Ricoh Company, Ltd.

Canon Kabushiki Kaisha

Other

Manuel Martinello

Mahdi Nezamabadi

Leonidas Spinoulas

Hector H Gonzalez-Banos

Noah Denis Bedard

Ivana Tosic

Evan G Levine

Sofia Karygianni

Pascal Frossard

Edward L Schwartz

Aashish Kumar

Jonathan Frieman Brand

Logical boundaries enclosing a physical area are defined. A segment of the logical boundaries is defined as a directional gate, wherein traversing the gate into the physical area is defined as an ingress and traversing the gate out of the physical area is defined as an egress. The directional gate is monitored, and ingresses and egresses are detected. An occupancy count of the physical area is maintained, based on monitoring the gate and detecting ingresses and egresses. One or more conditions are tracked in addition to the occupancy count. Artificial intelligence (AI) processing is applied to the maintained occupancy count and the additional tracked condition(s), in real-time as the monitoring, maintaining and tracking are occurring. One or more responsive actions are automatically taken as a result of applying the AI processing to the maintained occupancy count and the additional tracked condition(s).

Tracking conditions concerning an area to automatically generate artificial intelligence based responsive actions

Image data describing a physical space is received from one or more sensor(s). A corresponding image of the physical space is output to a user. User generated control signals are received, indicating to define one or more line(s) at specific coordinates in relation to the physical space. One or more line(s) are defined in response to the received control signals. Polarity is assigned to the line(s), wherein crossing a line in a first direction is defined as an ingress and crossing a line in a second, opposite direction is defined as an egress. An image of the line(s) is superimposed on the image of the physical space, wherein the image of the line(s) graphically indicates the assigned polarity. Physical objects are monitored in relation to the one or more line(s) in the physical space. An action is automatically taken in response to activity detected by the monitoring.

Detection-area fencing with polarity and object ingress/egress monitoring

The disclosure includes a system and method for providing visual analysis focalized on a salient event. A video processing application receives a data stream from a capture device, determines an area of interest over an imaging area of the capture device, detects a salient event from the data stream, determines whether a location of the detected salient event is within the area of interest, and in response to the location of the salient event being within the area of interest, identifies a portion of the data stream, based on the salient event, on which to perform an action.

On-demand visual analysis focalized on salient events

A method for person detection using overhead images includes receiving a depth image captured from an overhead viewpoint at a first location; detecting in the depth image for a target region indicative of a scene object within a height range; determining whether the detected target region has an area within a head size range; if within the head size range, determining whether the detected target region has a roundness value less than a maximum roundness value; if less than the maximum roundness value, classifying the detected target region as a head of a person and masking the classified target region in the depth image, where the masked region is excluded from detecting; and repeating the detecting to the masking to detect for and classify another target region in the depth image within the height range and outside of the masked region.

Person detection and identification using overhead depth images

A lightfield otoscope includes a housing with a tip configured to receive a disposable speculum. The otoscope also includes a microlens array, a sensor array and an optical train contained within the housing. The optical train includes an objective lens and a relay lens. The objective lens is positioned at least partially within the tip. The relay lens is used to relay an image plane of the objective lens to the microlens array and to relay a pupil plane of the objective lens to the sensor array. An active heating element is also contained within the housing and positioned to heat the front surface, thereby reducing fogging and/or condensation on the front surface.

Active lens defogging for digital imaging systems

Light-field data is masked to identify regions of interest, before applying to deep learning models. In one approach, a light-field camera captures a light-field image of an object to be classified. The light-field image includes a plurality of views of the object taken simultaneously from different viewpoints. The light-field image is pre-processed, with the resulting data provided as input to a deep learning model. The pre-processing includes determining and then applying masks to select regions of interest within the light-field data. In this way, less relevant data can be excluded from the deep learning model. Based on the masked data, the deep learning model produces a decision classifying the object.

Deep learning architectures for the classification of objects captured with a light-field camera

In one aspect, a color plenoptic imaging system captures a plenoptic image of an object. The plenoptic image is made up of a plurality of superpixels, and each superpixel includes a center subpixel. The collection of center subpixels from the plurality of superpixels forms a set of captured center view data for the object. The sensor array includes at least two arrays of different color sensors that capture subpixels of different colors. The microlens array and the sensor array are positioned such that, within the set of captured center view data, for each of the different colors, adjacent center subpixels of that color are separated by not more than three intervening center subpixels of a different color.

Reducing color artifacts in plenoptic imaging systems

Shape measurement of a specular object even in the presence of multiple intra-object reflections such as those at concave regions of the object. Silhouettes of the object are extracted, by positioning the object between a camera and a background. A visual hull of the object is reconstructed based on the extracted silhouettes, such as by image capture of shadows of the object projected onto a screen, and image capture of reflections by the surface of the object of coded patterns onto the screen. The visual hull is used to distinguish between direct (single) reflections of the coded patterns at the surface of the object and multiple reflections. Only the direct (single) reflections are used to triangulate camera rays and light rays onto the surface of the object, with multiple reflections being excluded. The 3D surface shape may be derived by voxel carving of the visual hull, in which voxels along the light path of direct reflections are eliminated. For surface reconstruction of heterogeneous objects, which exhibit both diffuse and specular reflectivity, variations in the polarization state of polarized light may be used to separate between a diffuse component of reflection and a specular component.

Shape reconstruction using electronic light diffusing layers (E-Glass)

A three-dimensional image of the eardrum is automatically registered. In one aspect, a three-dimensional image (e.g., depth map) of an eardrum is produced from four-dimensional light field data captured by a light field otoscope. The three-dimensional image is registered according to a predefined standard form. The eardrum is then classified based on the registered three-dimensional image. Registration may include compensation for out-of-plane rotation (tilt), for in-plane rotation, for center location, for translation, and/or for scaling.

Automatic eardrum registration from light field data

Eardrums are automatically classified based on a feature set from a three-dimensional image of the eardrum, such as might be derived from a plenoptic image captured by a light field otoscope. In one aspect, a grid is overlaid onto the three-dimensional image of the eardrum. The grid partitions the three-dimensional image into cells. One or more descriptors are calculated for each of the cells, and the feature set includes the calculated descriptors. Examples of descriptors include various quantities relating to the depth and/or curvature of the eardrum. In another aspect, isocontour lines (of constant depth) are calculated for the three-dimensional image of the eardrum. One or more descriptors are calculated for the isocontour lines, and the feature set includes the calculated descriptors. Examples of descriptors include various quantities characterizing the isocontour lines.

Automatic classification of eardrum shape

A plenoptic imaging system includes imaging optics, a microlens array and a sensor array. A method for generating a preview image of a plenoptic image captured by the plenoptic imaging system includes the following. A plenoptic image of an object captured by the plenoptic imaging system is accessed. The plenoptic image includes many superpixels. Each superpixel includes many subpixels. For each superpixel, a set of reliable subpixels is selected based at least in part on vignetting of the subpixels. A preview image is generated from the sets of reliable subpixels. The preview image includes pixels corresponding to the superpixels. The pixels in the preview image are generated from the set of reliable subpixels for the corresponding superpixels.

Preview generation for plenoptic imaging systems

A system includes cameras; a rotating stage, which has an axis of rotation; one or more computer-readable media; and one or more processors that are coupled to the one or more computer-readable media, the rotating state, and the cameras. Also, the one or more processors are configured to cause the system to control the cameras to capture respective images of a calibration target in a first pose, wherein the calibration target is mounted on the rotating stage; control the rotating stage to rotate through a predetermined angular increment to a second pose; control the cameras to capture respective images of the calibration target in the second pose; and calculate a transformation between two of the cameras based on the respective images of the calibration target in the first pose, on the respective images of the calibration target in the second pose, and on the predetermined angular increment.

Devices, systems, and methods for high-resolution multi-view camera calibration

3D shape reconstruction of objects including specular objects in which structured light patterns are projected toward an object which reflects the patterns onto multiple layers of EGlass arranged in spaced-apart relation with each other while each layer is in turn controlled to be in the diffuse mode with all others being in the transparent mode. Images are captured of the structured light pattern as reflected onto the EGlass layers by the mirror-like surface of the specular object. By projecting multiple different patterns, such as multiple different Gray code patterns, and by sequencing through each EGlass layer for each pattern, the 3D shape of the entirety of the visible surface of the object can be reconstructed by analysis of captured images of the distorted reflections of the patterns by the surface of the object. Embodiments are also described which can be used without reconfiguration for surface reconstruction of the shape of diffuse objects.

3D shape reconstruction using reflection onto electronic light diffusing layers

Devices, systems, and methods for estimating blur in an image obtain an image, identify pixels in the image that have intensity values that equal or exceed a first intensity threshold in a first color channel, and identify a trajectory based on the pixels in the image that have intensity values that exceed the first intensity threshold in the first color channel. Also, the devices, systems, and methods identify multiple instances of the trajectory in the image, identify an instance of the trajectory that includes pixels that have intensity values that do not equal or exceed a second intensity threshold in a second color channel, and generate a blur kernel based on the intensity values in the second color channel of the pixels in the instance of the trajectory.

Growing community of inventors

Mountain View, CA, United States of America

Manuel Martinello