Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
  • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
  • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/088—Non-supervised learning, e.g. competitive learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/044—Recurrent networks, e.g. Hopfield networks
  • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks
  • G06N3/0455—Auto-encoder networks; Encoder-decoder networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/0475—Generative networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/094—Adversarial learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods
  • G06N3/096—Transfer learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
  • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V20/00—Scenes; Scene-specific elements
  • G06V20/50—Context or environment of the image
  • G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
  • G06V2201/07—Target detection

Definitions

• Artificial cognitive systems are equipped with sensors configured to capture physical quantities and collect information.
• the raw data collected by the different sensors can be merged before processing.
• the purpose of merging data from multiple sensors or multiple data sources is to combine that data so that the resulting information has less uncertainty than what would be obtained when these data sources are used individually. Reducing uncertainty may mean obtaining more precise, complete, or more reliable information, or may refer to the outcome of an emergent view, such as stereoscopic viewing (e.g., calculating depth information by combining two-dimensional images from two cameras at different points of view).
• the data fusion algorithm is derived from a machine learning algorithm.
• the invention further provides a decision support process for a cognitive system from data originating from a plurality of data sources, comprising the steps of:
• the embodiments of the invention allow, through the reconstruction of sensor data, the detection of faults or anomalies of the sensors and the generation of virtual sensor data from existing sensors.
• the embodiments of the invention offer increased reliability and precision thanks to the combination of all the information available to reconstruct the data of an improved sensor.
• Figure 1 is a diagram showing a device 10 for decision support of a cognitive system 1, according to certain embodiments of the invention.
• the cognitive system 1 can be part of or can be a tracking system, a detection system, a surveillance system, a navigation system, or an intelligent transport system.
• an air traffic control radar for example a primary radar or a secondary radar
• the cognitive system 1 can be implemented in an airborne surveillance system operating, for example, inside a surveillance aircraft.
• a data source 11-i can be any type of image or video acquisition device configured to acquire image streams or video streams from the environment in which the cognitive system operates 1.
• Cognitive system 1 to robotics include connected autonomous vehicles (Internet of vehicles and vehicle-to-vehicle communications or vehicle-to-infrastructure or vehicle and any object), home automation / smart home, smart cities, technology portable and connected health.
• the data coming from the different data sources 11-i, for i varying from 1 to N, can have different representations in different data representation spaces.
• the embodiments of the invention provide a device 10 for decision support of a cognitive system 1 from data from the plurality of data sources 11-i, with i ranging from 1 to N, implementing machine learning techniques for merging and reconstructing data from data sources 11-i in order to provide decision unit 12 with a representation of data based on from which the decision unit 12 can determine one or more actions to be implemented by the cognitive system 1.
• the last layer of each auto encoder implemented in each encoding unit 1030-i can produce an average and variance vectors characterizing a multivariate normal distribution of the same dimension for the set of encoding units 1030-i associated with the plurality of data sources 11-i.
• the model of representation of the environment (or even the modeling of the latent world) can be used by the decision unit 12 to determine an action to be implemented by the cognitive system 1 and / or can be used for other cognitive system 1 tasks such as understanding the context, trajectory planning or any other decision-making task.
• a processing unit 103-i associated with the data source 11-i, for i varying from 1 to N can further comprise a comparison unit 1032-i configured to compare the data from the data source 11-i with the reconstructed representation of data determined by the data reconstruction unit 1031 -i associated with the data source 11-i.
• the comparison makes it possible to detect random or systematic errors (for example, false alarms due to 'phantom' detections which can be highlighted by comparing the data coming from the data source with the reconstructed data), calibrate data sources (for example detecting misalignment of the support of a sensor due to a shock or estimating a delay), and more generally, detecting if a data source deviates from its nominal operating point using data reconstructed as a reference for uncorrupted nominal data calibrated at the factory).
• calibrate data sources for example detecting misalignment of the support of a sensor due to a shock or estimating a delay
• generative antagonist networks As discriminators to help the data reconstruction units 1031 -1 to 1031 -3 to provide images of better fidelity.
• These generative antagonist networks are for example hybridized with generative variational autoencoders.
• the reconstruction of data according to the embodiments of the invention also allows the transfer of information between the data sources, by extending the set of training data for a data source by exploiting the data of the data sources. other data sources. Data transfer also ignores the nature and position of data sources.
• the decision unit 12 can be configured to determine an action to be implemented by the cognitive system 1 according to the model of representation of the environment and / or according to the comparisons made between the original data coming from the data sources 11 - i with the reconstructed data representations determined by the data reconstruction units 1031-i, for i ranging from 1 to N and / or as a function of the reconstructed representations.
• the data reconstruction units 1031 -i can also be co-located with the data sources 11-i and the encoding units 1030-i, for i ranging from 1 to N.
• a return channel from the modeling of the a posteriori world from the data fusion unit 105 to the processing unit of the data source 11-i can be implemented.
• This architecture advantageously makes it possible to co-locate the comparison units 1032-i online, which can be configured to implement general tasks of monitoring data sources 11-i including for example dynamic activation / deactivation, calibration, time stamping, anomaly detection, and communication of the general state of the data source 11-i.
• the data coming from the data sources can comprise data generated in real time by the data sources, data previously processed, and / or contextual data.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Evolutionary Computation (AREA)
• General Health & Medical Sciences (AREA)
• Health & Medical Sciences (AREA)
• Artificial Intelligence (AREA)
• Software Systems (AREA)
• Computing Systems (AREA)
• Data Mining & Analysis (AREA)
• Computational Linguistics (AREA)
• Mathematical Physics (AREA)
• General Engineering & Computer Science (AREA)
• Life Sciences & Earth Sciences (AREA)
• Biomedical Technology (AREA)
• Biophysics (AREA)
• Molecular Biology (AREA)
• Multimedia (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Databases & Information Systems (AREA)
• Medical Informatics (AREA)
• Image Analysis (AREA)
• Traffic Control Systems (AREA)
• Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=70614252

Family Applications (1)

Country Status (5)

Families Citing this family (4)

Citations (1)

Family Cites Families (8)

• 2020
  • 2020-03-23 FR FR2002789A patent/FR3108423B1/fr active Active
• 2021
  • 2021-03-08 EP EP21709420.0A patent/EP4128055A1/de active Pending
  • 2021-03-08 US US17/907,221 patent/US12499670B2/en active Active
  • 2021-03-08 WO PCT/EP2021/055762 patent/WO2021190910A1/fr not_active Ceased
  • 2021-03-08 CN CN202180030076.7A patent/CN115461756A/zh active Pending

Patent Citations (1)

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events