阅读说明：本技术 基于大数据的安全检测方法 (Safety detection method based on big data ) 是由 周宗明 于 2020-07-27 设计创作，主要内容包括：本发明涉及一种基于大数据的安全检测方法,其包括：图像预处理模块根据外观检测请求中的检查部位信息去除第一客机检测图像中的冗余信息以得到第二客机检测图像；特征提取模块提取第二客机检测图像中外观检测部位的几何信息；模型生成模块根据获取到的客机标准图像和位姿信息得到不同位姿下的客机位姿图像,并对所有客机位姿图像进行特征归一化判别以将通过判别的客机位姿图像存储至对应的图像匹配库中；视角分析模块根据几何信息进行位姿相关性分析以得到第二客机检测图像的视角信息,并根据所述视角信息从图像匹配库中获取对应位姿的客机位姿图像；外观识别模块根据所述客机位姿图像和第二客机检测图像识别出客机存在的外观缺陷。(The invention relates to a security detection method based on big data, which comprises the following steps: the image preprocessing module removes redundant information in a first passenger plane detection image according to the inspection part information in the appearance detection request to obtain a second passenger plane detection image; the feature extraction module extracts geometric information of an appearance detection part in a second passenger plane detection image; the model generation module obtains passenger plane pose images under different poses according to the obtained passenger plane standard images and pose information, and performs characteristic normalization judgment on all the passenger plane pose images so as to store the passenger plane pose images passing the judgment into a corresponding image matching library; the visual angle analysis module is used for performing pose correlation analysis according to the geometric information to obtain visual angle information of a second passenger plane detection image, and acquiring a passenger plane pose image with a corresponding pose from an image matching library according to the visual angle information; and the appearance recognition module recognizes the appearance defects of the passenger plane according to the passenger plane pose image and the second passenger plane detection image.)

1. A safety detection method based on big data is characterized in that an appearance inspection terminal sends an appearance inspection instruction to an unmanned aerial vehicle, wherein the appearance inspection instruction comprises an airliner number, an airliner type and inspection part information;

the unmanned aerial vehicle responds to the received appearance inspection instruction to acquire an image of a specified part of the passenger plane to obtain a first passenger plane inspection image, generates an appearance inspection request according to the type of the passenger plane, the inspection part information and the first passenger plane inspection image, and then sends the appearance inspection request to an appearance inspection platform;

an image preprocessing module of the appearance detection platform removes redundant information in a first passenger plane detection image according to the inspection part information to obtain a second passenger plane detection image;

the feature extraction module extracts geometric information of an appearance detection part in a second passenger plane detection image;

the model generation module obtains passenger plane pose images under different poses according to the obtained passenger plane standard images and pose information, performs characteristic normalization judgment on all passenger plane pose images, and then stores the passenger plane pose images passing the judgment into a corresponding image matching library;

the visual angle analysis module performs pose correlation analysis according to the geometric information to obtain visual angle information of a second passenger plane detection image, and acquires a target passenger plane pose image from an image matching library according to the visual angle information, the passenger plane type and the inspection part information;

and the appearance recognition module recognizes appearance defects of the passenger plane according to the pose image of the target passenger plane and the detection image of the second passenger plane.

2. The method of claim 1, wherein the first passenger aircraft inspection image is a passenger aircraft appearance image acquired by a high resolution image sensor of the unmanned aircraft, comprising exterior bypass inspection images at different viewing angles;

the second passenger plane detection image is an image obtained by removing spatial redundancy, visual redundancy, information entropy redundancy, structural redundancy and knowledge redundancy from the first passenger plane detection image;

the standard image of the passenger plane is an appearance image of the passenger plane in an original state acquired by an image sensor, and the original state is a state to be flown when no damage occurs to the appearance of the passenger plane.

3. The method of claim 2, wherein the geometric information is a geometric feature presented by an appearance detection site in the second passenger aircraft detection image;

and the pose correlation analysis is to calculate the slope and curvature of the appearance detection part according to the parallel, vertical, regular, equal and similar graphic features indicated by the geometric information and obtain the corresponding image shooting angle according to the slope and curvature.

4. The method of claim 3, wherein the inspection site information includes an appearance detection site and a site identifier; the appearance detection part comprises a porthole part, a wing part, an empennage part and a fuselage part of the passenger plane;

the wing portion includes; the fuselage part comprises a nose part, a left front fuselage, a right front fuselage, a left rear fuselage, a right rear fuselage and an undercarriage;

the wing portion includes a left side wing leading edge, a right side wing leading edge, a left wing middle portion, and a right wing middle portion.

5. The method of claim 4, wherein an appearance recognition module identifies an appearance defect present on the passenger aircraft based on the passenger aircraft pose image and a second passenger aircraft detection image:

the appearance recognition module extracts texture features and geometric features of the passenger plane pose image and a second passenger plane detection image respectively to obtain a first appearance mixed feature and a second appearance mixed feature;

the appearance identification module is used for calculating the characteristic difference value of the first appearance mixed characteristic and the second appearance mixed characteristic to obtain a defect characteristic string;

and the appearance recognition module recognizes the appearance defects existing in the passenger plane according to the defect feature string.

6. The method according to claim 5, wherein the first appearance mixture feature and the second appearance mixture feature comprise a plurality of feature strings and feature information corresponding to each feature string, and the feature information comprises texture feature point information, geometric feature point information and feature point serial information.

7. The method of claim 6, wherein the appearance recognition module calculates a feature difference between the first appearance mixture feature and the second appearance mixture feature to obtain a defect feature string, and further comprising: and the appearance identification module obtains a corresponding defect area in the passenger plane detection image according to the defect feature string and the feature information corresponding to the defect feature string.

8. The method according to claim 7, wherein the pose information includes angle-of-view information and orientation information, the angle-of-view information being a shooting angle of view with a target shooting location as a reference center and an angle size corresponding to the shooting angle of view, the shooting angle of view including: frontal view, lateral view, overhead view, and overhead view.

9. The method of claim 8, wherein the model generation module performing feature normalization discrimination on all passenger aircraft pose images comprises:

the model generation module extracts the characteristics of the standard image of the passenger plane to obtain the characteristics of the original image part;

the model generation module extracts the characteristics of each passenger plane pose image to obtain the image position characteristics of each passenger plane pose image;

the model generation module performs characteristic normalization processing on the image position characteristics of each passenger plane pose image to obtain unified pose characteristics of each passenger plane pose image;

and the model generation module calculates the characteristic similarity value of the unified pose characteristic of each passenger plane pose image and the original image position characteristic, and compares the characteristic similarity value with a similarity threshold value.

10. The method of claim 9, wherein the model generation module performing a feature normalization process on the image location features of each guest pose image to obtain a uniform pose feature for each guest pose image comprises:

the model generation module obtains a plurality of position feature vectors of each passenger plane pose image according to the image position features of each passenger plane pose image;

and the model generation module obtains the unified pose characteristics of each passenger plane pose image according to the plurality of position characteristic vectors of each passenger plane pose image.

Technical Field

The invention relates to the field of big data and security detection, in particular to a security detection method based on big data.

Background

After the airplane and the airborne equipment fly for a period of time, phenomena such as abrasion, looseness, corrosion and the like can occur, so that detailed and complicated inspection and maintenance work needs to be carried out before the airplane takes off, and the airplane is ensured to safely execute the next flight task.

The inspection of the external detour of the airplane is one of the necessary works before the flight, and the purpose is to ensure the integral state of the airplane, the good state of the visible components and equipment of the airplane and the flight safety of the airplane. For example, when the appearance of the aircraft is defective, the flight resistance is increased, the flight performance is degraded, and the aircraft vibrates abnormally. In addition, the defect regions can form stress concentrations that affect the fatigue resistance of the component.

Currently, the appearance of passenger aircraft is typically inspected by manual visual inspection for scratches, dents, paint defects, or other damage. Therefore, before the passenger plane takes off, the passenger plane quality inspector needs to use the lifting platform to inspect the surface quality of the passenger plane for several hours, the inspection efficiency is low, the inspection process is complicated and time-consuming, and the accuracy of the appearance quality flaw inspection highly depends on the working state of the passenger plane quality inspector.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides a security detection method based on big data, which comprises the following steps:

the method comprises the steps that an appearance inspection terminal sends an appearance inspection instruction to the unmanned aerial vehicle, wherein the appearance inspection instruction comprises a passenger plane number, a passenger plane type and inspection part information;

the unmanned aerial vehicle responds to the received appearance inspection instruction to acquire an image of a specified part of the passenger plane to obtain a first passenger plane inspection image, generates an appearance inspection request according to the type of the passenger plane, the inspection part information and the first passenger plane inspection image, and then sends the appearance inspection request to an appearance inspection platform;

an image preprocessing module of the appearance detection platform removes redundant information in a first passenger plane detection image according to the inspection part information to obtain a second passenger plane detection image;

the feature extraction module extracts geometric information of an appearance detection part in a second passenger plane detection image;

the model generation module obtains passenger plane pose images under different poses according to the obtained passenger plane standard images and pose information, performs characteristic normalization judgment on all passenger plane pose images, and then stores the passenger plane pose images passing the judgment into a corresponding image matching library;

the visual angle analysis module is used for carrying out pose correlation analysis according to the geometric information to obtain visual angle information of a second passenger plane detection image, and acquiring a passenger plane pose image corresponding to a pose from an image matching library according to the visual angle information, the passenger plane type and the inspection part information;

and the appearance recognition module recognizes appearance defects of the passenger plane according to the passenger plane pose image and the second passenger plane detection image, generates a detection report according to the appearance defects, and sends the detection report to a corresponding appearance inspection terminal.

According to a preferred embodiment, the first passenger plane detection image is a passenger plane appearance image collected by a high-resolution image sensor of the unmanned plane, and the passenger plane appearance image comprises external bypass inspection images at different viewing angles;

the second passenger plane detection image is an image obtained by removing spatial redundancy, visual redundancy, information entropy redundancy, structural redundancy and knowledge redundancy from the first passenger plane detection image;

the standard image of the passenger plane is an appearance image of the passenger plane in an original state acquired by an image sensor, and the original state is a state to be flown when no damage occurs to the appearance of the passenger plane.

According to a preferred embodiment, the high resolution image sensor comprises: a panoramic camera, a monocular camera, a binocular camera and a trinocular camera.

According to a preferred embodiment, the redundant information comprises spatial redundancy, visual redundancy, information entropy redundancy, structural redundancy and knowledge redundancy.

According to a preferred embodiment, the geometric information is a geometric feature presented by an appearance detection part in the second passenger aircraft detection image;

and the pose correlation analysis is to calculate the slope and curvature of the appearance detection part according to the parallel, vertical, regular, equal and similar graphic features indicated by the geometric information and obtain the corresponding image shooting angle according to the slope and curvature.

According to a preferred embodiment, the inspection site information includes an appearance detection site and a site identifier; the appearance detection part comprises a porthole part, a wing part, an empennage part and a fuselage part of the passenger plane;

the wing portion includes; the fuselage part comprises a nose part, a left front fuselage, a right front fuselage, a left rear fuselage, a right rear fuselage and an undercarriage;

the wing portion includes a left side wing leading edge, a right side wing leading edge, a left wing middle portion, and a right wing middle portion.

According to a preferred embodiment, the appearance recognition module recognizes appearance defects present on the passenger aircraft based on the passenger aircraft pose image and the second passenger aircraft detection image:

the appearance recognition module extracts texture features and geometric features of the passenger plane pose image and a second passenger plane detection image respectively to obtain a first appearance mixed feature and a second appearance mixed feature;

the appearance identification module is used for calculating the characteristic difference value of the first appearance mixed characteristic and the second appearance mixed characteristic to obtain a defect characteristic string;

and the appearance recognition module recognizes the appearance defects existing in the passenger plane according to the defect feature string.

According to a preferred embodiment, the first appearance mixture feature and the second appearance mixture feature comprise a plurality of feature strings and feature information corresponding to each feature string, and the feature information comprises texture feature point information, geometric feature point information and feature point serial information.

According to a preferred embodiment, the appearance recognition module calculates a feature difference between the first appearance mixture feature and the second appearance mixture feature to obtain a defect feature string, and the steps further include: and the appearance identification module obtains a corresponding defect area in the passenger plane detection image according to the defect feature string and the feature information corresponding to the defect feature string.

According to a preferred embodiment, the pose information includes angle of view information and orientation information, the angle of view information is a shooting angle of view with a target shooting position as a reference center and an angle size corresponding to the shooting angle of view, and the shooting angle of view includes: frontal view, lateral view, oblique lateral view, planar view, overhead view, and overhead view.

According to a preferred embodiment, the feature normalization discrimination of all the passenger aircraft pose images by the model generation module comprises the following steps:

the model generation module extracts the characteristics of the standard image of the passenger plane to obtain the characteristics of the original image part;

the model generation module extracts the characteristics of each passenger plane pose image to obtain the image position characteristics of each passenger plane pose image;

the model generation module performs characteristic normalization processing on the image position characteristics of each passenger plane pose image to obtain unified pose characteristics of each passenger plane pose image;

and the model generation module calculates the characteristic similarity value of the unified pose characteristic of each passenger plane pose image and the original image position characteristic, and compares the characteristic similarity value with a similarity threshold value.

According to a preferred embodiment, the step of performing feature normalization processing on the image position features of each passenger aircraft pose image by the model generation module to obtain the unified pose features of each passenger aircraft pose image includes:

the model generation module obtains a plurality of position feature vectors of each passenger plane pose image according to the image position features of each passenger plane pose image;

and the model generation module obtains the unified pose characteristics of each passenger plane pose image according to the plurality of position characteristic vectors of each passenger plane pose image.

According to a preferred embodiment, the step of performing feature normalization processing on the image position features of each passenger aircraft pose image by the model generation module to obtain the unified pose features of each passenger aircraft pose image includes:

wherein F is the uniform pose feature, i is the index of the position feature vector, n is the number of the position feature vectors, and w_iThe weight of the ith position feature vector is taken up,normalizing the processed part for the ith dimensionA bit feature vector v is a pose corresponding to the position image of the passenger plane,is a multidimensional loss function.

According to a preferred embodiment, the cosmetic defects are cosmetic defects which can adversely affect the flight conditions of the passenger aircraft and include cosmetic scratches, cosmetic depressions and paint defects.

According to a preferred embodiment, the target passenger aircraft pose image is a passenger aircraft pose image corresponding to a shooting pose of the unmanned aircraft.

The embodiment of the invention has the following beneficial effects:

the safety detection method based on big data provided by the invention automatically detects the defects existing in the appearance of the passenger plane through the passenger plane appearance image shot when the outer part of the unmanned plane winds around, so as to avoid the defects from causing adverse effects on the flight of the passenger plane, eliminate the influence of adverse appearance factors and ensure the good appearance state of the passenger plane. In addition, the invention can analyze the passenger plane appearance images captured by the unmanned plane at different visual angles in real time and identify the appearance defects in the images, and eliminate the influence of different shooting visual angles on the detection of the passenger plane appearance defects by generating the appearance pose images at the same visual angle as the characteristic reference of the passenger plane appearance detection images at the corresponding visual angles, thereby greatly shortening the time for manually and visually checking the passenger plane appearance, improving the passenger plane appearance checking efficiency and accuracy, saving the human resources and reducing the time cost.

Drawings

Fig. 1 is a flowchart of a big data based security detection method according to an exemplary embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings in conjunction with the following detailed description. It should be understood that the description is intended to be exemplary only, and is not intended to limit the scope of the present invention. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present invention.

It is to be understood that although the terms first, second, third, etc. may be used herein to describe various information, these information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present invention.

Referring to fig. 1, in one embodiment, a big data based security detection method may include:

and S1, sending an appearance inspection instruction to the unmanned aerial vehicle by the appearance inspection terminal, wherein the appearance inspection instruction comprises the type of the passenger aircraft and the inspection part information.

Optionally, the examination region information includes an appearance detection region and a region identifier; the appearance detection part comprises a porthole part, a wing part, an empennage part and a fuselage part of the passenger plane; the part identifier is used for uniquely identifying different detection parts of the passenger plane;

the wing portion includes; the fuselage part comprises a nose part, a left front fuselage, a right front fuselage, a left rear fuselage, a right rear fuselage and an undercarriage; the wing portion includes a left side wing leading edge, a right side wing leading edge, a left wing middle portion, and a right wing middle portion.

Alternatively, the passenger aircraft types are different kinds of civil aircraft produced and used by different airlines, such as boeing passenger aircraft, fast sail passenger aircraft and airbus.

S2, the unmanned aerial vehicle responds to the received appearance inspection instruction, acquires an image of the designated part of the passenger aircraft to obtain a first passenger aircraft detection image, generates an appearance inspection request according to the type of the passenger aircraft, the inspection part information and the first passenger aircraft detection image, and then sends the appearance inspection request to the appearance inspection platform.

Optionally, the first passenger aircraft detection image is a passenger aircraft appearance image collected by a high-resolution image sensor of the unmanned aircraft, and the passenger aircraft appearance image comprises external bypass inspection images at different viewing angles.

Optionally, the high resolution image sensor comprises: a panoramic camera, a monocular camera, a binocular camera and a trinocular camera.

S3, removing redundant information in the first passenger aircraft detection image according to the inspection part information by an image preprocessing module of the appearance detection platform to obtain a second passenger aircraft detection image.

Optionally, the appearance detection platform automatically detects an appearance detection image of the passenger plane shot by the unmanned plane in response to the received appearance detection request, and identifies defects existing in the appearance of the passenger plane. In practical application, the outside of the passenger plane is subjected to the plane winding inspection, namely whether defects such as scratches, depressions, paint defects and the like exist in the appearance of the passenger plane is an important preparation work before the passenger plane takes off, so that the good flying state of the passenger plane can be ensured, the appearance defects of the passenger plane can be found in advance, and the adverse effect of the appearance defects on the flying of the passenger plane is avoided. Meanwhile, the aircraft is beneficial to the flight safety of the passenger plane, and the maintenance cost is effectively reduced.

Currently, the appearance of passenger aircraft is typically inspected by manual visual inspection for scratches, dents, paint defects, or other damage. Therefore, before the passenger plane takes off, the passenger plane quality inspector needs to use the lifting platform to inspect the surface quality of the passenger plane for several hours, the inspection efficiency is low, the inspection process is complicated and time-consuming, and the accuracy of the appearance quality flaw inspection highly depends on the working state of the passenger plane quality inspector.

In the case of poor working conditions for aircraft quality inspectors, it may result in some of the appearance defects present on passenger aircraft not being detected, with serious consequences. For example, cracks appearing on the porthole of a passenger plane can cause the passenger plane to burst and fall off due to the porthole in the flying process after the cracks are not detected, and serious adverse effects are generated on the flying state of the passenger plane.

Optionally, the redundant information comprises spatial redundancy, visual redundancy, information entropy redundancy, structural redundancy and knowledge redundancy,

the spatial redundancy refers to redundancy caused by strong correlation existing between adjacent pixels in an image; visual redundancy refers to partial image information which cannot be perceived or insensitive by human eyes; the information entropy redundancy is that the bit number used by each pixel in the image is larger than the information entropy of the image; structural redundancy means that there is a strong texture or self-similarity in the image; knowledge redundancy means that the image also contains information about some verification knowledge.

Optionally, the second passenger aircraft detection image is an image obtained by removing spatial redundancy, visual redundancy, information entropy redundancy, structural redundancy, and knowledge redundancy from the first passenger aircraft detection image.

S4, the feature extraction module extracts geometric information of the appearance detection part in the second passenger aircraft detection image.

Optionally, the geometric information is a geometric feature presented by an appearance detection part in the second passenger aircraft detection image.

And S5, the model generation module obtains passenger plane pose images under different poses according to the obtained passenger plane standard images and pose information, performs characteristic normalization judgment on all the passenger plane pose images, and then stores the passenger plane pose images passing the judgment into a corresponding image matching library.

In practical application, when the appearance defects of the passenger plane are identified through the passenger plane appearance image shot by the unmanned plane, the passenger plane appearance defect identification method is influenced by different shooting visual angles, and appearance part characteristics extracted from appearance detection images under different visual angles are different, so that the accuracy of an identification result is reduced when the passenger plane appearance defect is identified by using uniform appearance part characteristics. And different shooting angles can shield corresponding appearance detection parts to different degrees, so that information of certain parts is lost, and the accuracy of identifying the appearance defects of the passenger plane is low.

The invention can generate the passenger plane pose images under different visual angles according to the standard passenger plane images, and automatically matches the passenger plane pose images under the same visual angle from the database to be used as image reference when detecting the passenger plane appearance images shot by the unmanned plane, thereby eliminating the influence of different visual angles on the passenger plane appearance detection result and improving the accuracy of passenger plane appearance defect detection.

Optionally, the model generation module performs feature normalization discrimination on all the passenger plane pose images to improve the degree of reality of the generated passenger plane pose images in different poses, so that the generated passenger plane pose images can be used as standard reference images for defect detection of passenger plane detection images in different poses.

Optionally, the pose information includes view angle information and orientation information, the view angle information is a shooting view angle using the target shooting position as a reference center and an angle corresponding to the shooting view angle, and the shooting view angle includes: frontal view, lateral view, oblique lateral view, planar view, overhead view, and overhead view.

Optionally, the standard image of the passenger aircraft is an image of the appearance of the passenger aircraft in an original state acquired by an image sensor, and the original state is a state to be flown when no damage occurs to the appearance of the passenger aircraft.

Specifically, the model generation module obtains passenger plane pose images at different poses according to the obtained passenger plane standard image and pose information, and the method comprises the following steps:

the model generation module extracts the characteristics of the standard image of the passenger plane to obtain the characteristics of the original image components;

the model generation module generates a basic feature image based on the original image component features, wherein the basic feature image is an image which retains the basic features of an appearance detection part in a standard image of the passenger plane;

the model generation module calculates deformation of basic features contained in the basic feature image under different visual angles according to different visual angle information and different azimuth information indicated by the pose information to obtain corresponding deformation features;

and the model generation module adjusts the basic characteristics of the appearance detection part in the basic characteristic image according to the deformation characteristics to generate the passenger plane pose images in different poses.

Optionally, the basic features include a region outline, a region color, and a region size, and the basic features and/or deformation features of the basic features are present in all guest pose images generated from the standard guest image.

Specifically, the feature normalization judgment of all the passenger plane pose images by the model generation module comprises the following steps:

the model generation module extracts the characteristics of the standard image of the passenger plane to obtain the characteristics of the original image part;

the model generation module extracts the characteristics of each passenger plane pose image to obtain the image position characteristics of each passenger plane pose image;

the model generation module performs characteristic normalization processing on the image position characteristics of each passenger plane pose image to obtain unified pose characteristics of each passenger plane pose image;

and the model generation module calculates the characteristic similarity value of the unified pose characteristic of each passenger plane pose image and the original image position characteristic, and compares the characteristic similarity value with a similarity threshold value.

Optionally, the model generation module compares the feature similarity value of the image part features in the uniform pose and the original image part features with a similarity threshold value,

when the feature similarity value is larger than the similarity threshold value, representing that all the generated passenger plane pose images pass through judgment, and storing all the passenger plane pose images into a corresponding image matching library by a model generation module;

and when the feature similarity value is not greater than the similarity threshold value, the generated all passenger plane pose images are not judged, and the model generation module deletes all passenger plane pose images and regenerates the passenger plane pose images.

Optionally, the similarity threshold is used to verify whether features from different images are similar, and the similarity threshold is a numerical value preset by a manager according to an actual situation.

Optionally, the passenger plane pose image is used as a feature reference of a passenger plane detection image at the same view angle, and the passenger plane pose image at the same view angle is consistent with the basic features of the appearance detection part of the passenger plane detection image, so that the accuracy of passenger plane appearance defect detection can be improved.

Specifically, the step of performing feature normalization processing on the image position features of each passenger plane pose image by the model generation module to obtain the unified pose features of each passenger plane pose image includes:

the model generation module obtains a plurality of position feature vectors of each passenger plane pose image according to the image position features of each passenger plane pose image;

and the model generation module obtains the unified pose characteristics of each passenger plane pose image according to the plurality of position characteristic vectors of each passenger plane pose image.

Optionally, the performing, by the model generation module, feature normalization processing on the image part features of each passenger aircraft pose image to obtain unified pose features of each passenger aircraft pose image includes:

wherein F is the uniform pose feature, i is the index of the position feature vector, n is the number of the position feature vectors, and w_iThe weight of the ith position feature vector is taken up,is a position feature vector after the ith dimension normalization processing, v is a position corresponding to the position and pose image of the passenger plane,is a multidimensional loss function.

And S6, the visual angle analysis module performs pose correlation analysis according to the geometric information to obtain visual angle information of a second passenger plane detection image, and acquires a passenger plane pose image corresponding to a pose from an image matching library according to the visual angle information, the passenger plane type and the inspection part information.

Optionally, the pose correlation analysis is to calculate a slope and a curvature of the appearance detection portion according to parallel, perpendicular, regular, equal and similar graphic features indicated by the geometric information, and obtain a corresponding image capturing angle according to the slope and the curvature.

And S7, the appearance recognition module recognizes appearance defects existing in the passenger plane according to the passenger plane pose image and the second passenger plane detection image, generates a detection report according to the appearance defects, and sends the detection report to a corresponding appearance inspection terminal.

Specifically, the appearance recognition module recognizes appearance defects existing in the passenger plane according to the passenger plane pose image and the second passenger plane detection image:

the appearance recognition module extracts texture features and geometric features of the passenger plane pose image and a second passenger plane detection image respectively to obtain a first appearance mixed feature and a second appearance mixed feature;

the appearance identification module is used for calculating the characteristic difference value of the first appearance mixed characteristic and the second appearance mixed characteristic to obtain a defect characteristic string;

and the appearance recognition module recognizes the appearance defects existing in the passenger plane according to the defect feature string.

Optionally, the first appearance mixture feature and the second appearance mixture feature include a plurality of feature strings and feature information corresponding to each feature string, where the feature information includes texture feature point information, geometric feature point information, and feature point serial information.

Optionally, the first appearance blending feature is a blend of texture features and geometric features of the passenger aircraft pose image; the second appearance mixture feature is a mixture of texture features and geometric features of the second passenger aircraft detection image.

Optionally, the appearance recognition module obtains the defect feature string by calculating a feature difference between the first appearance mixture feature and the second appearance mixture feature, and the step further includes: and the appearance identification module obtains a corresponding defect area in the passenger plane detection image according to the defect feature string and the feature information corresponding to the defect feature string.

Alternatively, the cosmetic defects are cosmetic defects that can adversely affect the flight conditions of the passenger aircraft, including cosmetic scratches, cosmetic depressions, and paint defects.

The safety detection method based on big data provided by the invention automatically detects the defects existing in the appearance of the passenger plane through the passenger plane appearance image shot when the outer part of the unmanned plane winds around, so as to avoid the defects from causing adverse effects on the flight of the passenger plane, eliminate the influence of adverse appearance factors and ensure the good appearance state of the passenger plane. The method can analyze the passenger plane appearance images captured by the unmanned plane at different visual angles in real time, identify the appearance defects in the images, and eliminate the influence of different shooting visual angles on the detection of the passenger plane appearance defects by generating the appearance pose images at the same visual angle as the characteristic reference of the passenger plane appearance detection images at the corresponding visual angles, thereby greatly shortening the time for manually and visually checking the passenger plane appearance, improving the passenger plane appearance checking efficiency and accuracy, saving human resources and reducing the time cost.

In one embodiment, a big data based security detection system includes an appearance inspection terminal, an appearance detection platform, and a drone. The appearance detection platform is in communication connection with the inspection terminal. The appearance inspection terminal is a device with a calculation function, a storage function and a communication function used by an inspector, and comprises: smart mobile phones, desktop computers, notebook computers and intelligent wearable devices.

The appearance detection platform comprises an image preprocessing module, a feature extraction module, a model generation module, a visual angle analysis module and an appearance identification module.

The image preprocessing module is used for removing redundant information in the first passenger aircraft detection image according to the inspection part information in the appearance detection request to obtain a second passenger aircraft detection image.

The feature extraction module is used for extracting geometric information of an appearance detection part in a second passenger plane detection image.

The model generation module is used for obtaining passenger plane pose images under different poses according to the obtained passenger plane standard images and pose information, performing characteristic normalization judgment on all the passenger plane pose images, and then storing the passenger plane pose images passing the judgment into a corresponding image matching library.

And the visual angle analysis module is used for performing pose correlation analysis according to the geometric information to obtain visual angle information of a second passenger plane detection image, and acquiring a passenger plane pose image corresponding to a pose from an image matching library according to the visual angle information, the passenger plane type and the inspection part information.

And the appearance recognition module is used for recognizing appearance defects of the passenger plane according to the passenger plane pose image and the second passenger plane detection image.

The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.