WO2021089221A1 - Method and device for secure management of an image bank for aircraft landing assistance - Google Patents

Abstract

The invention concerns a method implemented by computer and a device for secure management of an image bank for aircraft landing assistance. The method comprises at least steps of: - receiving sensor data, the data relating to at least one image collected by at least one sensor equipping an aircraft, the aircraft belonging to an image provider community comprising a plurality of aircraft, each collected sensor image being tagged with at least timestamp and image provider identification information, and a means for validating the integrity of the data; - validating the integrity of the received data using a distributed consensus validation process; - creating a validated data block if the integrity of the data is validated, the validated block comprising at least one link to the sensor data; and - making the validated data block available in a network of distributed registers.

Description

DESCRIPTION

METHOD AND DEVICE FOR SECURE MANAGEMENT OF AN IMAGE BANK FOR AIRCRAFT LANDING ASSISTANCE

The invention relates to a method and a device for secure management of a bank of images used by aircraft landing assistance systems. The main field of use of the invention is that of so-called EVS (“Enhanced Vision System”) landing assistance systems based on electro-optical, infrared or radar imaging cameras or sensors. to film the airport environment during the landing of an aircraft.

[0002] The invention addresses the general problem of assisting the landing of aircraft on an airstrip, in conditions of reduced visibility, in particular because of difficult weather conditions, for example in fog. The standards impose rules for obtaining visibility during the landing phase. These rules result in decision thresholds which refer to the altitude of the aircraft during its descent phase. At each of these thresholds, identified visual cues must be obtained to continue the landing maneuver, otherwise it must be abandoned. Abandoned landing maneuvers represent a real problem for air traffic management and for flight planning. Before take-off, the ability to land at destination must be estimated on the basis of more or less reliable meteorological forecasts, and, if necessary, plan for fallback solutions.

[0003] The problem of landing aircraft in conditions of reduced visibility has been the subject of the development of several techniques which are currently in use.

One of these techniques is the ILS instrument landing system ("Instrument Landing System"). The ILS system is based on several radiofrequency equipment installed on the ground, at the landing strip, and a compatible instrument placed on board the aircraft. The use of such a guidance system requires expensive equipment and a specific qualification of the pilots. Moreover, it cannot be installed at all airports. This system is not generalized and it will probably be decommissioned in the future. Another alternative is the landing aid by GPS ("Global Positioning System"). Although having sufficient precision, the reliability of this solution is too low since it can easily - intentionally or not - be subject to interference. Also, its integrity is not guaranteed.

[0006] An EVS augmented vision technique is also employed. The principle is to use sensors that are more efficient than the pilot's eye in degraded weather conditions, and to embed the information collected by the sensors in the pilot's field of vision, by means of a head-up display or on the visor of a helmet worn by the pilot. This technique relies mainly on the use of sensors to detect the radiation from the lamps arranged along the runway and on the approach ramp. Incandescent lamps produce visible light, but they also emit in the infrared range. Sensors in the infrared range can detect this radiation and the detection range is better than that of humans in the visible range, during poor weather conditions. An improvement in visibility therefore makes it possible to a certain extent to improve the approach phases and limit abandoned approaches. However, this technique is based on parasitic infrared radiation from the lamps present in the vicinity of the runway. For the sake of lamp durability, the current trend is to replace incandescent lamps with LED lamps. The latter have a less extensive spectrum in the infrared range. A side effect is therefore to cause technical obsolescence of EVS systems based on infrared sensors.

[0007] An alternative to infrared sensors is to obtain images by a radar sensor, in centimeter or millimeter band. Certain frequency bands chosen outside the water vapor absorption peaks exhibit very low sensitivity to harsh weather conditions. Such sensors therefore make it possible to produce an image through fog, for example. However, even though these sensors have fine range resolution, they exhibit much coarser angular resolution than optical solutions. The resolution is directly related to the size of the antennas used, and it is often too coarse to obtain precise positioning of the landing strip at a sufficient distance to perform the resetting maneuvers. [0008] Whatever the method (EVS, Radar) used in a landing aid system, there is always an image processing to allow the progress of flight procedures which use these sensors to help the aircraft. crew to position themselves in relation to the airstrip or the landing area.

[0009] The current image processing operations are linked to a particular type of image for each sensor, and are carried out a priori, by calibration and experience. However, image processing is calibrated on the basis of a few airports only.

In some cases, flights are performed on a clear day during the day to calibrate the sensors in question. However, given the cost of flights to generate images, the number of flights remains very limited and therefore the image bank which groups together all the images collected remains small.

Furthermore, the image bank remains incomplete because it does not take into account the diversity of meteorological situations, the variability of the environment (such as the presence of temporary obstacles for example), and it is moreover not exhaustive on earth.

[0012] This then results in reduced reliability of the piloting or display computers which use imagery for flight operations.

[0013] There is therefore a need to improve image processing for air operations, by increasing the volume of data in the image bank in order to achieve a reliability threshold required both in terms of precision and. terms of geographic coverage.

[0014] More generally, an aim of the invention is to obtain a critical mass of images from sensors on board aircraft participating in a community of image suppliers, whether they are images. in the visible, in the infrared or radar images. The critical mass of images must be sufficient for image processing operations to function with a reliability that improves the safety of aeronautical operations based on the use of such sensors.

Another object of the invention is to meet the need to encourage participation in the enrichment of such an image bank. This goal is achieved by setting up fair and certain compensation mechanisms for contributors who provide the images or who provide the processes allowing the processing. images. Indeed, another obstacle to improving the content of the image bank is the low rate of contributors and there is a need to encourage all actors to participate in this update of the image data they produce. or that they use, in a collaborative way and without intermediary. The actors who create and / or manage image data can be quite varied including, without limitation, image sensor suppliers, aircraft manufacturers, image processing specialists, states (designers of navigation procedures) , researchers, airlines.

[0016] Beyond the question of non-existent remuneration for an image supplier, one of the reasons for the current low participation is linked to the question of the risk of the use of images whose origin is not certain and / or whose reliability is not validated. Also, there is a need for secure validation mechanisms for the data coming from an image supplier, the certain validation having to relate at least to the date of obtaining and availability of the images, and to the authentication of the images. the issuer.

[0017] Thus, an object of the invention is to overcome the drawbacks of known techniques.

[0018] To this end, the object of the invention is to meet the aforementioned needs by providing a solution for the secure management of an image bank for air operations, in particular for assistance with the landing of aircraft.

To obtain the desired results, a computer-implemented method is proposed for the secure management of an image bank for aircraft landing assistance, the method comprising at least steps of : receiving sensor data, the data relating to at least one image collected by at least one sensor equipping an aircraft, said aircraft forming part of a community of image suppliers comprising a plurality of aircraft, each sensor image collected being tagged with at least timestamp and image provider identification information, and a means to validate data integrity;

- validate the integrity of the data received according to a validation process by distributed consensus; - create a validated data block, if the integrity of the data is validated, said validated block comprising at least one link to said sensor data; and

- make the validated data block available in a network of distributed registers.

[0020] According to alternative or combined embodiments:

- The distributed consensus validation step involves implementing a consensus algorithm to validate the integrity of the data by at least two image providers among the community of image providers.

- the distributed consensus validation step includes a step of verifying the quality of the sensor images via validation rules depending on the image providers.

- the data further comprises a summary of content and additional data relating to, in particular the weather forecast, the state of the navigation systems at the time of the shots, the 3D position of the aircraft at the time of touchdown on the runway, the 3D position and the 3D orientation of the aircraft with respect to the earth and / or of the sensor which took the image (s).

- the method comprises after the data reception step, a step of storing the sensor data in a sensor image database, and where the validation step is done on the stored data.

the network of distributed registers is a chain of blocks comprising a plurality of nodes.

- the distributed consensus validation step is internal to the blockchain and consists of proof of work, and the provisioning step of said validated block consists of adding said validated block to the blockchain.

- the distributed consensus validation step further includes a payment step for the node (s) of the blockchain contributing (s) to the validation.

- the payment step consists in granting a synthetic score VS to a node of the block chain which contributes to the validation. - the distributed consensus validation step is performed by software code called a smart contract stored on and executed by the blockchain.

- The method further includes a step of using the validated sensor data to train machine learning algorithms and generate trained image processing algorithm models.

- machine learning takes place on deep neural networks, in particular networks of the CNN convolutional neural network type or of the GNN generative antagonist network type.

- the method further comprises a step of storing the models of trained image processing algorithms in a database of trained image processing algorithms and / or a step of writing the models in a chain of blocks .

- the writing step is to write a hash in the blockchain.

- The method further comprises a step of sending one or more models of trained image processing algorithms to a consumer entity that has issued a request to use the model.

- the use request is a purchase order placed on the blockchain where the models are written, or is made by a smart contract to which the consumer entity is subscribed.

- the sending step is to send a decryption key for a required algorithm model.

- the method further comprises a step of verifying the ability of the consumer entity to receive the required model or models.

the method further comprises a payment step of the VS type for a consumer entity which participates in the validation of sensor data.

Advantageously, the invention uses secure time-stamping mechanisms for making available in a database of georeferenced sensor images, new sensor images taken during flights by a plurality of so-called aircraft. “Image suppliers”, each image being associated with parameters of the corresponding flight, and in particular the parameters of 3D position and 3D orientation of the device at the time of the shooting. Advantageously, the validation of the images to be added to the image bank is validation by consensus. In particular, the validation of the images comprises the verification of the quality of the sensor images which must be made available in the database, and which is done via validation rules depending on the image suppliers. In particular, the validation rules depend on the quality of each image supplier, the quality covering the quality of aircraft and crew members for example for procedures. Advantageously, the validation rules can also take into account the quality of actors specializing in image processing who can also validate the quality of the shots and their usefulness for training image processing algorithms.

Advantageously, the images made available in the sensor image database can be used to train various artificial intelligence algorithms to classify the images, in particular by deep neural networks or "Deep Learning" in English.

Advantageously, the sensor image database as constituted according to the principles of the invention can be used to develop robust trained image processing algorithms and produce a database of trained algorithms, the training of said algorithms being variable according to the suppliers of the sensor images.

The sensor image database can be used to validate the robustness or weakness of different algorithms with respect to scenarios ("use cases" in English) considered problematic, by making it possible to run in parallel the different algorithms on the data set already present in the image base and to detect excessive differences between the different algorithms on the results provided.

The present invention will find many fields of application and in particular applications for detection:

- runway, runway outline, light bar, approach lights;

- obstacles near the ground, taxiing buildings, vehicles or people;

- birds, all types of drones; - power lines.

[0027] The invention also covers a computer program product comprising code instructions for performing the steps of the claimed method, when the program is executed on a computer.

[0028] The invention further covers a device for secure management of an image bank for aircraft landing assistance, the device comprising means for implementing the steps of the claimed method.

[0029] Other characteristics, details and advantages of the invention will become apparent on reading the description given with reference to the accompanying drawings given by way of example and which represent, respectively:

[0030] [Fig.1] a "producer / validator" architecture of a secure management system of an image bank according to one embodiment of the invention;

[0031] [Fig.2] the steps of a method for providing a validated data block according to one embodiment of the invention;

[0032] [Fig.3] an architecture for processing sensor images of a secure management system of an image bank according to one embodiment of the invention;

[0033] [Fig.4] the steps of a method of providing trained image processing algorithms according to one embodiment of the invention;

[0034] [Fig.5] a "consumer" architecture of a secure management system of an image bank according to one embodiment of the invention;

[0035] [Fig.6] the steps of a method for using image processing algorithms trained according to one embodiment of the invention;

[0036] [Fig.7] an architecture of a secure management system of an image bank according to one embodiment of the invention.

FIG. 1 illustrates a so-called “producer / validator” architecture 100 of a system for secure management of an image bank according to the principles of the invention. The “producer / validator” system 100 according to one embodiment of the invention comprises a community of image suppliers 102. A community of image suppliers within the meaning of the present invention is understood to mean as a set of devices 102 where each device is able to capture and supply one or more images taken by one or more sensors. In one embodiment relating to assistance in the landing of aircraft, the images of an image supplier are images taken during a flight of an aircraft equipped with sensors. The community of image suppliers is then a fleet of aircraft, an aircraft within the meaning of the present invention taking a generic definition covering any flying device, whether it is an airplane, a drone, a balloon, whether it is piloted or no.

The system 100 comprises a distributed register type network 104 or DLT ("Distributed Ledger Technology" in English) which is composed of a plurality of computing entities (processors, computers) where a register is simultaneously recorded and synchronized on network entities. The network can evolve by adding new information previously validated by the entire network, and updating a distributed register has repercussions on the entire network. Each device or entity in the network always has the latest version of the registry.

In a particular embodiment, the network 104 is a chain of blocks ("blockchain" in English) where each block is linked to the previous one by a hash key. A blockchain is a distributed database that is secured by cryptographic techniques. The transactions exchanged are grouped into "blocks" at regular time intervals, securely by cryptography, and form a chain.

[0040] In an alternative embodiment, the system 100 includes an image database 106 which stores images collected and sent by image providers.

[0041] The system 100 further includes a data integrity validation module 108, coupled to the network 104. In one embodiment, the validation module is also coupled to the image database 106.

The validation module 108 is configured to allow collaborative validation of the data. Advantageously, the validation module implements a validation algorithm by distributed consensus. The validated data is made available in the DLT network 104 or according to an alternative embodiment in the blockchain. In one embodiment of the invention with a block chain, the validation by distributed consensus is internal to the block chain and can consist of a proof of work. The blockchain can be public or private, or according to intermediate governance, which can use different barriers to entry including validation by proof of work. A “public” blockchain works without a trusted third party (“trustless” model), generally with a validation by complex proof of work (eg hashcash). A public blockchain does not generally define any rule other than that of the code constituted by the protocol and software technology that composes it. A "private" blockchain includes nodes participating in the consensus which are defined in advance and then authenticated. Its operating rules may possibly be extrinsic.

FIG. 2 shows the steps of a method 200 for making validated data available in a DLT network 104 according to one embodiment of the invention. The method 200 corresponds to a so-called "Producer / Validator" configuration such as that illustrated in FIG. 1, where an image supplier publishes in the DLT network data relating to sensor images that it has collected. In one embodiment, data relating to the sensor images are stored in the image database 106, and a link is written in a chain of blocks containing for example an SHA (“Secure Hash Algorithm” in English) of the. 'picture.

Each image collected is labeled, and the data is sent 202 with at least time-stamping information (eg date of creation, validity) and identification of the image supplier, and at least one means of validating the data integrity (eg the association of a hash number to the data). The data can also include a summary of the content (eg parameters, units, quantity, image format) as well as additional data, such as the weather forecast, the state of the navigation systems at the time of the shots, the position 3D upon touchdown on the track, etc. Advantageously, the supplier can register the 3D position and the 3D orientation with respect to the earth of the aircraft and / or of the sensor that took the image (s). This set forms a “block” to be validated.

In one embodiment, the sensor images are a sequence of images produced at a fixed time interval. In an alternative, the images sensors are a sequence of images produced at a fixed distance from a reference element (eg a runway threshold). Advantageously, image compression algorithms can be applied, the data produced can be masked (by encryption for example).

[0047] In one embodiment, the data is sent in real time by the image provider over a digital data link. In an alternative, the data is sent once the aircraft has landed (e.g. via a "Gatelink" link when the aircraft is on a WiFi link at its door). In another alternative, the data is stored on board the aircraft and retrieved by downloading by an operator.

The method 200 makes it possible in a following step 204 to validate the data according to a collaborative process which allows the DLT network to reach a consensus. In one embodiment, the validation step includes performing a consensus algorithm that involves at least two image providers from the sensor image provider community.

[0049] In one embodiment, the data stored in the image database can be encrypted after the step of verifying their integrity and the authenticity of the producer.

In one embodiment to improve the robustness of the system to cheating, the validation step makes it possible to cross-reference the data from the image database (ie the identification of the image supplier) with data. external (for example with information if the flight is well scheduled at this time at this airport for this identifier).

In one embodiment of a block chain, the step of collaborative validation of the sensor images comprises a step of remunerating the contributor (s) to the validation (ie the node (s) of the chain of blocks). The compensation step can consist in granting a synthetic score VS (“Value Score” in English) to a node participating in the validation, the score making it possible to govern access to exchanges (as producer / supplier of images and / or validator and / or consumer).

In one embodiment, this score VS can be a function of i) the "value" of the information that it produces VS_PROD and ii) of the "value" of the information that it produces. consumes VS_CONS. VS value can be expressed as a score (cipher), symbol, value in cryptocurrency, real currency (fiat), etc. If the images are declared invalid, the blockchain is updated with a decrease in the VS_PROD by a certain amount for the producer, and an increase in the VS_PROD for the one who detected the anomaly. If the images are declared valid, the blockchain is updated with an increase in the VS_PROD by a certain amount for the validator node.

Returning to step 204 of validation, if the images are considered to be intact and therefore validated, the method makes it possible to create 206 a validated data register or block, then to send 208 the validated block to the DLT network 104 In one blockchain embodiment, the validated block that is created contains a link to the data.

In one embodiment of a blockchain, the step of collaborative validation of the sensor images can be carried out directly and automatically by a software code called a smart contract ("Smart Contract" in English) stored on and executed by the chain of blocks. With each new entry in the database (and attempt to write a block), a contract can check:

- if the producer is declared (if it is a consortium chain);

- include the producer's VS score in the assessment of its integrity;

- to include a "mark" in this same evaluation;

- perform functional checks on the data: data from the aircraft concerned can be correlated / cross-referenced with other sources of information on the aircraft (public or private sources from the company, air traffic control, etc.) emitted in real time and corresponding to the state of the aircraft (take-off time, status (in flight, on the ground, in cruise, etc.), current position, etc.).

As known from blockchain technologies, a smart contract is software code that is stored and executed on / by a blockchain and is triggered by external data, which allows other data to be changed, in blockchain or whatever. A smart contract ensures equal treatment (in value) between the parties to the contract. With compute resources distributed, data shared across the blockchain, executing a smart contract is inherently safe: the smart contract code is replicated in several nodes of the architecture implementing the blockchain, and being deterministic, the results of the different executions are identical. The code as well as the execution of the code are then safe.

As with any program or computer code, different programming languages are available for a smart contract, with different security and regulatory models (framework contract governing other contracts, cascade contracts, etc.). The forms taken by smart contracts can be diverse (e.g services, agents, snippets, scripts, SOA, API, add-ons, plug-ins, extensions, DLC, etc.). Mathematical logic (decision-making operating on data) can be that of classical logic, fuzzy, combinatorial, intuitionist, modal, propositional, partial, para-consistent, etc. or a combination of these logics. The software can be coded in part or in whole in hardware form (e.g. FPGA). A smart contract can be in whole or in part open source ("open source") and / or closed source ("closed source"). In the case of open source, the code is auditable or verifiable by the parties or third parties. A smart contract can combine open source parts (e.g. auditable, verifiable, improvable, etc.) with closed parts (proprietary, secret, sensitive, etc.). A closed source can be binary, possibly obfuscated or "hardened". Cryptographic techniques can be diverse: symmetric, asymmetric, “post-quantum”, “quantum-safe”, with the use of “Quantum-Key-Distribution”, etc.). A smart contract can be human and / or machine readable.

FIG. 3 illustrates an architecture 300 allowing the processing of sensor images and FIG. 4 shows the steps of a method 400 for making available image processing algorithms trained in a secure management system of an image bank according to one embodiment of the invention. System 300 includes a distributed register network 104 to which can connect a plurality of so-called image processing specialist entities 302, configured to train neural networks, and produce trained image processing algorithms which can be connected. stored in a database 304 of trained image processing algorithms. Advantageously, the DLT network 104 contains, according to the principle of the method 200 described above, blocks validated data. The method 400 begins when N (N> = 1) entities 302 decide to request or directly receive 402 one or more validated data blocks.

[0058] In the next step 404, the N entities specializing in image processing implement automatic learning algorithms on the corresponding data.

Different types of automatic learning (or machine) are possible. Machine learning is an area of computing that uses statistical techniques to give computer systems the ability to "learn" from data (for example, to gradually improve the performance of a specific task), without be explicitly programmed for this purpose.

Neural networks (hardware realization of machine learning, or software emulation) can be used to process new data. Machine learning can be performed on particularly large data, that is, using as much data as possible (e.g. stability, convergence, weak signals, etc.). New data can be added constantly and learning can be refined.

[0061] Different learning algorithms can be used, in combination with the features according to the invention. The method may comprise one or more algorithms from among the algorithms comprising: “support vector machines” or “wide margin separators” (in English “Support Vector Machine”, acronym SVM); "boosting" (classifiers); neural networks (in unsupervised learning); decision trees ("Random Forest"), statistical methods such as the Gaussian mixture model; logistic regression; linear discriminant analysis; and genetic algorithms.

Materially, according to embodiments, the method according to the invention can be implemented on or by one or more neural networks. A neural network according to the invention can be one or more neural networks chosen from neural networks comprising: a) an artificial neural network (in English “feedforward neural network”; b) an acyclic artificial neural network, eg a multi-layered perceptron, thus distinguishing itself from recurrent neural networks; c) a forward propagation neural network; d) a network Hopfield neurons (a discrete-time recurrent neural network model whose matrix of connections is symmetrical and zero on the diagonal and where the dynamics are asynchronous, with only one neuron being updated at each unit of time); e) a recurrent neural network (made up of interconnected units interacting non-linearly and for which there is at least one cycle in the structure); f) a convolutional neural network (in English "CNN" or "ConvNet" for "Convolutional Neural Networks", a type of feedforward acyclic artificial neural network, by multilayer stacking of perceptrons) or g) a generative antagonist network (in English " Generative Adversarial Networks ”acronym GAN, which is a class of unsupervised learning algorithm).

In one embodiment, a GAN network can be used to train and robustify the algorithms to cope with different situations: failure or cyber attack on the sensors, on the avionics system / computers, etc. Indeed, currently it is impossible to recover data that would correspond to such situations. GANs make it easy to generate large amounts of data, so GANs can be very useful for training machine learning algorithms.

[0064] The purpose of a GAN is to generate data which is very similar to that of a training game. To do this, the generator will generate "false" images which deceive the discriminator into believing that they are real. The discriminator must determine as best he can which images are "real" and which are "false". As the training progresses the two modules improve (the discriminator and the generator) until the discriminator is no longer able to distinguish the false images from the true ones, and it then produces predictions with 50 % precision (equivalent to randomly guessing the nature of the image). The quality of the images produced by the generator is then very similar to those of the game used for training.

Some examples of improving the robustness of algorithms using a GAN can be:

- to create "fake track data" using a GAN with a generator; - to enrich the data ("Data Enhancement"), the GANs making it possible for example to easily modify the weather in an image, we can consider that for each image (or other type of data) added to the image database , a GAN is implemented to automatically generate the same image with different weather;

- train algorithms to detect impossible situations by simulating these situations and generating the corresponding data;

- create "Adversarial Inputs" to study the reaction of algorithms and minimize the impact of a potential attack / failure of a sensor or a component of the functional chain.

Returning to the method 400 of FIG. 4, when the step 404 of training the image processing algorithms is completed, the method allows 406 to generate models of trained algorithms and to store them in a database. data 304 of trained image processing algorithms for use by users / co n so mm a rs.

In one embodiment, the results are artificial intelligence models of the "deep neural networks" type with their parameters (weight), their reliability (precision, recall).

[0068] In an alternative, the image processing can be standard and relate to the detection of contour, right, contrast.

In another alternative, the image processing algorithm can detect a level of interest in new images compared to an existing stock of images, and perform image rating operations by associating a tag to an image considered uninteresting or of low value.

[0070] The method 400 ends with the provision 408 of models of trained image processing algorithms. In one embodiment, the models of the trained algorithms are made available in a distributed register type network, which advantageously can be the same network 104 as that of the image providers.

In one embodiment, the models of the trained algorithms are written in a chain of blocks. In an alternative, the models are written in a database of trained image processing algorithms and a hash is written to a blockchain.

FIG. 5 illustrates a so-called “consumer” architecture 500 allowing the use of trained image processing algorithms and FIG. 6 shows the steps of a method 600 for using models of processing algorithms. images trained in a secure management system of an image bank according to one embodiment of the invention.

The system 500 comprises a database 304 storing models of trained image processing algorithms coupled to a distributed register type network 504. The method 600 of using trained image processing algorithms begins when a 'consumer' entity 502 sends a request for use 602 to the DLT network 504. In order to know if a data block can interest it and therefore make a request for use of a trained algorithm model, information such as the format and the summary of the contents of a network block (ie encrypted data) can be left unencrypted in a storage space.

[0074] In a blockchain embodiment, a consumer may subscribe or express an interest in retrieving a trained image processing algorithm model or optimized parameters by placing a purchase order on the chain. block 504 or by a smart contract. The contract reads the data in clear and is triggered if the consumer is subscribed to this data and if his balance (Value Score) is positive. If so, the contract can issue a data recovery request which results in creating a new block sending a decryption key for the required block.

If the data set corresponding to the buyer's request is deemed valid, decryption keys stored in the blockchain are retrieved by the contract and transmitted to the buyer. The transaction is written into the blockchain by the contract.

Advantageously, a consumer can send a request for several models of trained image processing algorithms, in order to improve the precision, integrity and availability or for redundancy issues (by implementation of a dissimilarity technique for example). After receiving the request from a consumer, the method makes it possible to check 604 whether the latter has the capacity required to download the trained model or the optimized parameters.

In an embodiment with a block chain where a consumer participates in the validation of blocks for the sensor images and obtains a VS type remuneration, the verification relates to the value of the VS of the consumer who must authorize him to recover the trained model (s).

[0079] In an alternative, a purchase order placed on the blockchain can be made available to validating entities which perform the verification of the VS. Reward and validation mechanisms can be activated when the verifications are considered valid.

[0080] If the request or transaction is valid, the required algorithm, or the optimized data set for an algorithm that the user already has, is passed 606 to the consumer. Advantageously, the latter can have the possibility of verifying its integrity (by the hash recovered in the blockchain).

In an alternative embodiment, the consumer can note the attractiveness of the block received (vis-à-vis his own interest). Scoring mechanisms can also be activated to reward the most useful data transmitters.

In another variant, a score can be calculated for a block, by counting the number of downloads of the images and / or by the number of interested consumers.

In another variant, a score can be set to meet regulatory needs, such as, for example, the validation of a new navigation procedure at a large frequented airport can have a high score for the images provided.

In another variant, a score may be set to meet international needs for global coverage, which may be the case for underserved airports.

FIG. 7 illustrates the architecture of a system 700 for secure management of an image bank for aid in the landing of aircraft according to a preferred embodiment of the invention with chains of blocks. The general architecture of the system is a Supplier / Consumer type architecture which implements the methods described in Figures 2, 4 and 6.

A community of sensor image suppliers 702 provides new data built on sensor images taken by sensors during flights, in particular during landing phases, and stored in a database 706. The data is validated by a consensus validation mechanism 708.

A smart provisioning contract detects the new validated data and stores metadata in a blockchain 704 (i.e. an identification, a date, an issuer, a score, a hash of the content of the data).

The data is used by a community 710 of image processing specialists to train sensor image processing algorithms and generate trained algorithms, stored in a database 712 of sensor image processing algorithms trained.

When purchasing data, a buy order is placed by a consumer 714 on a blockchain 716, and a data consumption smart contract is triggered. The contract checks the value score of the potential buyer / consumer, optionally verifies the reality of the data provided, downloads it and sends it to the buyer. The consumer contract also sends the information of the corresponding block of the blockchain, and the hash of the data content (which allows the buyer to verify the integrity of the data).

In one embodiment, the consumer contract indicates to the buyer the position of the data in the shared database and the keys for decrypting the data. In an alternative mode, the contract re-encrypts the data in the database so that each consumer pays and to avoid key sharing.

In an embodiment without a smart contract, a supplier can receive a purchase order and send the required data directly to the buyer. It writes the transaction (with date and hash) to the blockchain. On receipt, the buyer checks the block (with the hash) and validates the block or rejects it. Each network member is assigned a value score (VS) to enable them to participate to exchanges (as a producer and / or consumer of data). Scores are updated based on the result. A VS score is a function of:

- the value of the information it produces VS_PROD; and

- the value of the information it consumes VS_CONS.

The VS value can be expressed as a score, a cryptocurrency, a real currency. Advantageously, the value VS can be modified by conversion into a monetized value, when:

- an actor buys an amount VS_MONEY of right to consume in order to subsequently consume data (which can be useful if he consumes more than he produces);

- an actor transforms his SV into real currency or cryptocurrency for other uses, which decreases his SV by the transformed value VS_MONEY.

The VS score is required to change depending on the attractiveness of the data it produces (number of subscribers for example or number of downloads or quantity of downloads), the value of the data it consumes, and purchases / sales in the VS_MONEY alternative.

The VS_PROD and VS_CONS values can be calculated by one or a combination of the following methods:

- value in quantity (in Megabyte for example);

- value in quality (depending on the quality of the image, the rarity, the angle of view);

- consensus on a price (proposal by a consumer, validation by the producer or proposal by a producer vs. acceptance / rejection by a consumer, or negotiation);

- history of use;

- daily price depending on various parameters;

- a priori fixing of min / max terminals;

- a priori consensus on the values of the various data between actors;

- value related to the quality of the image processing algorithm; - value linked to the importance of the image to configure the image processing algorithm.

These values or methods of calculating these values are entered in a smart contract. The contract can be bi-partite (agreement between two players) or multi-party (agreement between 1 to N suppliers and 1 to M consumers).

The blockchain contains in real time the VS score for each actor in the network: when an actor’s score is updated, the blockchain is updated too. Advantageously, the use of the blockchain makes it possible to have this information distributed over several nodes and therefore to make it immutable and secure as well as to have a certain history.

When a consumer retrieves a data set in the blockchain, after checking the integrity (recalculated hash), the supplier's VS is updated (adding a sum if the block is validated, removing of a sum if the dataset was corrupted) and its own VS (of an inverse value to that of the sender) is updated. This transaction is written to the blockchain.

Each supplier can in turn publish information to other members of the network or "subscribe to the network" via a smart contract for example, and receive information about it automatically.

The system can be hosted in a “cloud” or on servers (nodes of the block chain) and is accessible on multiplatforms (EFB, WebbApp On the ground, etc.).

[0100] The present description illustrates a preferred implementation of the invention, but which is not limiting. Examples are chosen to allow a good understanding of the principles of the invention and a concrete application, but are in no way exhaustive and should allow those skilled in the art to make modifications and variants of implementation while keeping the same. principles.

[0101] The invention can be implemented from hardware and / or software elements. It may be available as a computer program product on computer readable medium and includes code instructions for performing the steps of the methods in their various embodiments.

Claims

1. A method (200) for secure management of an image bank for aircraft landing assistance, the method comprising at least the steps of: - receiving (202) sensor data, the data relating to at least one image collected by at least one sensor equipping an aircraft, said aircraft forming part of a community of image suppliers (102) comprising a plurality of aircraft , each collected sensor image being tagged with at least time stamping and identification information from the image provider, and a means of validating the integrity of the data; - validate (204) the integrity of the data received according to a collaborative validation process implementing a consensus algorithm within a network of distributed registers (108); - create 206 a validated data block, if the integrity of the data is validated, said validated block comprising at least one link to said sensor data; and - make available 208 the validated data block in the network of distributed registers 104. 2. The method of claim 1 wherein the validation step comprises implementing a consensus algorithm to validate the integrity of the data by at least two image providers among the community of image providers. 3. The method according to claim 1 or 2 wherein the validation step comprises a step of verifying the quality of the sensor images via validation rules depending on the image providers. 4. The method according to any one of the preceding claims wherein the data further comprises a summary of content and additional data relating to, in particular the weather forecast, the state of the navigation systems at the time of the shots, the position. 3D of the aircraft at the moment of touchdown on the track, the 3D position and the 3D orientation of the aircraft with respect to the earth and / or of the sensor which took the image (s). 5. The method according to any one of the preceding claims comprising, after the step of receiving the data, a step of storing the sensor data in a database of sensor images (106), and in which the validation step takes place. done on the stored data. 6. The method of any preceding claim wherein the distributed ledger network is a blockchain comprising a plurality of nodes. 7. The method of claim 6 wherein the validation step is internal to the blockchain and consists of a proof of work, and the step of providing said validated block consists of adding said validated block to the chain. of blocks. 8. The method of claim 6 or 7 wherein the validation step further comprises a payment step or nodes of the contributor blockchain (s) to the validation. 9. The method of claim 8 wherein the retribution step consists of granting a synthetic score VS to a node of the blockchain which is a contributor to the validation. 10. The method according to any one of claims 6 to 8 wherein the validation step is performed by software code called a smart contract stored on and executed by the blockchain. The method of any preceding claim further comprising a step of using the validated sensor data to train machine learning algorithms and generate trained image processing algorithm models. 12. The method of claim 11 wherein the machine learning is performed on deep neural networks, in particular networks of CNN convolutional neural network type or of GNN generative antagonist network type. 13. The method according to claims 11 or 12 further comprising a step of storing the trained image processing algorithm models in a database of trained image processing algorithms and / or a writing step. the models in a blockchain. 14. The method of claim 13 wherein the writing step consists of writing a hash in the blockchain. 15. The method according to any one of claims 11 to 14 further comprising a step of sending one or more models of trained image processing algorithms to a consumer entity having sent a request for use of the model. . 16. The method of claim 15 wherein the request for use is a purchase order placed on the blockchain where the models are written, or is made by a smart contract to which the consumer entity is subscribed. 17. The method of claim 16 wherein the sending step comprises sending a decryption key for a required algorithm model. 18. The method according to any one of claims 15 to 17 further comprising a step of verifying the ability of the consumer entity to receive the required model or models. 19. The method according to any one of claims 15 to 18 further comprising a VS type payment step for a consumer entity which participates in the validation of sensor data. 20. A device (700) for secure management of an image bank for aircraft landing assistance, the device comprising means making it possible to implement the steps of the method for secure management of a bank. of images for the aircraft landing aid according to any one of claims 1 to 19. 21. Computer program comprising code instructions for executing the steps of the method for secure management of an image bank for aircraft landing assistance according to any one of claims 1 to 19. , when said program is executed by a processor.