53 results about "Interest point detection" patented technology

Methods and apparatus for operating on images are described, in particular methods and apparatus for interest point detection and/or description working under different scales and with different rotations, e.g. for scale-invariant and rotation-invariant interest point detection and/or description. The present invention can provide improved or alternative apparatus and methods for matching interest points either in the same image or in a different image. The present invention can provide alternative or improved software for implementing any of the methods of the invention. The present invention can provide alternative or improved data structures created by multiple filtering operations to generate a plurality of filtered images as well as data structures for storing the filtered images themselves, e.g. as stored in memory or transmitted through a network. The present invention can provide alternative or improved data structures including descriptors of interest points in images, e.g. as stored in memory or transmitted through a network as well as datastructures associating such descriptors with an original copy of the image or an image derived therefrom, e.g. a thumbnail image.

Systems, methods and devices for determination of disease class scores for patient tissue are disclosed. The disease class scores may be based on the probability or probability-like metric for a disease condition or outcome. The system includes an imaging apparatus and a computing system with instructions executable by a processor. The computer system may locate one or more points-of-interest on the pre-processed images of the patient tissue using a point-of-interest detector and generate one or more disease spatial maps including one or more probability or probability-like metric of disease classifications on the detected points-of-interest. The information in the disease spatial maps is aggregated to produce disease class scores.

Systems, devices, and methods for training a neural network and performing image interest point detection and description using the neural network. The neural network may include an interest point detector subnetwork and a descriptor subnetwork. An optical device may include at least one camera for capturing a first image and a second image. A first set of interest points and a first descriptor may be calculated using the neural network based on the first image, and a second set of interest points and a second descriptor may be calculated using the neural network based on the second image. A homography between the first image and the second image may be determined based on the first and second sets of interest points and the first and second descriptors. The optical device may adjust virtual image light being projected onto an eyepiece based on the homography.

The invention relates to video processing, semantics extraction and the like. In order to provide a method for modeling abnormal events in multi-visual angle video monitoring based on temporal-spatial correlation information, the technical scheme is that: the method for modeling the abnormal events in the multi-visual angle video monitoring based on the temporal-spatial correlation information comprises the following steps of: (1) extracting temporal-spatial local characteristics, namely (a) detecting temporal-spatial characteristic interest points and (b) describing the temporal-spatial characteristic interest points; (2) constructing a word bag, namely (a) constructing a codebook and (b) projecting key points based on the codebook, wherein the work bag is a method for characterizing video units in the form of a statistical histogram by using extracted characteristic points; and (3) modeling the abnormal events, particularly for (a) videos acquired by monocular cameras and (b) the probability of the abnormal events which are output by an abnormal event detecting module under multi-visual angle monitoring videos. The method is mainly applied to the video processing.

The invention discloses a target tracking method and a system thereof, wherein, the method comprises the following steps: a video image of a tracking target is acquired; before a target model of the tracking target is established, the video image is analyzed to acquire an area corresponding to the complete target of the tracking target in the video image, and a target model of the tracking target is established according to the area corresponding to the acquired complete target; after the target model of the tracking target is established, the video image is analyzed to acquire possible locations in the target area of the tracking target in the current image, and the possible locations in the target area are taken as candidate target areas; the features of each candidate target area are respectively matched with the target mode based on detection of interest points, the partial image feature extraction technology and the pyramid matching algorithm, and the candidate target area with the largest matching result is taken as the current target area of the tracking target. The technical proposal provided by the invention can improve the success rate of tracking.

A hardware coprocessor architecture calculates the Difference-of-Gaussian (DoG) pyramid of an input image and extracts from this the interest points to be used in several image detection algorithms. Advantages of the architecture include the possibility to process the image by stripes, namely by blocks having one dimension coincident with the input image width, in the absence of an input frame buffer and the possibility to avoid RAM memory. The coprocessor is suitable to be tightly coupled with raw image sources like sensors.

The invention relates to a head posture estimation interest point detection method fusing depth and gray scale image characteristic points. The method comprises the following steps: extracting the characteristic points of a depth image; extracting the characteristic points of a gray scale image; fusing the characteristic points of the depth image and the gray scale image. The characteristic point detected on the basis of the depth image and the characteristic point detected on the basis of the gray scale image are combined to form certain characteristic points which are positioned accurately and are high in robustness, thereby inheriting the advantage of the detection of different characteristic points of the depth image and the gray scale image, and realizing maximum and rapid detection of characteristic points with great surface variations in the depth image and a pupil area with a great gray scale value in the gray scale image. In particular, a calculation mode for correcting a calculated Haar-like characteristic value in the depth image is provided, the finally-extracted characteristics have certain spatial rotation invariability, and the true values of human face characteristic points can be approached under the situation of large-angle rotation, thereby increasing the final characteristic point detection accuracy, and shortening the detection time.

The invention discloses a fast correlation neighborhood feature point-based sliding window target tracking method. The method includes the following steps of: S1, target window template generation; S2, fast correlation neighborhood feature point extraction; S3, optimal point of interest screening; S4, point of interest sliding window search; S5, feature point template matching update; and 6, decision voting output. The invention also discloses a fast correlation neighborhood feature point-based sliding window target tracking system. With the method and system of the invention adopted, the problem of poor real-time performance and low stability of target tracking in a complex condition can be solved. According to the method and system, fast correlation neighborhood feature points, adopted as point of interests, are detected, and therefore, the robustness of target feature description in a complex condition can be enhanced; point of interest screening is carried out through using a window cross-correlation relation, and therefore, the accuracy of the target description can be improved; and sliding window search and adaptive multi-scale template matching online update are adopted in template construction, and target window output is realized through adopting decision voting, and therefore, the accuracy and stability of target tracking in the complex condition can be improved.

The invention belongs to the technical field of image processing and particularly relates to an infrared and visible image fusion method based on a saliency map and interest point convex hulls. The infrared and visible image fusion method comprises the following steps that firstly, saliency detection is conducted on an infrared image, so that a binarized saliency map is obtained; secondly, interest point detection is conducted on the infrared image; thirdly, free interest points are removed, and salient interest points are obtained; fourthly, convex hulls of the salient interest points are obtained, AND operation is conducted on the binarized saliency map and a convex hull image, so that a target area is obtained, and the remaining portion of the image is used as a background area; finally, different fusion rules are adopted for the target area and the background area respectively, and the infrared image and the visible image are fused. By the adoption of the infrared and visible image fusion method, a few of interest points on a background can be removed, the obtained convex hulls of the salient interest points are more approximate to a real target, and by means of the saliency map, the target area can be more accurately extracted; the fusion performance is easily, rapidly and effectively improved.

The invention discloses a video image matching method based on mesh statistical constraint. The overall steps are as follows: FAST feature point detection is carried out on the input reference image Ia and the image Ib to be matched; After the feature points are obtained, the descriptor of a feature point is calculated by BRIEF algorithm. M point pairs are selected around the key pixel points P, and the comparison results of m point pairs are combined as descriptors; Searching for the nearest neighbor of each feature point on the image Ib to be matched on the reference image Ia; the two imagesare mesh-processed respectively, and all the mesh are matched. The real-time, super-robust matching method based on mesh statistics of the present invention takes motion smoothness as statistical likelihood of matching in a certain number of regions, and can quickly and reliably distinguish true from false matching, thereby obtaining high-quality matching point pairs.

The invention provides an interest point detection method and system of three-dimensional (3D) point cloud. The system comprises a 3D point cloud data acquisition module, a shape description module and an interest point extraction module, wherein the 3D point cloud data acquisition module is used for obtaining 3D point cloud data; the shape description module is used for generating a shape description vector of a surface wherein a pixel and a neighbourhood point are positioned by the distribution information of directed distance between the neighbourhood point of the pixel and a tangent plane where the pixel is positioned in the input 3D point cloud; and the interest point extraction module extracts the interest point on the basis of the generated shape description vector.

The invention discloses a structural similarity-based rapid and robust visible light image and SAR image registration method. The method comprises the following steps of: firstly carrying out coarse registration on a to-be-registered visible light image and a to-be-registered SAR image by using a rational polynomial coefficient model; carrying out point of interest detection on the visible light image by using an improved corner detection method, and uniformly selecting a certain quantity of points of interest as control points; constructing similarity measurement by using an improved directional Harris corner histogram HIOHC so as to carry out key point matching in the SAR image; after getting rid of obvious mistaken matching, solving a global conversion matrix by using a least square method LSM and obtaining a final registration result. The method has the effects of breaking through dependency, on image descriptors, of traditional methods, breaking through sensitivity, to different data sources, of the traditional methods, obtaining relatively ideal effects, realizing high-precision registration of visible light images and SAR images, and having important practical significance for disaster monitoring, a change detection and loss estimation.

The present invention relates to the technical field of artificial intelligence and more particularly to a commodity alignment method based on unsupervised feature point detection. The commodity alignment method based on unsupervised feature point detection comprises the following steps: S1, preparing feature point detection training data; S2, training the detection frame model; 3, feature point detection; 4, carry out affine transformation alignment according to that coordinate of the characteristic points. The invention can be used on the commodity subsequent identification network after thefeature points are detected and aligned, and the identification accuracy rate is obviously higher than that of the commodity subsequent identification network without alignment, because the network is easier to identify the forward object than the inclined object. For the existing supervised feature point alignment, this method can save labeling cost.

The invention discloses a multiple athletes behavior detection method based on scale adaptation local spatiotemporal features, and the method comprises the following steps: S1, a normalized Laplace operand is adopted to estimate a local scale and establish a scale adaptation local spatiotemporal feature detection algorithm, and a Harris spatiotemporal interest point detection operand and the Laplace operand are combined to infer a Harris-Laplace spatiotemporal interest point detection operand; S2, the normalized Laplace operand is generalized to a spatiotemporal domain; S3, local spatiotemporal feature descriptors are generalized to three-dimensional soccer match video images; then before a K-means clustering algorithm is adopted to generate a spatiotemporal codebook, each spatiotemporal interest point is normalized to ensure that the each spatiotemporal interest point is invariable under scaling and shifting.

The invention relates to a similarity matching method of cutting tool pictures, and belongs to the technical field of information. First of all, a gray-scale map is generated; then an image including a tool portion is obtained by removing angle squares at the two sides of a tool image; then performing gradient field operation of a GVF smooth image, eliminating details with quite small noise and scales, and obtaining an image with clear tool contour; then carrying out feature point detection; performing main direction distribution on detected feature points; according to main directions of the feature points, generating feature descriptors; and obtaining the size of similarity by performing feature point matching on feature vectors in the feature point descriptors by use of an Euclidean distance formula. The method can effectively solve the problem of incapability of classifying cutting tools due to concentration of the feature points in local small details (features with quite small scales) of an object by use of a standard SIFT algorithm, completes a comparison algorithm of image similarities through GVF-Harris-SIFT, and finally realizes type classification of the cutting tools of an engineering project.

The invention provides a user environment classification method based on low sampling rate sensing information; the method uses the following units: a data acquisition module, a data processing module, a position of interest (POI) detection module and a POI classification module; the method comprises the following steps: using a low sampling rate to collect intelligent mobile phone data and updating the data to a database, carrying out denoising, time synchronization and data alignment for the data, and using a verification algorithm based on stay points to extract the POIs in the data; using a cluster algorithm with noises and based on density to cluster similar POIs; carrying out POI analysis and classification based on information fusion environment classification framework. The method can effectively collect real world information, and can use the low sampling rate model effectiveness to express high noise time domain reality data according to the existing reference cluster method, thus solving the limited battery electric quantity disadvantages in intelligent mobile phone high sampling rate data acquisition, and improving the POI environment type distinguishing accuracy.

An intended interest point detection method provided for detecting an intended POI of a person from a plurality of first images and at least one second image includes obtaining a plurality of gaze information of the person in the plurality of first images; obtaining a location information of at least one POI in the at least one second image; determining energy value by the plurality of gaze information and the at least one POI; and inferring the intended POI by minimizing the energy value. The energy value is decreased when the gaze point location is correlated to a POI, and/or when the intended POI is consistent in contiguous time frames.

The embodiment of the invention discloses a method, a device and equipment for determining the position of an actual mining interest point and a readable storage medium, which relate to the artificialintelligence technology, in particular to the technical field of computer vision and electronic maps. According to the specific implementation scheme, interest point detection is carried out on an actual mining image to obtain a target interest point and a target direction of the target interest point relative to an image shooting place; and the geographic position of the target interest point isdetermined based on a ground area passing along the target direction from the image shooting place. According to the embodiment, the accuracy and the intelligent degree of determining the actual mining interest point position can be improved.