obtain images after the image denoising processing, the image deblurring processing and the image super-dividing processing. The image evaluation module calculates an image denoising index score, an image deblurring index score and an image superscore index score respectively.

After image processing, when an image denoising index score, an image deblurring index score and an image superscore index score are respectively calculated, referring to fig. 7, the embodiment of the invention uses a non-reference image quality evaluation method based on deep learning to evaluate, utilizes a scene statistical model of a local normalized brightness coefficient (MSCN) to quantify the loss of 'naturalness' of an image, and uses the MSCN method to extract feature vectors of the processed image to obtain MSCN parameter feature vectors of the processed image, and finally uses a support vector model to map the parameter feature vectors to quality scores, wherein the MSCN parameter feature vectors and the quality scores have a certain mapping relation.

In one possible embodiment, it may be understood that the index hierarchy model is established when the comprehensive analysis module calculates the final image assurance information evaluation comprehensive score according to the score of the relevant image assurance information index, the score of the image processing relevant assurance information index, and the score of the image analysis relevant assurance information index in the process of acquiring and transmitting the picture. The image acquisition module, the image transmission module, the image processing module and the image analysis module are used as secondary indexes, the image guarantee information comprehensive indexes are used as primary indexes, the established index level model can be shown in fig. 8, and the score of each secondary index is calculated by weighting and summing according to the score of the tertiary index included in each secondary index; the score of the image assurance information comprehensive index is calculated by weighted summation based on the score of each secondary index.

It can be understood that referring to fig. 8, in the embodiment of the present invention, the image guarantee information index is divided into four element groups including an image acquisition, an image transmission, an image analysis and an image processing, where the image acquisition element group corresponds to the real-time and fidelity index, the image transmission element group corresponds to the confidentiality, transmission delay and compression degree index, the image analysis element group corresponds to the SAR detection and remote sensing detection index, and the image processing element group includes an image denoising, an image deblurring and an image superdistribution index. As shown in fig. 9, in addition to there being an association between element groups, there are also associations between sub-elements within an element group and sub-elements of other element groups. The embodiment of the invention builds a network analysis model by considering the relation among element groups on the basis of an analytic hierarchy process.

The network analysis model decomposes the problems into different composition factors according to the nature of the problems and the total target to be achieved, and aggregates and combines the factors according to the mutual correlation influence among the factors and the membership according to different layers to form a multi-layer analysis structure model, so that the problems are finally classified into the determination of the relative importance weight of the lowest layer (scheme for decision, measure and the like) relative to the highest layer (total target) or the arrangement of the relative priority order.

In the embodiment of the invention, the workflow of the network analysis model is as follows:

(1) And establishing a hierarchical structure model. The decision target, the considered factors (decision criteria) and the decision object are divided into a highest layer, a middle layer and a lowest layer according to the interrelationship between the decision target, the considered factors (decision criteria) and the decision object, and a hierarchy chart is drawn. Specifically, according to actual requirements, the importance degree of functions among all sub-modules and inside all sub-modules in the image guarantee flow is determined, and the importance degree is measured through relative importance. The hierarchical structure of the multi-level index established in the embodiment of the present invention can be shown in fig. 8.

(2) And (5) establishing a comparison matrix. According to the hierarchical structure model established in the step (1), all factors are compared in pairs instead of being put together by a consistent matrix method, and relative scales are adopted in the process so as to reduce the difficulty of comparing the factors with different properties as much as possible, thereby improving accuracy. If a certain criterion is adopted, the schemes below the criterion are compared pairwise, and the schemes are rated according to the importance degree. Wherein, the importance of the index is quantitatively scaled as shown in table 1:

table 1 importance scale

Equally important	Slightly important	Important is	Much more important	Extremely important is
					1	3	5	7	9

The relative importance level is obtained by comparing two indexes with each other, for example, the relative importance level between two indexes in the image transmission element group can be seen in table 2.

TABLE 2 relative importance of image transmission element groups

	Security performance	Degree of compression	Transmission delay	Real-time performance
					Security performance	Equally important	Equally important	Slightly important	Is obviously important
Degree of compression	-	Equally important	Slightly important	Is obviously important
					Transmission delay	-	-	Equally important	Slightly important
Real-time performance	-	-	-	Equally important

(3) The weights are normalized. And (3) comparing the matrix to obtain a maximum eigenvector, normalizing the maximum eigenvector to obtain a weight vector, and finally weighting and summing the weights of the corresponding indexes in a specific image information guarantee comprehensive evaluation system to obtain the image comprehensive information guarantee capability score.

The network analysis mode can be visually displayed through the following flow, taking an image transmission flow as an example, layering the importance degrees of functions such as confidentiality, compression degree and the like, converting the qualitative description of the relative importance degrees between every two single indexes into quantitative description in a comparison matrix, solving the maximum characteristic vector of the comparison matrix, normalizing the maximum characteristic vector to serve as a weight vector, and finally weighting and summing the weights of the corresponding indexes in a specific image information security comprehensive evaluation system to obtain the image comprehensive information security capability score. For example, consider the image transmission element group as an example, and the comparison matrix of the image transmission element group is shown in table 3.

TABLE 3 image transmission element group comparison matrix

	Security performance	Degree of compression	Transmission delay	Real-time performance
					Security performance	1	1	3	5
Degree of compression	1	1	3	5
					Transmission delay	1/3	1/5	1	3
Real-time performance	1/3	1/5	1/3	1

The weights of the image transmission element groups obtained from the comparison matrix of table 3 are shown in table 4:

TABLE 4 image transmission element group weights

Security performance	Degree of compression	Transmission delay	Real-time performance
				0.391	0.391	0.15	0.068

The method is adopted to determine the weights of a plurality of three-level indexes belonging to the same two-level index, and the score of the corresponding two-level index is calculated based on the score of each three-level index and the corresponding weight. Then, for each secondary index, the weight of each secondary index is determined in the same manner, and a final composite score is calculated based on the score of each secondary index and the corresponding weight.

The image guarantee information evaluation system provided by the embodiment of the invention mainly has the following effects:

(1) The new remote sensing image detection and SAR image detection are provided, a Dense characteristic pyramid structure is used, dense connection similar to Dense-Net is added on the basis of an FPN structure, namely all outputs of a front layer are used as inputs of a rear layer, and meanwhile, a spliced characteristic diagram is adopted for channel connection instead of bit superposition, so that characteristics are fully reused, and the detection performance of a small target is improved.

(2) On the basis of common target detection, the inclined candidate frames are used for generating inclined anchor frames with different angles, different scales and different proportions by generating a network; the inclined area is used for pooling in the classification positioning module to convert the inclined area into a characteristic vector with a fixed length for classification and regression, and finally, the remote sensing target detection at any angle is realized, so that the method is applicable to target detection at any angle and any shape.

(3) The method for evaluating the quality of the reference-free image is provided, wherein the loss of the naturalness of the image is quantified through a scene statistical model of a local normalized brightness coefficient, in order to quantitatively describe the mapping relation between the natural scene statistical coefficient and the distortion degree, the MSCN is utilized to construct adjacent coefficient inner products in different directions of the processed image, so as to obtain MSCN parameter feature vectors, and finally, the support vector model is used for mapping the parameter feature vectors to quality scores.

(4) On the basis of the analytic hierarchy process, a network analysis model is constructed by considering the relationship of the mutual influence of the sub-elements of each element group. Dividing the image guarantee information evaluation index into three levels, respectively obtaining a second-level index weight by adopting a hierarchical analysis method, obtaining a third-level index weight by adopting a network analysis method, and calculating to obtain a final image guarantee information evaluation score, thereby realizing scoring of the image information guarantee process and intelligent analysis of the image.

In the foregoing embodiments, the descriptions of the embodiments are focused on, and for those portions of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.

It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the invention.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims (7)

1. The image guarantee information evaluation system is characterized by comprising an image acquisition module, an image transmission module, an image processing module, an image analysis module, an image evaluation module and a comprehensive analysis module;

the image acquisition module is used for acquiring pictures from a remote server or locally and recording a picture acquisition mode;

the image transmission module is used for transmitting the acquired picture and recording network delay and coding format during picture transmission;

the image processing module is used for performing image processing on the acquired picture;

the image analysis module is used for analyzing and detecting targets in the pictures to obtain analysis and detection results;

the image evaluation module is used for calculating scores of related image guarantee information indexes in the process of acquiring and transmitting the images according to the image acquisition mode, the network delay and the encoding mode; after the image processing, calculating the score of the image processing related guarantee information index; after the image analysis, calculating the score of the image analysis related guarantee information index according to the analysis detection result;

the comprehensive analysis module is used for calculating a final image guarantee information evaluation comprehensive score according to the score of the related image guarantee information index, the score of the related image processing guarantee information index and the score of the related image analysis guarantee information index in the process of acquiring and transmitting the picture;

the image acquisition mode comprises an acquisition mode from a remote server or a local acquisition mode, and the image evaluation module is used for calculating scores of relevant image guarantee information indexes in the acquisition and transmission processes of the image according to the image acquisition mode, the network delay and the coding mode, and comprises the following steps:

the image evaluation module calculates a real-time index score and a fidelity index score of the picture in the acquisition process according to the picture acquisition mode, and calculates a confidentiality index score, a transmission delay index score and a compression degree index score of the picture in the transmission process according to the network delay and the coding mode in the picture transmission process;

the image analysis module is used for analyzing and detecting targets in the pictures to obtain analysis and detection results, and comprises the following steps:

detecting and analyzing targets in the pictures based on a remote sensing image detection and analysis mode or an SAR image detection and analysis mode to obtain target detection and analysis results;

the remote sensing image detection analysis mode and the SAR image detection analysis mode are used for identifying targets in the picture based on a target detection network, and respectively identifying the type, the number and the position of each target in the picture and the identification confidence of each target;

the image evaluation module is configured to calculate, after image analysis, a score of an image analysis related guarantee information index according to an analysis detection result, and includes:

after the image analysis, calculating a remote sensing image detection analysis index score and a SAR image detection analysis index score according to the types, the number and the positions of the targets and the identification confidence coefficient of each target in the images identified by the remote sensing image detection analysis mode and the SAR image detection analysis mode respectively;

the image processing module is configured to perform image processing on an acquired picture, and includes:

respectively carrying out image denoising processing, image deblurring processing and image super processing on the picture;

correspondingly, the image evaluation module is configured to calculate a score of the image processing related guarantee information index after the image processing, and includes:

after image processing, an image denoising index score, an image deblurring index score and an image superscore index score are calculated respectively.

2. The image assurance information evaluation system of claim 1, wherein the target detection network comprises a dense feature pyramid module, a sloped candidate region generation module, and a classification location module;

the intensive feature pyramid module is used for extracting feature graphs of the pictures;

the inclination candidate region generation module is used for generating a plurality of inclination candidate regions representing each target position in the picture according to the feature map of the picture;

the classifying and positioning module is used for correcting the positions of the plurality of inclined candidate areas based on the feature map of the picture, and identifying the types, the number and the positions of the targets in the picture and the identification confidence of each target.

3. The image guarantee information evaluation system according to claim 2, wherein the dense feature pyramid module is a structure added with tight connection on the basis of an FPN structure, the dense feature pyramid module comprises C2, C3, C4 and C5 convolution layers and P2, P3, P4 and P5 convolution layers, the C2, C3, C4 and C5 convolution layers are sequentially connected in a convolution manner, the P2, P3, P4 and P5 convolution layers are connected in a convolution dense manner, the C2 convolution layer is connected with the P2 convolution layer, the C3 convolution layer is connected with the P3 convolution layer, the C4 convolution layer is connected with the P4 convolution layer, and the C5 convolution layer is connected with the P5 convolution layer;

and the output characteristic diagram of the C2 convolution layer is subjected to characteristic diagram splicing on the P2 convolution layer to obtain a spliced characteristic diagram, and 3X 3 convolution is carried out on the spliced characteristic diagram to obtain a characteristic diagram of the picture.

4. The image assurance information evaluation system of claim 2, wherein the tilt candidate region generation module is configured to generate a plurality of tilt candidate regions characterizing respective target positions in a picture from a feature map of the picture, comprising:

the inclination candidate region generation module is used for generating inclination anchor point frames representing different angles, different scales and different proportions of each target position in the picture according to the feature map of the picture;

correspondingly, the classifying and positioning module is configured to correct positions of a plurality of tilt candidate areas based on a feature map of a picture, and includes:

and extracting a region frame with a fixed length for each inclined candidate region based on a rotation region pooling method according to the feature map of the picture.

5. The image assurance information evaluation system of claim 1, wherein the computing of the image denoising index score, the image deblurring index score, and the image superscore index score after image processing, respectively, comprises:

according to the processed image, calculating an MSCN parameter feature vector of the processed image based on a local normalized luminance coefficient method MSCN;

and obtaining a corresponding index score based on the mapping relation between the MSCN parameter feature vector and the index score.

6. The image assurance information evaluation system of claim 1, wherein the comprehensive analysis module is configured to calculate a final image assurance information evaluation comprehensive score according to a score of a relevant image assurance information index, a score of an image processing relevant assurance information index, and a score of an image analysis relevant assurance information index in the process of acquiring and transmitting a picture, and includes:

establishing an index hierarchical structure module, wherein related image guarantee information indexes, image processing related guarantee information indexes and image analysis related guarantee information indexes in the image acquisition and transmission process are used as three-level indexes, an image acquisition module, an image transmission module, an image processing module and an image analysis module are used as two-level indexes, and an image guarantee information comprehensive index is used as a first-level index;

calculating the score of each secondary index by weighted summation according to the score of the tertiary index included by each secondary index;

the score of the image assurance information comprehensive index is calculated by weighted summation based on the score of each secondary index.

7. The image assurance information evaluation system of claim 6, wherein the weight of each tertiary index or the weight of each secondary index is obtained by:

and converting qualitative descriptions of relative importance degrees between every two single indexes into quantitative descriptions in a comparison matrix for the multiple three-level indexes or the multiple two-level indexes under the same two-level index, solving the maximum eigenvector of the comparison matrix, normalizing the comparison matrix to obtain a weight vector, and corresponding to the weight of each three-level index or each two-level index in the image guarantee information.