46 results about "Contour segmentation" patented technology

The invention relates to the technical field of industrial visual inspection, and discloses a leather surface defect identification method and system based on image processing. The method comprises the steps of performing preprocessing and multi-scale feature extraction on a leather image, generating a defect thermodynamic diagram through an attention mechanism, and dividing candidate regions; innovatively constructing a regional association graph, updating node embedding by using a graph attention network, and fusing global context information; and generating instance segmentation results through adaptive clustering and classifying the instance segmentation results. According to the method, region relevance is modeled through a graph structure, so that recognition of each region is benefited from global context information, and the problem that segmentation of adhesion and irregular defects is inaccurate in a traditional method is effectively solved; a clustering type instance generation mechanism can automatically aggregate discrete regions according to deep feature similarity, so that the defect contour segmentation precision and instance distinguishing capability under a complex texture background are improved, and accurate conversion from pixel-level prediction to instance-level segmentation is realized.

The invention relates to the technical field of computer vision and image processing, in particular to an end-to-end image registration method based on a monocular camera and micro rendering. The objective of the invention is to solve the problems of low registration precision and insufficient real-time performance caused by complex calculation, serious error propagation and poor adaptability to multi-modal data in a traditional image registration method. According to the main scheme, the method comprises the steps of extracting structured boundary information of a target object in an input image through a contour segmentation model, and generating a binary mask image; based on the contour segmentation result, estimating a 6D pose parameter of the target object through a 6D pose estimation deep network fusing a convolutional neural network CNN and a Transform module; and inputting the 6D pose parameters into a micro-renderable module, generating a prediction image, comparing the prediction image with an original input image, and performing end-to-end optimization and adjustment on model parameters to realize high-precision registration.

The application discloses a kind of based on the improved U-net's ultrasonic image's disassembly method, it is related to ultrasonic image technical field, by combining depth learning technique and image post-processing algorithm, can efficiently handle ultrasonic image type (single figure, double jigsaw, four jigsaw) The identification of operation, ultrasonic regional map contour segmentation and jigsaw disassembly, through the depth learning framework of U-net network, in combination with the feature extraction capability of ResNet50, efficient classification and segmentation of jigsaw image are realized, compared with traditional image processing method, the present application can process large-scale different organ tissue ultrasonic image in short time;Through accurate classification and segmentation, the contour of ultrasonic region in ultrasonic image can be accurately extracted, and jigsaw is effectively disassembled, the integrity of ultrasonic image is restored, avoids the case of misremoval or loss of information, and has wide clinical application prospect.

The invention discloses a corn seed variety identification and contour segmentation method and device based on deep learning, and relates to the technical field of corn seed category identification. The problems that in the prior art, a trained corn variety recognition model is difficult to achieve integrated detection of variety recognition and seed segmentation, tiny differences between similar varieties and morphological changes between seeds of the same variety in a natural placement state are difficult to effectively process, the model parameter quantity is large, and the calculation cost is high are solved. The method comprises the steps that a training data set is constructed, each corn seed image comprises a plurality of corn seeds, an improved YOLOv11 model is constructed, on the basis of the YOLOv11 model, a backbone network is replaced with a ConvNeXt V2 network to optimize feature extraction efficiency and network weight reduction, an image-level variety classification head is newly added, a loss function adopts a double-task collaborative loss function, and the number of the corn seed images is smaller than the number of the corn seed images. Parallel output of image-level variety classification and seed contour segmentation is realized, and an improved YOLOv11 model is trained and applied.

The application relates to a method and system for automatically identifying defects on a radiographic inspection film, belonging to the technical field of nondestructive testing. The method comprises collecting the film to obtain a digital image; based on the structural features of the weld, the image is preprocessed by partitioning, the weld area, heat-affected zone and base material background are identified and segmented, and different image enhancement strategies are adopted for different areas; the preprocessed image is input into an improved deep learning model combining the YOLOv8 target detection network and the U-Net segmentation network for detection, and the positioning, classification and contour segmentation information of the defects are output; based on the pre-set confidence threshold, the identification results are filtered and distinguished as clear defects and suspicious defects; a film evaluation report is generated, and the digital fingerprint of the film image is calculated, the fingerprint and the report are encrypted and uploaded to the blockchain storage for subsequent authenticity verification. The application realizes full automation and intelligentization of the film evaluation process, improves the identification accuracy and efficiency, and ensures the credibility and non-tamperability of the detection results.

The present application relates to a kind of intelligent detection method and detection system of surface coating micron-sized pore, scheme as follows: obtain coating surface image to be detected and execute standardization preprocessing;Preprocessing image is input into trained hole profile segmentation model, and hole profile coordinates are obtained;Based on hole profile coordinates, construct hole connected region by morphological operation, and according to geometric feature, Gap region that does not need to be punched is screened;Preprocessing image is input into trained missed hole detection model, and real missed hole region is obtained;According to hole profile coordinates, calculate geometric feature parameter, determine collapse hole and calculate hole distribution density, finally output detection quality result.The present application realizes the high-precision, automation detection of micron-sized pore by the combination of deep learning model and morphological operation, geometric parameter calculation, solves the inherent defects of artificial detection, meets the high consistency, large batch fast control demand of modern manufacturing industry to surface coating quality.

The application provides a digital human display method and system based on intelligent VR space, and relates to the technical field of three-dimensional models.The method comprises the following steps: performing multi-dimensional feature extraction on an original character video to obtain a dynamic feature dataset; performing joint node cloud construction and registration based on the dynamic feature dataset to obtain a space-time consistency skeleton model; generating a surface mask for video frames of the original character video through the space-time consistency skeleton model to obtain a dynamic human body grid separated from a background; and performing physical light fusion on the dynamic human body grid and VR environment parameters to obtain a digital human VR entity model.The method combines dynamic feature dataset construction, joint node cloud registration, parameterized human body grid generation and physical light fusion, realizes high-fidelity digital human migration from an ordinary video to a VR space, solves key problems such as motion distortion and rendering mistakes, reduces motion trajectory errors and improves human body contour segmentation accuracy.

The invention provides a crab instance segmentation system and method based on multi-branch feature fusion. The method comprises the following steps: acquiring a high-resolution crab image in a unified illumination and fixed environment by using image acquisition; performing fine polygon segmentation marking on crab shells and step feet of the crab images, converting the crab shells and the step feet into a COCO format, and constructing a high-quality data set; a multi-branch feature fusion model is constructed in model training, shallow details, a local structure and high-level semantic information are respectively extracted by introducing a three-branch feature fusion path, and feature enhancement is performed on a weak texture region and a fuzzy boundary in combination with a feature enhancement module, so that the target boundary perception capability and the small target segmentation precision are improved, and the target segmentation accuracy is improved. Training to obtain a crab identification model; and performing instance segmentation on an acquired image by using the crab identification model, and outputting a target contour, a segmentation mask and related feature information. Therefore, high efficiency and cost advantages are achieved while high-precision instance segmentation is guaranteed, and the method has good practicability and popularization value.

The application discloses a curve segmentation fitting method and device for image segmentation and a storage medium, and comprises the following steps: pre-processing an image to obtain contour data; performing a contour segmentation algorithm of a DCE evaluation standard according to the contour data to obtain a segmentation point list; and performing a curve fitting algorithm based on an Euler curve model according to the segmentation list to obtain a final fitting result. The automatic segmentation of the contour can be realized, and subjective errors caused by human intervention are reduced.

The invention discloses a three-dimensional model reconstruction method based on a sheet metal bent part expansion drawing, and the method is characterized in that the method comprises the following steps: S1, analyzing a sheet metal expansion drawing, classifying the analyzed primitive entities, and storing the primitive entities in corresponding primitive sets; s2, primitive preprocessing: performing de-weighting, breaking and discrete processing on the primitive, and screening out a metal plate bending line; s3, closed contour searching; s4, judging whether a closed contour exists or not, if not, repeating S3, and if yes, entering S5; s5, contour segmentation and topology construction: performing inner row segmentation on the outer contour and the inner contour by using a bending line, and establishing a geometric adjacency relation between the inner contour and the outer contour; s6, bending deduction is conducted, specifically, a deduction value is calculated according to the workpiece and bending information, and contour deduction is conducted on the contour; and S7, performing three-dimensional reconstruction, and constructing a three-dimensional entity model. According to the method, the technical defects that three-dimensional model reconstruction cannot be correctly carried out on a complex sheet metal expansion drawing and the model size error is large are overcome.

The embodiment of the invention provides a robot environment object recognition method and system, and relates to the technical field of robot visual recognition, and the method comprises the steps: obtaining a background diffuse reflection light spot, and collecting a multispectral image data stream which comprises a visible light image and a near-infrared image; analyzing the multispectral image data stream, and quantifying the response difference of the background diffuse reflection light spot and the coffee cup under different spectral bands to obtain a target spectral response difference; generating a de-interference image based on the target spectral response difference; performing contour segmentation on the interference-removed image to obtain the contour of the coffee cup; and calculating the spatial position and posture of the coffee cup based on the contour. The success rate of grabbing operation of the robot and the overall working efficiency can be improved.

This disclosure relates to a method, apparatus, and storage medium for cloaked target detection in millimeter-wave images based on deep learning. The method includes: preprocessing the acquired image to statistically analyze the target distribution patterns; segmenting the target contours according to the target distribution patterns and fusing them with a background sample set; detecting specific target objects based on a target detection network; and detecting cloaked targets in millimeter-wave images based on training an improved target detection model. Through the various embodiments of this disclosure, the generalization ability of the model and its ability to detect targets at different scales are effectively improved, meeting the needs of practical applications.

The application discloses a multi-target recognition and contour segmentation method based on an infrared thermal image temperature gradient, and particularly relates to the technical field of infrared image recognition, and is used for solving the problem that in the existing multi-target recognition of an infrared thermal image, adjacent target thermal zones are connected due to continuous temperature transition and unclear boundary gradient, and thus the target contour is connected, and it is difficult to accurately distinguish multiple targets and the contour attribution of the targets. Through local standardization processing on the infrared thermal image of a to-be-recognized scene, a gradient stable boundary band is formed, and when a same candidate thermal connected region contains multiple target temperature peak seed regions, a temperature descending section, a gradient direction reverse section and a gradient amplitude contraction section are extracted to determine a gradient saddle channel, and then a target separation boundary is generated, so that when temperature transition connection occurs in the adjacent target thermal zones, the separation position can be accurately determined, the target merging recognition risk is reduced, and the multi-target recognition and contour segmentation accuracy of the infrared thermal image is improved.

The application discloses a three-dimensional body construction method and device based on CT images, a storage medium and a terminal. A CT sequence image of a target object is acquired. In response to a user's manual contour segmentation operation on a first target CT image in the CT sequence image, the first target CT image is subjected to first contour segmentation, and a first contour image of the target object is obtained. Based on each first contour image, a second target CT image in the CT sequence image, except for the first target CT image, is subjected to second contour segmentation, and a second contour image of the target object is obtained. A three-dimensional body of the target object is constructed according to each first contour image and each second contour image. Since the second contour segmentation is active processing and calculation on the manually segmented first contour image, when the second contour image is obtained, manual operation on a large number of images can be avoided, human-computer interaction is reduced, the three-dimensional body construction efficiency is improved, and the accuracy of the three-dimensional body of the target object can be ensured by the manually segmented first contour image.

The invention discloses a flowering branch thorn analysis method based on machine vision, and belongs to the technical field of image processing. According to the method, firstly, an image of a flowering branch with thorns is acquired through image acquisition equipment, and position detection of the flowering branch and the thorns in the image is realized by adopting a target detection algorithm of adaptive feature fusion and space consistency constraint. Then, single thorns most suitable for analysis are screened, thorn region information is obtained, region contour segmentation is carried out on the thorn region information, and color and morphological characteristics are extracted; and based on the extracted information, a support vector machine (SVM) classifier is utilized to automatically discriminate the stabbing form, and stabbing type identification is realized. According to the method, the overall structure and detail feature extraction can be taken into consideration, the detection precision is improved through multi-scale fusion and a spatial constraint mechanism, and the automatic identification and analysis efficiency of the flowering branch thorns is improved.

The application discloses a power distribution network wire state recognition method and device, terminal equipment and storage medium, and belongs to the power distribution network field. The method is: inputting a flight image of a wire to be recognized into a preset backbone network to obtain a preliminary feature map output by the backbone network; inputting the preliminary feature map into an MDPE module to enable the MDPE module to perform multi-dimensional feature perception processing and noise suppression processing on the preliminary feature map to obtain a deep semantic feature map; inputting the deep semantic feature map into a preset HSFPN module to enable the HSFPN module to perform attention feature fusion on the deep semantic feature map to obtain a multi-scale fusion feature map; inputting the multi-scale fusion feature map into a detection head to obtain key point coordinates and contour segmentation masks of the wire to be recognized, and determining a wire state of the wire to be recognized according to the key point coordinates and the contour segmentation masks. Through implementation of the application, the problem of low power distribution network wire state recognition efficiency in the prior art can be solved.

The application relates to the technical field of image segmentation, and particularly discloses a focused ultrasound micro-damage image segmentation method based on a composite energy field guided Snake model. In view of the problems of single external energy construction, insufficient anti-interference capability, poor time sequence stability and the like of the existing focused ultrasound micro-damage image segmentation method in processing weak boundary, strong noise and dynamic expansion ultrasound images, the external energy item based on regional texture features, boundary structure and dynamic information is integrated into a composite energy field, multi-source joint driving is applied to the contour evolution of a Snake active contour segmentation model from two dimensions of spatial structure and time evolution, the interference of false critical points and the phenomena of contour drift, collapse and border crossing are significantly inhibited in weak boundary, strong noise and damage dynamic expansion ultrasound images, stable and accurate segmentation of the damage region in a complex ultrasound background is realized, and good time sequence consistency is maintained.

The application relates to a post-processing optimization method and device based on image instance segmentation, which comprises the following steps: obtaining one or more segmentation images based on image instance segmentation; extracting one or more contour segmentation lines of a discrete mask in each segmentation image, and judging the hole attribute of each discrete mask according to the pixel difference in each contour segmentation line; filling each discrete mask with the hole attribute; calculating the area and length relationship between each mask based on the mask segmentation image of each segmentation image and the segmentation image with the hole attribute, identifying whether each mask overlaps according to the area and length relationship, filtering the overlapping area, and identifying and filtering the island area of each segmentation image. According to the application, the overlapping, island and hole existing after image instance segmentation are identified and filtered through the overlapping area and length relationship of the mask, so that the segmentation precision of the instance contour is improved.

This invention relates to the technical field of fire extinguishing monitoring, and more particularly to a fire extinguishing monitoring system based on image processing. The system acquires smoke concentration and images within a space through a smoke acquisition unit and an image acquisition unit, respectively. An image preprocessing unit preprocesses the acquired images, making them more effective. Simultaneously, a YOLOv7 model in the fire area contour recognition unit identifies the fire area, and a threshold method in the fire area contour segmentation unit segments the identified fire area. This allows for a more accurate determination of the actual area ratio of the fire area to the entire image. The system then uses this actual area ratio to determine whether the monitored fire area can be accurately identified, and based on the determination results, identifies the reasons for non-compliance, thereby providing corresponding fire extinguishing plans and improving fire extinguishing efficiency.

The invention relates to the technical field of intelligent sorting of juvenile crabs, in particular to a parameter optimization method for self-adaptive sorting of juvenile crabs. Obtaining images of juvenile crabs; preprocessing the image, extracting the abdomen contour features of the juvenile crabs, and obtaining shared working condition parameters at the same time; outputting a sex discrimination result and discrimination confidence based on the abdomen contour features; a shared working condition parameter and a contour parameter are introduced to establish a belly contour recognition function relationship, and a natural recognition set for representing a double-view recognition result and credibility is formed; further introducing an asymmetric judgment parameter, generating a judgment threshold and an adjustable range thereof, forming a judgment set, and enabling the tolerance of the first type of error classification to be lower than that of the second type of error classification; and constructing an optimization target based on a natural identification set and a judgment set, solving in a parameter adjustable range to obtain an identification parameter optimization set comprising a contour segmentation threshold value, denoising smooth intensity, a dual-view fusion threshold value and an asymmetric misclassification weight parameter, and updating identification parameters according to the identification parameter optimization set so as to adapt to working condition changes and reduce misclassification risks.

Embodiments of the present application relate to the technical field of digital medical treatment, and specifically provide a training method and device of an image processing model, a terminal and a storage medium. The method comprises: obtaining a first picture and a second picture; performing image filtering processing on the first picture by using an initial image processing model to obtain a third picture; performing image segmentation on the second picture to obtain a first foreground image and a first background image, and performing image segmentation on the third picture to obtain a second foreground image and a second background image; obtaining a contour segmentation error according to the first foreground image, the first background image, the second foreground image and the second background image; calculating a pixel error of the second picture and the third picture; constructing a loss function according to the contour segmentation error and the pixel error; and iteratively updating the initial image processing model based on target training data and the loss function to obtain a target image processing model. The model improves the reliability of image filtering and provides good support for subsequent image tasks.

The invention provides an ultrasonic image sequence segmentation method and device, and the method is characterized in that the method comprises the following steps: an original template obtaining step: obtaining an original contour template of a target organ, and the original contour template shows the contour form of the target organ in a natural state; an image acquisition step: acquiring an ultrasonic image sequence of a plurality of image frames of the target organ in a deformation state due to extrusion of an instrument, and positioning an iteration starting frame in the ultrasonic image sequence; a deformation field construction step: based on the original contour template and the iteration start frame, establishing a deformation field representing a space mapping relation between a natural state and a deformation state of the target organ; and a contour segmentation step: based on the original contour template and the deformation field, obtaining contour information of the target organ in each image frame of the ultrasonic image sequence in an inter-frame iteration mode, and obtaining a contour sequence of a deformation state of the target organ.

The application provides an angiography image segmentation method and device, electronic equipment and readable storage medium. The method comprises the following steps: acquiring an angiography image, and inputting the angiography image into a pre-trained neural network model; after each layer of the neural network model performs convolution operation and activation operation on the input of the layer, the output of the layer is obtained through a blood vessel edge information enhancement operation of an edge and connected domain enhancement module; and the neural network model outputs a blood vessel contour segmentation result of the angiography image. By using the proposed edge and connected domain enhancement module and adding it to the neural network model layer by layer, the blood vessel edge information of different feature scales is fused, the segmentation result can better retain the edge information of the blood vessel contour and the connectivity between the contour regions, which is conducive to realizing an accurate and complete blood vessel contour segmentation result and helps to improve the accuracy and time efficiency of evaluating the blood vessel stenosis degree.

The application discloses a target organ contour segmentation method and related products. The method comprises the following steps: acquiring a to-be-processed image, wherein the to-be-processed image comprises a target organ; extracting a first image containing high-frequency information and a second image containing low-frequency information from the to-be-processed image; performing up-sampling processing on the second image to obtain a third image; fusing the first image and the third image to obtain a fourth image; and segmenting the contour of the target organ in the to-be-processed image according to the fourth image to obtain a first segmentation image.

The application belongs to the technical field of medical devices, and particularly relates to a flexible endoscope autonomous path planning method, device and system, wherein the flexible endoscope autonomous path planning method comprises: pre-processing an endoscope image; obtaining a visual depth map from the endoscope image and constructing a three-dimensional point cloud map; identifying and contouring a target in the endoscope image to obtain target position information; fusing the three-dimensional point cloud map and the target position information to construct a dynamic environment model; generating a safe path according to the dynamic environment model and movement constraint conditions of an endoscope catheter; and using a deep deterministic policy gradient algorithm to generate a movement strategy according to a current environment state, with the safe path as a reference trajectory. Through visual depth recovery and reinforcement learning strategy, the application realizes real-time three-dimensional environment reconstruction, autonomous obstacle avoidance and accurate tracking of a moving target of the endoscope in a dynamic human body cavity.

The application relates to the technical field of image segmentation, and particularly discloses a focused ultrasound micro-damage image segmentation method based on a composite energy field guided Snake model. In view of the problems of single external energy construction, insufficient anti-interference capability, poor time sequence stability and the like of the existing focused ultrasound micro-damage image segmentation method in processing weak boundary, strong noise and dynamic expansion ultrasound images, the external energy item based on regional texture features, boundary structure and dynamic information is integrated into a composite energy field, multi-source joint driving is applied to the contour evolution of a Snake active contour segmentation model from two dimensions of spatial structure and time evolution, the interference of false critical points and the phenomena of contour drift, collapse and border crossing are significantly inhibited in weak boundary, strong noise and damage dynamic expansion ultrasound images, stable and accurate segmentation of the damage region in a complex ultrasound background is realized, and good time sequence consistency is maintained.

The present application relates to the technical field of flower phenotype analysis, and particularly relates to a field fresh flower intelligent grading and labeling method and system, three-dimensional images are generated by fusing two-dimensional and depth images collected synchronously, a shape contour is extracted through an instance segmentation network, a crown, a stem and a leaf feature map are segmented and output, and corresponding shape, size and appearance parameters are calculated, then a spatial offset of a crown barycenter and a stem axis is calculated, and the spatial offset is combined with the aforementioned parameters to construct a comprehensive growth feature vector, then a current quality grade is determined based on weighted summation of breed evaluation weights, meanwhile, the vector is combined with a historical sequence to fit a growth trajectory, a dynamic growth evolution model is constructed, and a remaining growth time is obtained, finally, the quality grade and the remaining time are encoded, and are mapped to a perspective projection position corresponding to a two-dimensional image to generate a grading and labeling image. The present application combines static calculation with dynamic deduction, and realizes visual and intuitive labeling of fresh flower states.

The invention discloses an industrial component contour segmentation method based on diffusion enhancement and guided filtering, and belongs to the technical field of computer vision and image processing, and the method comprises the steps: a rough mask rapid generation step: carrying out the time sequence correlation and region focusing through a lightweight segmentation network cooperating with a tracker, and generating a rough mask; a diffusion edge iterative optimization step: carrying out multi-step condition denoising iterative correction on the edge region by adopting a discrete diffusion process; a guided filtering high-frequency transmission step: transmitting high-frequency edge information to the mask by taking the image of the industrial component as guidance; according to the method, the sub-pixel-level edge positioning precision is realized on the premise that the single-frame reasoning time does not exceed 0.6 second, and the technical problem that the segmentation precision and the speed are difficult to consider at the same time in the high-speed operation scene of the industrial production line is effectively solved.

The application belongs to the technical field of computer vision and aerial intelligent perception, and particularly relates to a kind of aerial small target detection segmentation method and system based on reliability guidance.Method includes: collecting the RGB aerial image and infrared aerial image of the same scene and performing coarse registration;Quality degradation assessment is carried out, and modal reliability map is generated;Estimate cross-modal local offset field and generate sparse candidate area combined with modal reliability map;RGB image, infrared image, modal reliability map and offset information are input into reliability-offset joint gated fusion network to obtain fusion feature map;Based on the fusion feature map, small target detection, frame guided contour segmentation and detection segmentation closed loop correction are carried out, and target category, position and contour information are output.The application improves the detection and segmentation stability of sparse small targets in complex environment in dual-mode aerial images through modal reliability modeling, local offset compensation, region-level joint gated fusion and closed loop refinement mechanism.

The invention relates to a hysteroscope focus segmentation quantification method and system based on deep learning. The method comprises the steps of video quality perception key frame screening, color space normalization and adaptive contrast enhancement preprocessing, focus segmentation of a residual attention encoder and decoder, automatic quantization of focus geometric parameters, multi-view three-dimensional reconstruction volume estimation and structured report generation and longitudinal comparison. The method has the advantages that the focus contour segmentation precision is enhanced through the boundary perception loss function, multi-dimensional geometric parameters are automatically extracted, the three-dimensional volume is estimated, and objective data support is provided for disease grading and curative effect evaluation.