70 results about "Aspect ratio (image)" patented technology

Disclosed is a image processing apparatus and a method for controlling the image processing apparatus. The image processing apparatus according to an embodiment of the present disclosure may identify an object from an acquired image, determine whether the object is hidden by another object by using an aspect ratio of a bounding box of the detected object, and based on the object being hidden, estimate an entire length of the object based on coordinate information of the bounding box. Accordingly, the size information of the hidden object may be efficiently identified while a large amount of database is applied or resources of the apparatus is minimized. The present disclosure may be in connection with a surveillance camera, an automotive driving vehicle, an artificial intelligence module of at least one of a user terminal or a server, a robot, an augmented reality (AR) device, a virtual reality (VR) device, a device related to a 5G service, and the like.

Systems and methods for image processing (e.g., image extension or image uncropping) using neural networks are described. One or more aspects include obtaining an image (e.g., a source image, a user provided image, etc.) having an initial aspect ratio, and identifying a target aspect ratio (e.g., via user input) that is different from the initial aspect ratio. The image may be positioned in an image frame having the target aspect ratio, where the image frame includes an image region containing the image and one or more extended regions outside the boundaries of the image. An extended image may be generated (e.g., using a generative neural network), where the extended image includes the image in the image region as well as generated image portions in the extended regions and the one or more generated image portions comprise an extension of a scene element depicted in the image.

The invention relates to the technical field of computer image processing, and particularly discloses a multi-image intelligent compression splicing method and splicing system capable of configuring the column number, and the splicing method comprises the following steps: 1, achieving the intelligent compression of images through the iteration quality attenuation, and balancing the image quality and storage space; step 2, dynamically calculating an image layout size based on the number of columns configured by a user; 3, keeping the aspect ratio of the image by adopting a cell centering strategy; and 4, realizing multi-image splicing by a dynamic canvas synthesis technology. The method has the advantages that intelligent image compression is achieved, and mass and storage space are balanced; a compression strategy can be dynamically adjusted according to the number of images; maintaining the original aspect ratio of the image; flexible multi-column image arrangement is supported; and splicing parameters can be dynamically configured.

The invention provides a real-time open vocabulary target detection model AOV-RTDETR for an aerial image, and aims to solve the problems that a traditional detection model cannot recognize unseen categories, small targets are difficult to detect and the real-time performance is insufficient. The model is based on an RT-DETR framework, cross-modal feature alignment of images and texts is achieved in combination with a CLIP visual language model, region-text comparison loss is introduced to break through predefined category limitation, and detection of unseen categories is supported. In order to improve the detection capability of a small target and a high aspect ratio target, a large strip-shaped convolution structure is provided to enhance the feature modeling capability, and a multi-scale spatial pyramid attention encoder and a text-guided multi-stage fusion decoder are designed to improve the feature expression and semantic alignment capability of the small target. The method has high precision and real-time balance capability, and is suitable for application scenes with high requirements for target detection precision and speed, such as aerial remote sensing, emergency rescue, unmanned aerial vehicle patrol and the like.

The invention relates to a Badam sorting method and device based on an X-ray image, and the method comprises the steps: obtaining an original X-ray image, and carrying out the preprocessing of the original X-ray image; processing the preprocessed X-ray image to obtain an image to be sorted; a density defect fusion descriptor, a penetrating empty shell recognition descriptor, a structure and texture stability descriptor and a special-shaped impurity recognition descriptor are extracted from each to-be-sorted image, and the density defect fusion descriptor fuses a gray average value, a gray standard deviation and a low-gray connected region proportion; the penetrable empty shell recognition descriptor is fused with density gradient and packing ratio, the structure and texture stability descriptor is fused with contour area and roundness, and the special-shaped impurity recognition descriptor is fused with length-width ratio, pixel energy and local texture ambiguity; and obtaining a sorting result. Compared with the prior art, the method has the advantages of high classification precision and the like.

Methods of particle shape and size characterization of powder particles include imaging the particles at a pixel-scale resolution to acquire images of the particles in which perimeter irregularities appear, increasing the resolution of the images and then performing fine-graining analysis on the images to accentuate at least some of the perimeter irregularities, calculating aspect ratios and elliptical form factors of the particles from the images; and determining the perimeter irregularities of the particles and elongation of the particles by mapping particles on an aspect ratio versus elliptical form factor plane to decouple the perimeter irregularities of the particles from the elongation of the particles and yield a statistical description of shapes of the particles.

To determine a projection area of an image on a projection surface so as to effectively use a drawing area of a panel of a projector.SOLUTION: A determining method of a projection area PAR, includes steps of: specifying coordinates of four vertices P10s, P11s, P12s, and P13s, which correspond one-to-one to four vertices P10p, P11p, P12p, and P13p of a rectangle RE10p defining a drawing area DAR1 in a panel PL, in a screen coordinate system ΣS of a projection surface SC; identifying a first intersection point coordinate which is a coordinate in the screen coordinate system ΣS of a first intersection point PI1s at which two diagonal lines of a quadrangle QU10s having vertices P10s, P11s, P12s, and P13s as vertices intersect with each other; identifying a largest rectangle which falls within a range of the quadrangle QU10s and has a first aspect ratio centered on the first intersection point coordinate as a first rectangle RE1s; and determining a projection area PAR on the projection surface SC based on the first rectangle RE1s.SELECTED DRAWING: Figure 11

The present invention discloses an image defect detection method, comprising: S1: extracting gradient features of an image to be detected to obtain a gradient map; inputting the image to be detected into a convolutional neural network for training to obtain a feature map; S2: fusing the gradient map with the feature map to obtain a feature map of a fused gradient attention mechanism; S3: obtaining a number of anchor boxes with adaptive aspect ratios; S4: obtaining the deep relationship between each anchor box; S5: clustering and merging the anchor boxes according to the deep relationship between each anchor box to obtain a text output area; S6: comparing the text output area with a standard image to complete defect detection of the image to be detected. The present invention generates an anchor box with adaptive aspect ratios, and uses a graph neural network to obtain the deep relationship between each anchor box, so that the anchor boxes can be better clustered and merged, and the text information on the image to be detected can be quickly and accurately located, which facilitates the processing of defect detection.

The present invention relates to a method for measuring dislocation density, the method comprising: obtaining an image of an object to be measured; detecting etch pits in the obtained image according to preset etch pit screening conditions, and statistically calculating an average area value S0 and an average aspect ratio R0 of the long side to the short side of a single etch pit in the obtained image; performing contour detection on the black shapes in the obtained image to obtain a contour area S and an aspect ratio R of the long side to the short side of the bounding rectangle of the black shape; determining the number of etch pits of the black shape based on the relationship between the contour area S and the aspect ratio R of the long side to the short side of the black shape and the average area value S0 and the average aspect ratio R0 of the long side to the short side of the bounding rectangle; and calculating the dislocation density of the object to be measured based on the total number of etch pits in the obtained image and the total area of the obtained image. The present invention also relates to a dislocation density counting device and a non-transitory computer-readable medium.

The invention relates to a feature fusion method based on a special-shaped convolution kernel, and aims to solve the problem that traditional feature fusion is poor in adaptability to object direction and length-width ratio change. Comprising the following steps: extracting input image features; multiple special-shaped convolution kernels are adopted to extract multi-direction features in parallel; splicing the feature maps along the height dimension; and fusing multi-scale information through expansion convolution, and adding the fused multi-scale information with original feature residual errors. According to the feature fusion method, the detection precision of targets with changeable directions and large length-width ratio difference is remarkably improved, mAPs on SIXray and CLCXray data sets are improved by 2.22% and 1.84% respectively, and the feature fusion method is suitable for target detection of complex scenes such as security check and remote sensing.

The present invention relates to the field of artificial intelligence technology, and specifically discloses an image processing method, device, storage medium and computer program product. When segmenting an image to be processed according to the resolution of the image to be processed and the resolution of a visual encoder, a first optimization condition for reducing the deviation between the aspect ratio of a first sub-image and the aspect ratio of a second sub-image and a second optimization condition for constraining the area of ​​each second sub-image and the difference with the area of ​​the image to be processed are set, thereby ensuring that the aspect ratio of the second sub-image input to the visual encoder after segmentation is as small as possible compared with the aspect ratio of the image to be processed to avoid large deformation when adapting to the resolution of the visual encoder, and avoiding excessive calculation amount and loss of local integrity of the image due to too large a number of segmentations, thereby ensuring the processing quality of the image processing task of artificial intelligence from the perspective of quality and efficiency.

A method includes obtaining image frames that comprise a video, each of the image frames having a first aspect ratio, and selecting at least one of the image frames as a condition frame. The method further includes generating, from each condition frame based on an image outpainting model, an outpainted condition frame that has a second aspect ratio different from the first aspect ratio. The method further includes generating, from at least one of remaining image frames based on a video outpainting model and the outpainted condition frames, an outpainted target frame that has the second aspect ratio, each outpainted target frame having spatial consistency with the image frame from which it was generated and temporal consistency with neighboring outpainted frames in the video.

The invention discloses a longitude and latitude extraction method, device and equipment and a readable storage medium, and is applied to the field of image processing, and the method comprises the steps: carrying out the threshold segmentation of an original image containing a grid-type material, and obtaining a binary image; performing convolution operation on the binary image based on a convolution kernel straight line extraction method to obtain a convolution image; processing the convolution graph based on morphological operation and the contour aspect ratio to obtain a processed image; and performing longitude and latitude skeleton extraction and duplicate removal processing on the processed image to obtain pixel coordinates of each longitude and latitude skeleton. According to the invention, the line extraction method of the convolution kernel can rapidly extract the warp and weft contours; irrelevant features can be eliminated by using morphological operation and a contour aspect ratio; the image is subjected to longitude and latitude skeleton extraction and de-duplication processing, the longitude and latitude skeleton coordinates in the grid type material image can be accurately positioned, the cutting path can be effectively optimized in the later period, and the cutting quality and efficiency are improved.

Problem: high-precision detection of impersonation of a person using a planar object such as a photograph. Solution: imaging means 11 that images a plurality of frames of the verification target at different angles; feature detection means 14 that detects the aspect ratio of the face of the verification target, i.e., the ratio of the vertical length to the horizontal length, from each of the plurality of frames of images imaged by the imaging means 11; and impersonation detection means 14 that detects impersonation of a person using a planar object based on the aspect ratio of the face of the verification target imaged at different angles.

This disclosure provides a character recognition method, apparatus, electronic device, and storage medium. The character recognition method includes: acquiring an image of a target character to be tested, wherein the target image includes at least one character to be tested; extracting and determining the grayscale value and aspect ratio of each character to be tested based on the image of the target character to be tested; extracting and determining the directional gradient, geometric invariant moments, and gray-level co-occurrence matrix of each character to be tested based on the grayscale value; determining the geometric features of each character to be tested based on the aspect ratio; determining the edge features of each character to be tested based on the directional gradient and the geometric invariant moments; determining the texture features of each character to be tested based on the gray-level co-occurrence matrix; inputting the geometric features, the edge features, and the texture features into a character recognition model; and outputting the recognition result of each character to be tested through the character recognition model.

The present invention relates to a body fluid test method, comprising: a step for identifying the ROI region of each sperm in each of n set image frames; reducing the image brightness of each of the n image frames by a set value; a step of creating an accumulated image frame in which the images of each sperm in the n image frames are accumulated; a step of confirming a movement path region of each sperm in the accumulated image frame; for each sperm, calculating the width of the moving path area, and judging the activity according to the calculated width; and calculating the aspect ratio of a rectangle surrounding the movement path region for each sperm, and determining the straightness according to the calculated aspect ratio.

The display device provided by the present application, the controller is configured to: receive the screen-cast media resource, and determine the media resource attribute; if the media resource attribute is a video, detect the portrait / landscape state of the display, obtain the video aspect ratio, and determine whether the video aspect ratio matches the portrait / landscape state of the display; if the media resource attribute is a picture, fill the screen for display according to the original ratio of the current image aspect ratio. The display device provided by the present application can process images and videos separately according to the attributes of the received screen-cast content. For images, they are directly displayed, while for videos, it is determined whether rotation is required based on the video aspect ratio and the portrait / landscape state of the display, so that the video playback better adapts to the ratio of the display, improving the user experience.

The disclosed approach performs image compression using a variational auto-encoder that is capable of greater compression than conventional approaches while maintaining high fidelity of the decompressed image. Examples persist a bottleneck layer output of a variational auto-encoder into a compressed image in the form of potential tensors. The potential tensor is decompressed into a restored image in the pixel space by a variational self-decoder. In some examples, different encoder / decoder pairs are trained on a particular image type based on feature attributes. For example, maps have lines that are narrower compared to their length (e.g., with high aspect ratios), which differ from features within photographs of persons and scenes. Some examples utilize contrast language image pre-training (CLIP) and / or bootstrap language image pre-training (BLIP) models to store inserts, each of which is associated with a compressed image to enable a natural language search of a set of compressed images without requiring decompression.

The invention discloses a network-based rail B-display image recognition method, which belongs to the field of railway track flaw detection and deep learning, and comprises the following steps: 1, acquiring a rail B-display image, dividing the B-display image into five types of labels, namely a rail end label, a rail head nuclear flaw label, a normal screw hole label, an abnormal screw hole label and a lead hole label, and constructing a data set after performing data enhancement on small sample types; 2, a data set is put into the model for training, an attention module is embedded into the model, the sensing ability of key features is enhanced through channel re-calibration, and the detection precision of small target damage is improved; and 3, aiming at a thin and long target with an extreme aspect ratio, an EIoU loss function is adopted to optimize bounding box regression, and the positioning precision and the convergence speed of the model in a complex form are remarkably improved. According to the invention, fine defects such as rail head nuclear injury, abnormal screw holes and the like can be effectively identified, the omission ratio and the false detection ratio are reduced, intelligent identification and automatic analysis of B-display images are realized, the manual dependence is greatly reduced, and the detection efficiency and the railway operation safety are improved.

Aspects of the technology described herein identify geologic features within seismic data using modern computer analysis. An initial step is the development of training data for the machine classifier. The training data comprises an image of seismic data paired with a label identifying points of interest that the classifier should identify within raw data. Once the training data is generated, a classifier can be trained to identify areas of interest in unlabeled seismic images. The classifier can take the form of a deep neural network, such as a U-net. Aspects of the technology described herein utilize a deep neural network architecture that is optimized to detect broad and flat features in seismic images that may go undetected by typical neural networks in use. The architecture can include a group of layers that perform aspect ratio compression and simultaneous comparison of images across multiple aspect ratio scales.

The present application relates to the field of image processing technology, and specifically to an olive leaf profile analysis method and system based on image processing. First, a standard model for olive leaf profile analysis is constructed, and then a comparison grayscale image of the leaf to be analyzed is obtained, and a leaf profile diagram of the leaf to be analyzed is separated from the image; then, the actual aspect ratio, actual symmetry, and actual circularity of the leaf to be analyzed are calculated from the leaf profile diagram, and a weighted judgment index is generated based on the actual aspect ratio, actual symmetry, and actual circularity. Finally, a comparison analysis result of the leaf to be analyzed is output based on the weighted judgment index and the standard model. The present application realizes automatic sampling and automatic recognition of olive leaves, which effectively reduces the workload of staff and can effectively improve analysis efficiency. Compared with traditional manual analysis, the present application adopts a unified standard for analysis, which can effectively avoid the influence of the subjective feelings of technicians on the analysis and improve the objectivity of the analysis.

The invention relates to the technical field of image enhancement, in particular to a blade scratch recognition method based on image enhancement, which comprises the following steps of: defining a texture sliding window with the size of N * N according to a to-be-detected blade grayscale image, and controlling the texture sliding window to carry out pixel-by-pixel traversal on a pixel coordinate system of the to-be-detected blade grayscale image, and calculating and generating a local entropy value. According to the method, the fused matrix is processed through morphological closed operation, the discontinuous high-response area is connected through the structural elements, internal holes are filled, the scratch fracture phenomenon caused by uneven illumination or metal reflection is repaired, and the completeness and continuity of the defect form are ensured. And performing connected domain marking based on a binary mask, extracting geometric parameters of an external rectangle, strictly screening the connected domains according to a length-width ratio form standard, and removing non-strip interference noise such as spots and pits, so that high-precision positioning of fine scratches on the surface of the blade is realized.

This invention relates to a method and apparatus for sorting almonds based on X-ray images. The method includes: acquiring original X-ray images and preprocessing them; processing the preprocessed X-ray images to obtain images to be sorted; extracting density defect fusion descriptors, penetrating shell identification descriptors, structure and texture stability descriptors, and irregular impurity identification descriptors from each image to be sorted. The density defect fusion descriptor incorporates the mean gray level, gray level standard deviation, and proportion of low-gray-level connected regions; the penetrating shell identification descriptor incorporates density gradient and fill ratio; the structure and texture stability descriptor incorporates contour area and roundness; and the irregular impurity identification descriptor incorporates aspect ratio, pixel energy, and local texture ambiguity. The sorting result is then obtained. Compared with existing technologies, this invention has the advantages of high classification accuracy.

This invention relates to a method and apparatus for object detection and determining training samples for object detection. The method, when dividing positive and negative samples, determines positive samples of target boxes based on the aspect ratio of the target boxes in the training image. Furthermore, when the aspect ratio of a target box is large, a first preset threshold inversely proportional to the aspect ratio is set. This results in a smaller first preset threshold when the aspect ratio of the target box is large. Using the intersection-union ratio (IUU) of candidate boxes and target boxes, along with the first preset threshold, can also effectively determine whether a candidate box is a positive sample, avoiding missed detections. This lays the foundation for training an object detection model based on positive and negative samples of target boxes in the training image. The trained object detection model can better learn the position and feature information of the object to be detected, reducing the missed detection rate of objects with large aspect ratios, thereby improving the accuracy of object detection.

A system and a method for 3D surface imaging of a 3D object, the method comprising directing a light beam to a digital micromirror device that displays binary fringe patterns, converting the binary fringe patterns to grayscale fringe patterns at an intermediate image plane; rotating the grayscale fringe patterns onto the intermediate image plane to match an aspect ratio of the object, projecting the greyscale fringe patterns from the intermediate image plane to the object, capturing deformed structure images reflected by the object; and transferring the captured images to a computer connected to a frame grabber.

The invention discloses a multi-plane millimeter wave image target frame fusion method and system. The method comprises the following steps: firstly, grouping mapping results of a plurality of millimeter wave images on the same surface, and then calculating fusion scores of all target frames in groups; and fusing the target frame of each group mapping result based on the fusion score. And finally obtaining the current-round fusion result of each group, replacing the'mapping result 'for grouping, and recalculating the fusion score to carry out the next-round fusion operation. And repeating the steps until a final unique fusion result is obtained, and mapping the final unique fusion result on the standard doll graph for display. According to the method, three indexes of intersection-to-union ratio score, width-to-height ratio score and area ratio score in the vertical direction are designed, and the three scores are comprehensively considered to perform target frame fusion. By applying the method, the accuracy of the target mapping result can be effectively improved, and then the visual presentation effect is remarkably improved.

The application discloses an aragonite whisker preparation method and system based on image recognition and multi-model prediction, and comprises the following steps: in each sampling period, a microscopic image of reaction slurry is subjected to mask extraction, skeleton thinning, cross point breaking and ellipse fitting processing, an average observation vector of crystal size is decoupled and output, a 2D-PBM is input with estimated liquid phase ion concentration, a two-dimensional growth kinetics model is combined, and a posterior state vector at the current moment is obtained; an ESN proxy model is used to replace the 2D-PBM, and the crystal morphology evolution trajectory in the future time domain is deduced within the current sampling period; air inlet component fluctuation is fed forwardly compensated, and adjustment instructions for temperature and air inlet flow are output and executed. The application realizes direct and real-time closed-loop control on the aspect ratio of acicular crystals, suppresses morphology degradation caused by impurity fluctuation, and improves product quality and aragonite phase purity.

The embodiment of the invention relates to the technical field of image processing, in particular to a sample balancing method suitable for small target detection, and the method comprises the steps: obtaining a plurality of anchor frames and truth value bounding boxes for a target object in a feature map of a target image, the scale parameters and aspect ratio parameters of different anchor frames being different, and the truth value bounding boxes being different; pixel points occupied by the target object in the target image are smaller than a preset value; based on the position coordinates of the truth value bounding box and the position coordinates of the anchor frames, respectively calculating detection quantities between the truth value bounding box and the anchor frames; wherein the detection quantity is used for measuring the similarity degree between the anchor frame and the true value bounding box; dividing the plurality of anchor frames into a positive training sample, a negative training sample and an invalid sample based on the detection quantity between the true value bounding box and the plurality of anchor frames; and outputting a positive training sample and a negative training sample, wherein the positive training sample and the negative training sample are used for subsequent classification operation and regression operation of the target object. According to the method, the small target detection capability can be greatly improved.

The present invention discloses a method for detecting high-density foreground targets in optical remote sensing images based on semantic fusion and frame alignment. The execution steps of the method are: (1) preprocessing data; (2) using a backbone network to extract features; (3) inputting the extracted features into a prediction network to generate prediction results; (4) determining whether it is a training mode; (5) in the training mode, the model will calculate the training loss and optimize the network parameters; (6) in the non-training mode, the prediction results are output. The method includes two feature re-identification convolution modules that can effectively distinguish internal texture features, thereby improving the accuracy of small targets. At the same time, an auxiliary point balance IoU (ABIoU) loss is also proposed. The ABIoU loss can not only avoid the local optimal solution of the distance term in the single-point regression, but also avoid the non-convergence of the IoU loss of the local aspect ratio, thereby achieving the stability of the loss value and the direct measurement of the side length, thereby reducing the problem of target misdetection and missed detection in a complex texture feature environment.

A method includes obtaining image frames that comprise a video, each of the image frames having a first aspect ratio, and selecting at least one of the image frames as a condition frame. The method further includes generating, from each condition frame based on an image outpainting model, an outpainted condition frame that has a second aspect ratio different from the first aspect ratio. The method further includes generating, from at least one of remaining image frames, based on a video outpainting model and the outpainted condition frames, an outpainted target frame that has the second aspect ratio, wherein the outpainted target frame having spatial consistency with the image frame from which the outpainted target frame was generated and temporal consistency with neighboring outpainted frames in the video.