A system for plant leaf identification includes: a plant image capturing unit which captures an image of a target plant to generate a plant image; a plant area image extraction unit which separates a background area and a plant area in the plant image to generate a plant area image including the plant area; a plant area image skeletonization unit which skeletonizes the plant area image to generate a skeletonized plant area image; a candidate leaf path generation unit which identifies a root vertex, a junction vertex and a leaf tip vertex in the skeletonized plant area image, and generates a plurality of candidate leaf paths by calculating all possible paths from the root vertex to the leaf tip vertex; and a final leaf path reconstruction unit which reconstructs a final leaf path matching the plant image by selecting the plurality of candidate leaf paths.

A topographic feature estimation processing that includes: classifying a plurality of measurement points, acquired by three-dimensional measurement of a scene and respectively including measurement information, into a plurality of point group sub-regions, each of which corresponds to a respective one of the plurality of classification vectors; and estimating topographic features of the scene by: for each of the plurality of point group sub-regions that have been classified for each of the measurement points included in the point group sub-region corresponding to a reference plane, taking a distance from the reference plane to each of the measurement points as a height of each of the measurement points, and by applying a progressive morphological filter to each of the plurality of point group sub-regions, removing a measurement point corresponding to a non-ground object from the plurality of measurement points acquired by the three-dimensional measurement.

Systems and methods of improving segmentation and classification in multi-channel images of biological specimens are described herein. Image channels may be arranged and processed sequentially, with the first image channel being processed from an unmodified image, and subsequent images processed from attenuated images that remove features of previous segmentations. For each segmented image channel in sequence, a binary image mask of the image channel may be created. A distance transform image may be computed from the binary mask. Attenuation images computed for all previous channels may be combined with a current attenuation image to create a combined attenuation image. The next image channel in the sequence may then be attenuated to produce an attenuated next image channel, which may then be segmented. The steps can be repeated until all image channels have been segmented.

A computer implemented method using artificial intelligence for matching images with locations and directions by acquiring a plurality of panoramic images, detecting objects and their locations in each of the panoramic images, acquiring a floorplan image, detecting objects and their locations in the floorplan image, comparing the objects and locations detect in each of the panoramic, image to the objects and locations detected in the floorplan image, and determining a location in the floorplan image where each panoramic image was taken.

A computer implemented method for determining a centerline of a three-dimensional tubular structure is described. The method includes providing an edge-detected data set of voxels that characterize a boundary of the tubular structure according to a three-dimensional voxel data set for the tubular structure. A gradient field of a distance transformation is computed for the edge-detected dataset. A voxel data set corresponding to a centerline of the tubular structure is computed according to derivative of gradient field. A trajectory within the tubular structure is computed based on the centerline.

Systems and methods for generating a centerline for an object in an image are provided. An exemplary method includes receiving an image containing the object. The method also includes generating a distance cost image using a trained first learning network based on the image. The method further includes detecting end points of the object using a trained second learning network based on the image. Moreover, the method includes extracting the centerline of the object based on the distance cost image and the end points of the object.

An image segmentation method and device, a computer device and a non-volatile storage medium are provided. The image segmentation method includes: performing super-pixel segmentation on an image to be segmented to obtain a super-pixel image, and binarizing the image to be segmented to obtain a binary image; combining the super-pixel image and the binary image to obtain a binary super-pixel image; performing distance transformation on the binary super-pixel image, to obtain a grayscale super-pixel image; marking seed points in the grayscale super-pixel image, to obtain a seed point super-pixel image in which grayscale values of the seed points are greater than a first value, and grayscale values of pixel blocks other than the seed points in target regions are the first value; and marking and filling the pixel blocks, grayscale values of which are the first value in the seed point super-pixel image, to obtain a segmented image.

A method of image-distance transformation using bi-directional scans is provided. The method includes the steps of: performing a first scan on each pixel of an input image using a first mask in a first order to generate an intermediate image; and performing a second scan on each pixel of the intermediate image using a second mask in a second order to obtain distance information of each pixel in the input image. A first current pixel in the input image that is not compared with prior pixels in the first order and in a first current segment is used in the first comparison process in the first scan, and a second current pixel that is compared with prior pixels in the second order and in a second segment is used in the second comparison process in the second scan.

Techniques are provided for determining a cell count within a whole slide pathology image. The image is segmented using a global threshold value to define a tissue area. A plurality of patches comprising the tissue area are selected. Stain intensity vectors are determined within the plurality of patches to generate a stain intensity image. The stain intensity image is iteratively segmented to generate a cell mask using a local threshold value that is and gradually reduced after each iteration. A chamfer distance transform is applied to the cell mask to generate a distance map. Cell seeds are determined on the distance map. Cell segments are determined using a watershed transformation, and a whole cell count is calculated for the plurality of patches based on the cell segments. A client device may be configured for real-time cell counting based on the whole cell count.

A data processing unit receives a reference image (IMG1.sub.3D) of a deformable physical entity, a target image (IMG2.sub.3D) of said physical entity, and a first region of interest (ROI1.sub.3D) defining a first volume in the reference image (IMG1.sub.3D) representing a reference image element. The reference image (IMG1.sub.3D), the target image (IMG2.sub.3D) and the first region of interest (ROI1.sub.3D) all contain 3D datasets. In response to user commands (c1; c2), the data processing unit defines a first contour (C1.sub.2D) in a first plane through the target image (IMG2.sub.3D), which is presented to a user via a display unit together with graphic data reflecting the reference image (IMG1.sub.3D), the target image (IMG2.sub.3D) and the first region of interest (ROI1.sub.3D). The first contour (C1.sub.2D) is aligned with at least a portion of a first border (IEB1) of a target image element (IE.sub.3D) in the target image (IMG2.sub.3D). The target image element (IE.sub.3D) corresponds to the reference image element in the reference image (IMG1.sub.3D). Based on the first contour (C1.sub.2D), the target image (IMG2.sub.3D) and the first region of interest (ROI1.sub.3D); the data processing unit determines a second region of interest (ROI2.sub.3D) defining a second volume in the target image (IMG2.sub.3D).