Abstract: An anomaly detector uses two neural networks, the first, a general purpose classifying convolutional neural network operates as a teacher neural network, while a second neural network in an auto-encoder type configuration. Each of the two neural networks receives the same input stream, and generates respective feature outputs at different levels, corresponding to different resolutions for image data. The respective outputs of the two neural networks are compared at each level, and the resulting difference values consolidated across the difference levels to obtain a final difference value. In a training phase this difference value is used to drive the determination of the weights and biases of the auto-encoder, so as to obtain a auto-encoder trained for a particular input type, under the influence of the teacher neural network. In an operational mode, the difference value is compared to a threshold to determine whether a particular sample is anomalous or not.

Abstract: A device comprising at least one processing logic configured for: obtaining an input vector representing an input data sample; until a stop criterion is met, performing successive iterations of: using an autoencoder trained using a set of reference vectors to encode the input vector into a compressed vector, and decode the compressed vector into a reconstructed vector; calculating a reconstruction loss between the reconstructed and the input vectors, and a gradient of the reconstruction loss; updating said input vector for the subsequent iteration using said gradient.

Abstract: The drone comprises a camera, an inertial unit measuring the drone angles, and an extractor module delivering data of an image area (ZI) of reduced size defined inside a capture area (ZC) of the sensor. A feedback-control module dynamically modifies the position and the orientation of the image area inside the capture area, in a direction opposite to that of the angle variations measured by the inertial unit. The sensor may operate according to a plurality of different configurations able to be dynamically selected, with a base configuration using a base capture area (ZCB) for low values of roll angle (?), and at least one degraded mode configuration using an extended capture area (ZCE) of greater size than the base capture area (ZCB), for high values of roll angle (?).

Abstract: The drone comprises a camera (14) having a rolling shutter digital sensor which sends video data (l) line by line. An inertial unit (26) sends a gyrometric signal representative of the variations in attitude (?, ?, ?) of the camera at a given instant. An image processing module (30) comprising an anti-wobble module receives the video data (l) and the gyrometric signal as inputs, and outputs video data processed and corrected for artifacts introduced by the vibrations of the motors of the drone. A complementary filtering module (36) applies a predetermined compensating transfer function to the gyrometric signal at the input of the anti-wobble module, which transfer function is an inverse transfer function of the frequency response of the gyrometric sensor of the inertial unit.

Abstract: The drone comprises a camera, an inertial unit measuring the drone angles, and an extractor module delivering data of an image area (ZI) of reduced size defined inside a capture area (ZC) of the sensor. A feedback-control module dynamically modifies the position and the orientation of the image area inside the capture area, in a direction opposite to that of the angle variations measured by the inertial unit. The sensor may operate according to a plurality of different configurations able to be dynamically selected, with a base configuration using a base capture area (ZCB) for low values of roll angle (?), and at least one degraded mode configuration using an extended capture area (ZCE) of greater size than the base capture area (ZCB), for high values of roll angle (?).

Abstract: The drone comprises a camera, an inertial unit measuring the drone angles, and an extractor module delivering data of an image area (ZI) of reduced size defined inside a capture area (ZC) of the sensor. A feedback-control module dynamically modifies the position and the orientation of the image area inside the capture area, in a direction opposite to that of the angle variations measured by the inertial unit. The sensor may operate according to a plurality of different configurations able to be dynamically selected, with a base configuration using a base capture area (ZCB) for low values of roll angle (?), and at least one degraded mode configuration using an extended capture area (ZCE) of greater size than the base capture area (ZCB), for high values of roll angle (?).

Abstract: The apparatus comprises a camera (10) with a digital sensor read by a mechanism of the rolling shutter type delivering video data (Scam) line by line. An exposure control circuit (22) adjusts dynamically the exposure time (texp) as a function of the level of illumination of the scene that is captured. A gyrometer unit (12) delivers a gyrometer signal (Sgyro) representative of the instantaneous variations of attitude (?, ?, ?) of the camera, and a processing circuit (18) that receives the video data (Scam) and the gyrometer signal (Sgyro) delivers as an output video data processed and corrected for artefacts introduced by vibrations specific to the apparatus. An anti-wobble filter (24) dynamically modifies the gain of the gyrometer signal as a function of the exposure time (texp), so as to reduce the gain of the filter when the exposure time increases, and vice versa.

Abstract: A vertical-view camera (16) delivers an image signal (ScamV) of the ground overflown by the drone. Gyrometric sensors (102) measure the Euler angles (?, ?, ?) characterizing the attitude of the drone and delivering a gyrometric signal (Sgyro) representative of the instantaneous rotations. Rotation compensation means (136) receive the image signal and the gyrometric signal and deliver retimed image data, compensated for the rotations, then used to estimate the horizontal speeds of the drone. The camera and the inertial unit are piloted by a common clock (160), and it is provided a circuit (170) for determining the value of the phase-shift between the gyrometric signal and the image signal, and to apply this phase-shift value at the input of the rotation compensation means (136) to resynchronize the image signal onto the gyrometric signal before computation of the retimed image data.

Abstract: The drone comprises a camera (14) having a rolling shutter digital sensor which sends video data (l) line by line. An inertial unit (26) sends a gyrometric signal representative of the variations in attitude (?, ?, ?) of the camera at a given instant. An image processing module (30) comprising an anti-wobble module receives the video data (l) and the gyrometric signal as inputs, and outputs video data processed and corrected for artifacts introduced by the vibrations of the motors of the drone. A complementary filtering module (36) applies a predetermined compensating transfer function to the gyrometric signal at the input of the anti-wobble module, which transfer function is an inverse transfer function of the frequency response of the gyrometric sensor of the inertial unit.

Abstract: The drone (10) comprises a camera with a hemispherical-field lens of the fisheye type pointing to a fixed direction (?) with respect to the drone body. A capture area (36) of reduced size is extracted from the image formed by this lens (42), the position of this area being function of the signals delivered by an inertial unit measuring the Euler angles characterizing the attitude of the drone with respect to an absolute terrestrial reference system. The position of this area is dynamically modified in a direction (44) opposite to that of the changes of attitude (38) of the drone detected by the inertial unit. The raw pixel data are then processed to compensate for the geometric distortions introduced by the fisheye lens in the image acquired in the region of the capture area.

Abstract: A vertical-view camera (16) delivers an image signal (ScamV) of the ground overflown by the drone. Gyrometric sensors (102) measure the Euler angles (?, ?, ?) characterizing the attitude of the drone and delivering a gyrometric signal (Sgyro) representative of the instantaneous rotations. Rotation compensation means (136) receive the image signal and the gyrometric signal and deliver retimed image data, compensated for the rotations, then used to estimate the horizontal speeds of the drone. The camera and the inertial unit are piloted by a common clock (160), and it is provided a circuit (170) for determining the value of the phase-shift between the gyrometric signal and the image signal, and to apply this phase-shift value at the input of the rotation compensation means (136) to resynchronize the image signal onto the gyrometric signal before computation of the retimed image data.

Abstract: The drone comprises a camera, an inertial unit measuring the drone angles, and an extractor module delivering data of an image area (ZI) of reduced size defined inside a capture area (ZC) of the sensor. A feedback-control module dynamically modifies the position and the orientation of the image area inside the capture area, in a direction opposite to that of the angle variations measured by the inertial unit. The sensor may operate according to a plurality of different configurations able to be dynamically selected, with a base configuration using a base capture area (ZCB) for low values of roll angle (?), and at least one degraded mode configuration using an extended capture area (ZCE) of greater size than the base capture area (ZCB), for high values of roll angle (?).

Abstract: The apparatus comprises a camera (10) with a digital sensor read by a mechanism of the rolling shutter type delivering video data (Scam) line by line. An exposure control circuit (22) adjusts dynamically the exposure time (texp) as a function of the level of illumination of the scene that is captured. A gyrometer unit (12) delivers a gyrometer signal (Sgyro) representative of the instantaneous variations of attitude (?, ?, ?) of the camera, and a processing circuit (18) that receives the video data (Scam) and the gyrometer signal (Sgyro) delivers as an output video data processed and corrected for artefacts introduced by vibrations specific to the apparatus. An anti-wobble filter (24) dynamically modifies the gain of the gyrometer signal as a function of the exposure time (texp), so as to reduce the gain of the filter when the exposure time increases, and vice versa.

Abstract: The drone (10) comprises a camera with a hemispherical-field lens of the fisheye type pointing to a fixed direction (?) with respect to the drone body. A capture area (36) of reduced size is extracted from the image formed by this lens (42), the position of this area being function of the signals delivered by an inertial unit measuring the Euler angles characterizing the attitude of the drone with respect to an absolute terrestrial reference system. The position of this area is dynamically modified in a direction (44) opposite to that of the changes of attitude (38) of the drone detected by the inertial unit. The raw pixel data are then processed to compensate for the geometric distortions introduced by the fisheye lens in the image acquired in the region of the capture area.