CN-121999411-A - Airport aviation display content security-oriented multistage detection method and system

CN121999411ACN 121999411 ACN121999411 ACN 121999411ACN-121999411-A

Abstract

The invention belongs to the technical field of airport flight information display, and discloses a multilevel detection method and system for airport aviation display content security. According to the method, template similarity is detected, whether the similarity of the front image and the rear image of the same device is lower than a threshold value is detected, long-sequence detection is carried out, the content in the current image is identified by using an OCR algorithm, whether sensitive words exist in the content is checked, semantic analysis is carried out by using a BERT classification algorithm, whether illegal and tampered content exists in the content in the current image is analyzed by the image content and a prepared instruction word through a generated multi-mode model, and if yes, an alarm module is triggered to prompt that the content is possibly tampered. The invention discovers the problem that the content displayed by the airport flight information display system is tampered and gives an alarm at the first time, thereby minimizing the possible economic loss and reputation damage caused by the content security problem.

Inventors

LI FUCONG
ZHANG XINHUA
LIU FENGLING
Geng xue
Shan Yisheng

Assignees

青岛民航凯亚系统集成有限公司

Dates

Publication Date: 20260508
Application Date: 20260122

Claims (10)

1. The multi-level detection method for airport aviation display content security is characterized by comprising the following steps of: The method comprises the steps of S1, collecting image data being displayed by all flight display screens by utilizing an image acquisition module, detecting template similarity by utilizing a similarity analysis module according to the displayed images, detecting whether the similarity of the front and rear images of the same equipment is lower than a threshold value or not by comparing the similarity of the front and rear images of the same equipment, if the similarity is lower than the threshold value, tampering the content, and if the content is tampered, detecting; S2, long sequence detection is carried out through a long sequence detection module, similarity comparison values calculated by a plurality of frames of the equipment image are accumulated, a similarity prediction value is carried out through an LSTM+transducer prediction module, and if the predicted values of the continuous steps are lower than a threshold value, the continuous steps are tampered, and detection is continued; S3, recognizing the content in the current image by using an OCR algorithm through a semantic analysis module, checking whether sensitive words exist in the content, and performing semantic analysis by using a BERT classification algorithm, wherein whether tampered illegal content exists in the content in the image; s4, analyzing whether the illegal and tampered content exists in the content in the current image through the multi-mode analysis module and the instruction words prepared in advance, triggering the alarm module to prompt the user equipment that the content is tampered if the illegal and tampered content exists, and re-detecting the content from the beginning if the illegal and tampered content does not exist in the current image through the multi-mode analysis module, and returning to the step S1.
2. The multi-level detection method for airport aviation content security of claim 1, wherein in step S1, collecting image data being displayed on all flight displays by using an image acquisition module comprises: (1) Setting a display screen set in an airport internal flight information display system as Each element in the collection represents a display device; (2) Traversing the read collection by the image acquisition module Content being displayed by the device in question, and a content collection is generated , For a first frame image of a first device, For a first frame image of a second device, Is the first A first frame image of the device.
3. The multi-level detection method for airport aviation content security of claim 1, wherein in step S1, detecting template similarity using a similarity analysis module comprises: (a) Recording a current frame image and acquiring a next frame image by utilizing a similarity analysis module, comparing a front frame image and a rear frame image in the same equipment, carrying out graying treatment on the front frame image and the rear frame image through image preprocessing, and respectively weighting the images according to red, green and blue 3 channels Realizing; The brightness values of the red, green and blue channels respectively; (b) Using structural similarity index Calculating the similarity of the front frame image template and the rear frame image template, wherein the expression is as follows: ; in the formula, In order for the degree of similarity of the brightness, In order for the contrast to be similar, Is the structural similarity; (c) Will be And comparing the calculated value with a preset threshold value, and if the calculated value is lower than the threshold value, the matching of the task template fails and the task template is tampered.
4. The multi-level detection method for airport aviation content security of claim 3, wherein the expression of the brightness similarity is: ; in the formula, Respectively two frames of images And Is used for the luminance of the light source, Is a constant value, and is used for the treatment of the skin, , A constant of 0.01 for avoiding zero denominator L as the range of pixel values; The expression of contrast similarity is: ; in the formula, Respectively two frames of images And Is defined by the local standard deviation of (c), For avoiding a range of pixel values where denominator is zero L, A constant of 0.01; The structural similarity is expressed as: ; in the formula, For two frames of images And Is used to determine the local covariance of (1), 。
5. The multi-level detection method for airport aviation content security of claim 1, wherein in step S2, the long sequence detection by the long sequence detection module comprises: the long sequence detection module is used for each device on the basis of the frame image set P Calculating the structural similarity of the current frame and the historical frame through SSIM to form a feature vector: ; in the formula, Is the first A structural similarity index between every two frames of the first frame and the second frame of the device; And forming a sequence with the length of 10 by the characteristic values according to time sequence, predicting the next SSIM value in the sequence by a fusion algorithm model of LSTM+transducer, and judging tampering if a plurality of predicted SSIM values are continuously lower than a threshold value, and detecting the existence of the tampered sequence in the next stage.
6. The airport aviation content security-oriented multi-level detection method of claim 5, wherein the lstm+transform fusion algorithm model predicts a next SSIM value in the sequence, comprising: Constructing a data set according to the prepared SSIM value sequence prepared in advance; training a fusion algorithm model, and training an SSIM value sequence data set by fusing LSTM+a transducer; The next SSIM value is predicted by the trained lstm+transducer model.
7. The multi-level detection method for airport aviation content security according to claim 1, wherein in step S3, the semantic analysis module recognizes the content in the current image using OCR algorithm, checks whether there is a sensitive word in the content, and performs semantic analysis using BERT classification algorithm, comprising: Step 1, a semantic analysis module recognizes a current image frame acquired by an image acquisition module through an OCR text recognition model aiming at each element in a set P which is judged to be suspected to be tampered at two stages, and extracts information of images in the current frame to form a set: ; in the formula, Is the first Device No. Text information of the frame image; Detecting whether the content recognized by OCR contains sensitive word information or not through a sensitive word library, and if so, falsifying the content; Step 2, for And classifying the texts through the BERT model, detecting whether illegal contents exist in the text contents in the images through semantic understanding of the BERT model, and if so, displaying that the contents are suspected to be tampered.
8. The airport aviation content security-oriented multi-level detection method of claim 7, wherein classifying text by BERT model comprises: (1) Training BERT model recognition to recognize the vocabulary of the sensitive word stock; (2) The content in the sentence is classified by bert model, and it is confirmed whether the sentence has illegal meaning.
9. The multi-level detection method for airport aviation display content security according to claim 1, wherein in step S4, the multi-modal analysis module analyzes the image content and the instruction word prepared in advance, and the generated multi-modal model analyzes whether there is a tampered content in the current image, comprising: through a preset instruction word set , wherein, Instruction words set for a certain task are used for detecting whether the type of problem exists in the image or not, and the instruction words are assembled Generating content judgment by the instruction word in the image and the image acquired by the current frame through a generated multi-mode model, judging whether the content in the image has illegal conditions, and considering that the tampered conditions exist if the model judges that the illegal conditions exist; Will be assembled The method comprises the steps of constructing an instruction word set G for carrying out accurate analysis on a non-passing illegal scene, wherein gm is an instruction word for a specific certain scene, analyzing the instruction word and the image one by one through a generated multi-mode model, analyzing whether the certain illegal scene exists in the image according to the requirement of the instruction word by the generated multi-mode model, if yes, returning fasle is not existed, and if false is returned, the image does not have illegal content.
10. An airport-oriented content security multi-level detection system, characterized in that the system operates the airport-oriented content security multi-level detection method of any one of claims 1-9, the system comprising: the image acquisition module is used for collecting images which are being displayed by all flight display screens; The similarity analysis module is used for detecting the similarity of the templates, and detecting whether the similarity of the front and rear images of the equipment is lower than a threshold value or not, and if the similarity is lower than the threshold value, the content is tampered; The long sequence detection module is used for detecting a long sequence, accumulating similarity comparison values calculated by a plurality of frames of the equipment image, carrying out a similarity prediction value of the next step through the LSTM+transducer prediction module, and falsifying the similarity comparison values to continue the next step detection if the predicted values of the continuous steps are lower than a threshold value; The semantic analysis module is used for identifying the content in the current image by using an OCR algorithm, checking whether the content has sensitive words or not, performing semantic analysis by using a BERT classification algorithm, and judging whether the content in the image has tampered illegal content or not; The multi-mode analysis module is used for analyzing whether the illegal and tampered content exists in the content in the current image through the generation of the multi-mode model by using the image content and the instruction words prepared in advance, if so, the execution triggering alarm module prompts the user equipment that the content is tampered, and if not, the first-step image collection module is used for re-detecting from the head.

Description

Airport aviation display content security-oriented multistage detection method and system Technical Field The invention belongs to the technical field of airport flight information display, and particularly relates to a multistage detection method and system for airport aviation display content security. Background Currently, an airport flight information display system is taken as an important infrastructure in the civil aviation field and bears important responsibilities of providing key information such as real-time flight dynamics, boarding gates, luggage turntables and the like for passengers, airlines and airport staff. The accuracy, timeliness and safety of the content are directly related to the normal running order of the airport and the travel experience of passengers. However, with rapid development of information technology and increasing complexity of network environments, flight information display systems face increasing content security risks. Conventional content management approaches may be difficult to address a variety of potential threats to malicious tampering, illegal information insertion, virus attacks, or system failures. Once the system content is tampered with or improper information is released, the passengers are panicked and the flights are delayed by light weight, and serious social influence and economic loss can be caused by heavy weight. Therefore, there is a great need for a method that can comprehensively and effectively detect and prevent the content security threat of flight information display systems. In the prior art, although different detection means aiming at the safety of an information system exist, the content characteristics of the system are displayed by aiming at airport flight information, and various detection methods are comprehensively utilized, so that the solution of the content safety is comprehensively ensured. The invention is put forward under the background, and aims to provide a method for comprehensively detecting by a plurality of methods, so as to ensure the safety of the content of the airport flight information display system. Disclosure of Invention In order to overcome the problems in the related art, the disclosed embodiment of the invention provides a multistage detection method and a multistage detection system for airport aviation display content safety, which are based on an airport flight information display system, relate to the content safety detection of the airport flight information display system in the civil aviation field, and are a method for comprehensively detecting through a plurality of methods to ensure the content safety of the aviation display system. The safety of the content of the airport flight information display system is guaranteed. The technical scheme is that the airport aviation display content security-oriented multistage detection method comprises the following steps of: The method comprises the steps of S1, collecting image data being displayed by all flight display screens by utilizing an image acquisition module, detecting template similarity by utilizing a similarity analysis module according to the displayed images, detecting whether the similarity of the front and rear images of the same equipment is lower than a threshold value or not by comparing the similarity of the front and rear images of the same equipment, if the similarity is lower than the threshold value, tampering the content, and if the content is tampered, detecting; S2, long sequence detection is carried out through a long sequence detection module, similarity comparison values calculated by a plurality of frames of the equipment image are accumulated, a similarity prediction value is carried out through an LSTM+transducer prediction module, and if the predicted values of the continuous steps are lower than a threshold value, the continuous steps are tampered, and detection is continued; S3, recognizing the content in the current image by using an OCR algorithm through a semantic analysis module, checking whether sensitive words exist in the content, and performing semantic analysis by using a BERT classification algorithm, wherein whether tampered illegal content exists in the content in the image; s4, analyzing whether the illegal and tampered content exists in the content in the current image through the multi-mode analysis module and the instruction words prepared in advance, triggering the alarm module to prompt the user equipment that the content is tampered if the illegal and tampered content exists, and re-detecting the content from the beginning if the illegal and tampered content does not exist in the current image through the multi-mode analysis module, and returning to the step S1. In step S1, image data being displayed on all flight display screens is collected by using an image acquisition module, including: (1) Setting a display screen set in an airport internal flight information display system as Each element in the c