METHOD, SYSTEM AND MACHINE-READABLE STORAGE MEDIA FOR DETECTING OBJECTS USING RECURRENT NEURAL NETWORK AND LINKED FEATURE MAP Russian patent published in 2019 - IPC G06K9/46 G06K9/62 

FIELD: physics.

SUBSTANCE: group of inventions relates to computer engineering and can be used to identify objects or visual features. System comprises a sensor component and a detection component. Sensor component is configured to receive a first data stream from the sensors and a second data stream from the sensors, wherein each of the first stream and the second stream contain a plurality of frames from the sensor. Detection component is configured to generate a linked feature map based on the frame from the first type sensor and the second type sensor frame, detecting one or more objects based on the linked feature card. One or more of the formation and detection involves generation or detection using a neural network with a recurrent link which provides information on features or objects from previous frames.

EFFECT: technical result is classification of objects in dynamic environment.

20 cl, 8 dwg