450 results about "Low contrast" patented technology

The invention relates to an underwater image enhancement method based on wavelet Mama, which aims at the problems of color shift, low contrast ratio and fuzzy details of an underwater image, extracts shallow layer features of an underwater low-quality image, inputs the shallow layer features into a multi-scale coding network, and performs joint enhancement on low-frequency color information and high-frequency detail features by using a wavelet Mama unit. Global semantic features are obtained by combining down-sampling layer-by-layer compression, and global modeling of color offset correction and contrast enhancement is completed; the spatial resolution is recovered through up-sampling, and reconstruction and enhancement of texture details and color information are realized in combination with jump connection and a wavelet Mama unit; and finally, the features are mapped to an image domain and fused with input image residual errors, and an enhanced underwater image with natural color, clear texture and balanced contrast is generated. According to the method, frequency domain-space domain joint feature modeling is realized through the wavelet Mama unit, and the color authenticity, the structural definition and the detail perceptibility of the underwater image are remarkably improved.

The invention belongs to the field of computer vision and industrial defect detection, and discloses a steel coil end face defect detection method and system based on CBAM-BiFPN and multi-loss optimization, and the method comprises the steps: image preprocessing and region extraction: extracting a steel coil end face region through OpenCV and other image processing methods; dividing the high-resolution image into a plurality of small blocks and performing data enhancement; the method comprises the following steps: constructing a YOLOv11 network based on CBAM-BiFPN, introducing a CBAM attention mechanism and BiFPN feature fusion structure, and constructing an improved YOLOv11 detection network; multi-loss function joint optimization training is carried out, and model training is carried out in combination with loss functions such as Focal Loss and WIoU; and fusion of detection results and defect reconstruction output: reconstructing original image defects of all tile image detection results in a space coordinate mapping and redundant region fusion mode, and realizing high-precision overall detection output. Compared with a traditional method, the method still has the high recognition capability in a complex background and low-contrast scene, the detection precision and stability are obviously improved, and higher industrial adaptability and practical value are achieved.

The invention discloses a multi-modal fusion defect detection method and system, and belongs to the technical field of intelligent detection and machine vision, and the system comprises a visible light sensor, a thermal infrared sensor, a hyperspectral sensor, a multi-band light source trigger control system, a modal preprocessing and alignment module, a cross-modal feature fusion module, and a defect detection and output module. According to the invention, three sensors are used to construct a multi-modal visual perception system, and a multi-band light source triggers a control system to complete image acquisition; after multi-modal information is subjected to preprocessing and cross-modal alignment through the modal preprocessing and alignment module, multi-modal fusion is achieved through the cross-modal attention module and the multi-scale feature fusion pyramid structure, and finally defect recognition and output are conducted through the defect detection and output module. According to the method, the defect identification precision can be obviously improved, and the method has obvious advantages especially for low-contrast and early-stage hidden crack defects, and has good expandability and deployment suitability at the same time.

The invention relates to an improved YOLO11n underwater target identification and detection method based on local and global perception, and belongs to the field of underwater intelligent identification. According to the method, a local bottleneck module is constructed to extract local fine-grained features, a multi-scale large convolution kernel module is introduced to expand a receptive field, and a multi-branch channel attention module, a space and channel combination grouping attention module and a cross-dimension feature fusion module are combined to adaptively enhance effective features and suppress environmental noise interference. And a local-global bottleneck module is further constructed, and local detail and global semantic collaborative modeling is realized through a cascade structure. And a Bottleneck module of the original YOLO11n is replaced by the module, so that an improved model is formed. After training, the method has higher detection precision and robustness on underwater data sets such as DUO, RUOD and the like, and is suitable for real-time target identification tasks in a complex underwater environment. According to the method, the problems of color distortion, low contrast, fuzzy details and difficulty in multi-scale target feature extraction of the underwater image are solved, and high-precision and real-time balance detection is realized.

The invention discloses a wild animal target detection method based on improved YOLO11. According to the invention, double-dynamic convolution is designed to improve the feature extraction and nonlinear expression capability of the feature extraction module on the wild animal image so as to flexibly adapt to the feature difference between visible light and infrared images; a multi-scale expansion attention mechanism is introduced to improve the multi-scale target detection precision; unified-IoU loss is introduced to solve the problem of unbalanced quality of a prediction frame; and a DySample dynamic up-sampling operator is introduced, so that the detail recovery capability is improved, and the feature distortion is reduced. The problems that a traditional YOLO feature extraction module cannot effectively process an image shot by a field infrared camera, and an infrared image depends on thermal radiation, is low in contrast, weak in texture, large in environmental thermal noise and large in feature difference with a visible light image and is difficult to adapt are solved, and traditional CIoU loss easily pays attention to and fits a low-quality prediction frame. And a sampling module cannot effectively reconstruct low-resolution infrared image features.

The invention relates to the technical field of tunnel crack detection and artificial intelligence edge calculation, and provides a tunnel apparent disease detection method based on deep learning and knowledge distillation, which comprises the following steps: step 1, introducing spectral domain information enhancement to an original tunnel image, the edge texture features of the disease area in the image are enhanced through methods such as multi-scale wavelet transform and small-scale enhancement. Step 2, constructing a high-performance teacher model, introducing a flexible up-sampling structure to adapt to feature recovery requirements of different levels of semantic information, introducing an efficient visual coding module to enhance feature fusion capability of different scale channels, and designing a scale adaptive weighted loss function at the same time; by introducing a frequency spectrum enhancement mechanism, structural features of disease areas with low contrast, fuzzy edges and the like are remarkably enhanced in an image preprocessing stage, clearer information input is provided for a model, and the stable recognition capability of a system in environments of uneven illumination, complex background and the like is enhanced.

The invention provides a medical image segmentation method and system based on spatial detail enhanced vision, and relates to the technical field of image processing. The method comprises the following steps: performing initial feature mapping on an input and preprocessed medical image to obtain an embedded feature map; the embedded feature map is input into an encoder for feature extraction, and multi-scale features are obtained; the encoder comprises a plurality of encoding stages, and the number of channels is doubled and the spatial resolution is halved through down-sampling operation between the encoding stages; the multi-scale features are subjected to up-sampling and spatial enhancement reconstruction step by step through a decoder, and a high-precision segmentation result is generated; wherein the decoder comprises a plurality of decoding stages, the decoding stages correspond to the encoding stages, and feature fusion is carried out between the corresponding stages of the encoder and the decoder through jump connection. According to the method, the boundary description precision and the segmentation robustness of the low-contrast image can be improved without increasing the linear complexity, and the method is suitable for medical image segmentation scenes of skin lesions, gastrointestinal polyps and the like.

The invention discloses a camouflage target detection method and system based on a double-domain fusion enhanced network, and relates to the technical field of target detection. Through the nonlinear double-domain fusion module, in combination with nonlinear mapping of a spatial domain and a frequency domain, key difference characteristics of a frequency domain amplitude spectrum and a phase spectrum are captured, the problem that the detection performance is reduced in a scene of low contrast and the like depending on an RGB spatial domain is solved, and the target discrimination degree is improved; based on a lightweight scale perception modulation converter and a double-feature fusion module, multi-scale features are extracted, aligned and fused, a semantic relation is integrated by means of cross attention, and the problems of detail loss and boundary fuzziness caused by scale diversity are solved; the context feature enhancement module integrates cross attention and edge auxiliary injection, accumulates multi-layer feature integration, gives consideration to a global boundary and a local structure, effectively reduces false detection, missing detection and edge roughness, and further enhances robustness through multi-layer auxiliary supervision.

The invention discloses a three-dimensional blood vessel image segmentation method and system. Belongs to the technical field of medical image processing and particularly relates to the technical field of three-dimensional blood vessel image segmentation. The method solves the following problems existing in a blood vessel segmentation task in an existing method: small blood vessel features are difficult to accurately extract in a CTA image which is low in contrast and contains noise and artifacts; global context modeling is difficult to consider and local correlation is difficult to guarantee, so that long-distance dependent modeling is insufficient or a local structure is fractured; limited by a fixed geometrical shape of a traditional convolution kernel, the traditional convolution kernel is difficult to adapt to deformation characteristics of a complex topological structure of a blood vessel, resulting in discontinuous segmentation or fuzzy boundary of a branch region. Channel dynamic grouping and energy-driven attention generation are achieved through a grouping self-adaptive attention module, and self-adaptive modeling of a blood vessel complex branch structure and a geometrical shape is achieved through a multi-scale space structure aggregation module in combination with a strip-shaped deformable convolution and cross-scale guiding mechanism.

The invention belongs to the technical field of image processing, and particularly relates to a composite material ultrasonic scanning image defect feature extraction method and system, and the method comprises the steps: obtaining an ultrasonic scanning image of a composite material, and carrying out the filtering processing; constructing a gradient outer product matrix of neighborhood pixel points of the pixel points and accumulating to obtain a local structure matrix; performing characteristic decomposition on the local structure matrix to obtain a characteristic value, and calculating a structure coherence factor according to the characteristic value; coupling potential energy is constructed in combination with a gray value and a structural coherence factor, and a low-gray and disordered defect signal is highlighted through nonlinear gain; and performing region segmentation by using the high-coupling potential energy points as anchor points, and extracting defect blocks. According to the method, the problem of low-contrast defect leak detection under the strong texture background is effectively solved by utilizing the essential difference between the background texture and the defect in the physical topology, and the defect feature extraction precision is remarkably improved.

The invention relates to the technical field of medical image processing, in particular to a medical image segmentation method based on a frame prompt enhanced SK-VM + + network, and the method comprises the steps: collecting a to-be-segmented image in a data set, and carrying out the three-channel copying of the to-be-segmented image; extracting focus bounding box information in the text annotation file to generate a binary mask; generating a box prompt visualization graph by using the binary box prompt mask; fusing the binary box prompt mask and the copied three-channel image to obtain a four-channel tensor; and taking the four-channel tensor as input, constructing an SK-VM + + model, and performing segmentation processing by using the SK-VM + + model. The medical image segmentation method solves the problem that the existing medical image segmentation method is difficult to accurately deduce the spatial context only by means of the pixel intensity in a low-contrast and high-noise environment, and is easy to cause false detection and missing detection.

The invention relates to the field of endoscope image processing, and particularly discloses an automatic lesion recognition system for endoscopy of the digestive system department, which is characterized in that after a preprocessed original endoscope image is acquired, a double-branch parallel processing architecture is used to acquire characteristics with low resolution and rich semantic information through a deep context branch, and the characteristics of the original endoscope image are acquired. The potential area of the focus is accurately deduced; meanwhile, the fine texture of the mucous membrane is captured in a lossless manner through shallow detail branches which keep high resolution in the whole process. Furthermore, through a context-guided asymmetric enhancement mechanism, a global view of a deep branch is utilized to generate an uncertainty perception attention map as a reference, and weak detail features corresponding to a potential focus area in a shallow branch are accurately irradiated and adaptively enhanced. Thus, a conservative enhancement strategy is adopted in an uncertain focus area, background noise is effectively inhibited, and therefore the detection sensitivity and robustness of low-contrast and flat focuses are fundamentally improved.

The invention provides a method and a system for restoring a bamboo-strip character image based on multi-granularity feature guidance, and innovatively designs a coarse-fine two-stage restoration network and end-to-end multi-task loss joint training aiming at the problems of structure-texture confusion, non-uniform degradation, low contrast ratio and the like of the bamboo-strip image. In the coarse repair stage, a font texture and structure double-reconstruction sub-network is used for separating semantics from a source; in the fine repairing stage, multi-scale dynamic range distribution diagram self-attention (Mdma) is provided, pixels are dynamically classified according to degradation intensity, and long-short range dependence joint modeling is achieved; an adaptive mask is designed to sense pixel shuffling downsampling (Ampd), sampling is guided by mask confidence, damage position information is kept, and artifacts are inhibited. Five mainstream methods are compared on a homemade 313 bamboo strip single word data set, the PSNR, the SSIM and the FID are optimal under 0-60% irregular masks, visual evaluation of real missing samples is natural in texture, the structure is complete, and the readability of the bamboo strip characters and the subsequent recognition accuracy are effectively improved.

The invention discloses an underwater image enhancement method and system based on binocular vision and polarization imaging. The method comprises the following steps: firstly, synchronously acquiring an orthogonal polarization image pair through a binocular polarization imaging unit, recovering scene depth information from the polarization image pair by using a binocular stereoscopic vision technology, and calculating a transmissivity graph; analyzing the polarization characteristic difference between target reflected light and backscattered light, establishing a polarization difference model, and separating background scattered light; solving an underwater imaging equation in combination with the depth information and a polarization difference model to obtain a preliminary restoration result of the target reflected light; and finally, further improving the image quality through the steps of multi-scale spectrum analysis, adaptive filter construction and post-processing by adopting a spectrum adaptive image enhancement technology. The system comprises a binocular polarization imaging unit, a calculation processing unit and an active illumination unit. According to the method, the problems of color distortion, low contrast and fuzzy details of the underwater image are solved, and the visual quality and usability of the underwater image are improved.

The invention discloses an infrared small target detection method based on comparative learning and frequency gradient feature fusion, and aims to improve the detection precision. A current method faces three challenges that the contrast of an infrared image is low, space information is limited and is interfered by clutters, so that global context is missing, and robustness is insufficient; target structure features are weak and are similar to background gray, and texture extraction is difficult; an image background is complex, a target is tiny, missing detection and false alarm are caused frequently, and the distinguishing capacity of a foreground and the background is affected. For the first problem, a frequency attention perception fusion module is designed to improve semantic consistency and positioning precision. In order to solve the second problem, a textural feature enhancement module is designed to enhance textural feature representation. In order to solve the third problem, a temperature sensing comparison learning module is designed to improve the foreground and background distinction degree. In combination with the three designs, the model designed by the invention can accurately detect the small target in the infrared image, and is worthy of vigorous popularization.

The invention relates to the technical field of image recognition, in particular to a medical image intelligent evaluation system based on image recognition. The system comprises an image registration module, an image segmentation module, a preliminary fusion module, an image evaluation module, an optimization feedback module and an image output module. According to the method, the CT image and the MRI image are subjected to image registration, spatial alignment is ensured, then the region of interest is segmented and fused, namely, the skeleton contour in the CT image is superposed on the MRI image, and due to the fact that motion artifacts generated by movement of a patient in the scanning process possibly exist in the original CT image, the skeleton contour in the CT image is fused with the motion artifacts in the MRI image. If the skeleton contour does not exist in the MRI image, the overlapping degree and the blank degree of the skeleton contour and the anatomical structure edge of the MRI image are analyzed, and an optimized registration parameter or segmentation parameter is fed back, so that when the segmentation network is trained, the segmentation precision under the conditions of artifacts and low contrast is improved, spectrum and texture information of the two images is reserved to the maximum extent, and the fusion effect is guaranteed.

The invention provides an optic disk real-time continuous tracking method and device, and relates to the technical field of medical image processing. According to the method, the video sequence of the ophthalmologic operation is acquired and preprocessed, and then the preprocessed video sequence is subjected to multi-scale feature extraction and optic disc target detection, so that an optic disc detection target is obtained; the optic disk detection target is based on Kalman filtering prediction and appearance feature matching, an adaptive gating matching mechanism is introduced to carry out inter-frame target trajectory association, a target trajectory is maintained in combination with an appearance feature cache updating mechanism and ReID cosine similarity retrieval, and a trajectory tracking result is output; wherein the trajectory tracking result comprises a continuous tracking result of the optic disc position, the confidence coefficient, the trajectory identifier and the timestamp. According to the invention, the detection precision of the small-scale optic disc is improved, the shielding recovery and track continuity are enhanced, and high-precision, high-robustness and real-time optic disc continuous tracking is realized in complex operation scenes of strong reflection, low contrast, local shielding and the like.

The invention relates to the field of image processing and target detection, in particular to a crack detection and segmentation method based on depth perception and structural feature enhancement. The method comprises the following steps: acquiring original image data containing cracks; performing multi-scale feature extraction on the image; multi-scale feature extraction is realized in combination with a C3K2SAConv module capable of switching cavity convolution; introducing a SimAM attention mechanism to enhance crack area response; a CCM crack convolution module is designed to extract a slender and fractured crack edge structure, and detection and segmentation accuracy and boundary integrity are improved; and finally completing positioning and segmentation output of the crack region. The method improves the recognition accuracy and edge integrity of the structural crack in a complex background and low contrast environment, and is suitable for application scenes of structure monitoring, intelligent maintenance, automatic driving environment perception and the like of traffic infrastructures such as roads, bridges and the like.

The invention discloses an EPE co-extrusion adhesive film surface defect detection system based on image recognition, and relates to the technical field of industrial product quality detection, and the EPE co-extrusion adhesive film surface defect detection system comprises an image acquisition unit, an image processing unit and a result output unit, and the image acquisition unit comprises at least three light sources with different optical configurations and a camera which are used for synchronously or sequentially imaging the same detection area. According to the EPE co-extrusion adhesive film surface defect detection system based on image recognition, by adopting a multi-mode optical imaging and depth feature decoupling fusion technology, the problems of background texture interference and difficulty in low-contrast defect recognition in EPE co-extrusion adhesive film surface defect detection are effectively solved. The surface information is obtained from different physical dimensions through multi-light-source configuration, background textures and defect features are separated in combination with a feature decoupling algorithm, the recognition capacity of difficult-to-detect defects such as weak indentations, water waves and superficial impurities is improved, the occurrence probability of false detection and missing detection is reduced, and the reliability of a detection result is ensured.

The invention discloses a PICC catheter recognition method and system based on image recognition, and relates to the technical field of medical image analysis, and the system comprises a medical image input module which is used for obtaining X-ray image data; the improved YOLO-Seg segmentation module is connected with the medical image input module, and the improved YOLO-Seg segmentation module comprises a CSPDarknet backbone network of which the channel number is expanded to 1280, a feature pyramid network for outputting a 4-scale feature map, and a segmentation head containing 32-channel prototype mask generation; and the double attention enhancement module is integrated in the improved YOLO-Seg segmentation module, comprises a channel attention module and a space attention module, and is used for enhancing low-contrast catheter feature response. According to the method, the slender feature extraction capability is enhanced through the 32-channel prototype mask generator, and a catheter continuity reconstruction algorithm based on skeleton analysis is adopted, so that the Dice coefficient of the PICC catheter is improved, the fracture rate of the catheter is reduced, the problem of slender structure segmentation failure is solved, and the catheter segmentation precision is improved.

The invention discloses an intelligent optical sensing system based on multispectral fusion, and relates to the technical field of optical sensing, and the system comprises a multispectral image collection module which is used for synchronously collecting original image data of a target scene under three spectral channels of visible light, near-infrared and short-wave infrared; the space-time registration and preprocessing module is used for carrying out high-precision space-time registration and radiation correction on the images of different spectrum channels; the self-adaptive feature extraction and fusion module is used for extracting multi-scale spectral features from the registered multi-spectral image and dynamically selecting and weighting a fusion strategy according to scene content; and the lightweight decision network module outputs a final target identification and state discrimination result. According to the technical scheme, the time-space consistency of multispectral data acquisition can be realized, the cross-band registration precision is improved, the robustness under the conditions of low contrast, shielding and severe weather is enhanced, meanwhile, the calculation complexity and memory occupation are greatly reduced, and the method is suitable for edge calculation equipment with limited resources.

The invention discloses a method and a system for visually detecting outer surface defects of steamed stuffed buns, relates to the technical field of visual detection of defects, and aims to stretch a low-contrast area to a higher gray level space through area mean value logarithm mapping processing, enhance the overall image layering sense and contribute to revealing tiny defects hidden under a strong interference background. Through fusion calculation of first-order gradient and two-dimensional Laplacian response in the horizontal direction and the vertical direction, composite response can be formed to structural changes of recesses, cracks, wrinkles and the like in different directions on the surface of the steamed stuffed bun, so that texture contours which are originally weakened in strong light, moisture or compressed areas are presented in a highlight manner; the problem that the structure edge is fuzzy is solved. And local high-frequency detail enhancement processing is performed on the normalized image, so that low-contrast defect areas which are visually fuzzy due to humidity or illumination interference originally, such as light-color wrinkles, local collapse, edge indentations and the like, are enhanced and expressed in a response diagram.

The invention discloses a long-distance binocular camera calibration optimization method based on multistage feature enhancement, and the method comprises the steps: carrying out the processing of a condition that a long-distance calibration plate has a highlight region and is fuzzy in angular points, employing a bilateral filter to suppress the reflection of a highlight mirror surface, reducing the reflection of light, and maintaining the definition of an image; the checkerboard texture is enhanced by adopting a CLAHE algorithm in combination with blocking processing and a contrast gain threshold value; a Sobel operator and a Harris corner response function are fused, gradient direction distribution characteristics of a pixel neighborhood are extracted through the Sobel operator, weighted fusion is carried out on the gradient direction distribution characteristics and the Harris corner response function, and the response intensity and specificity of a corner area are remarkably enhanced. A three-level image preprocessing framework including reflection suppression, contrast enhancement and corner enhancement is constructed, the problem of feature extraction of a traditional method under a complex illumination condition is solved, the corner area detection capability is enhanced, the problems of specular reflection noise, low contrast and corner blur are effectively solved, and the calibration precision is improved.

The invention discloses an automatic focusing method and system of a cell slide scanner, and relates to the related field of optical equipment focusing technology.The method comprises the steps that a slide is previewed and scanned, a multi-mode preview image and physical height sensing data are synchronously collected and preprocessed, and then the pre-processed image is obtained; the multi-modal preview image and the corresponding height data are input into a pre-training focus prediction model, a focusing strategy graph aligned with a slide coordinate space is output, and the graph comprises the predicted optimal focus offset, the focusing function recommendation identifier, the focus verification confidence coefficient and the search radius of each view point; according to the focusing strategy graph, a final focus is determined through feedforward positioning, limited verification and conditional adjustment, a scanner objective table is controlled to carry out formal scanning at the final focus position, and a complete scanning image of the cell slide is collected. The problems that a traditional focusing method is long in time consumption and poor in adaptability to complex samples with low contrast, high background fluorescence and scratch pollution are solved, and the adaptive capacity of the focusing process is improved.

The invention provides an unmanned aerial vehicle inspection tour anti-dazzle visual angle optimization method and system based on background perception. Acquiring a key detection area of the inspection target; candidate shooting poses of the unmanned aerial vehicle are generated for the key detection area; constructing a glare risk function based on the material type parameters, the solar incident angle and the deviation angle between the camera observation direction and the mirror reflection direction; constructing a contrast scoring function based on the brightness corresponding to the background type and the reflection brightness corresponding to the key detection area; a target shooting position is selected according to the glare risk function and the contrast scoring function, and an unmanned aerial vehicle inspection track is generated; the shooting visual angle of the unmanned aerial vehicle is automatically selected and adjusted in a complex natural light environment, the image failure risk caused by high reflection, silhouette and low contrast is reduced, and the detectability of defects of key parts is improved.

The invention relates to a scratch detection method, device and equipment for a transparent film and a medium, and the method comprises the steps: firstly carrying out image multiplication processing and dynamic range mapping on a transparent film gray level image, and obtaining a target gray level image through threshold segmentation; calculating a pixel gradient magnitude based on the image, synthesizing an edge gradient image, and generating a first salient image through Gaussian filtering and gray linear transformation; meanwhile, logarithmic transformation is carried out on the original target image to obtain a second salient image; secondly, respectively calculating gray average values of the two salient images, determining a self-adaptive segmentation threshold by combining a preset threshold, extracting a defect region through double-image threshold segmentation, and obtaining an intersection to obtain an initial scratch region; and finally, screening according to a preset area condition to obtain a final scratch area. According to the method, through the multi-feature fusion and self-adaptive threshold technology, the problems of low contrast, uneven illumination and the like in scratch detection of the transparent film are effectively solved, the detection efficiency and accuracy are remarkably improved, and the method has high engineering application value.

The invention discloses an aerial target detection method based on dynamic attention and double-frequency feature enhancement. The method comprises the following steps: 1) extracting multi-scale features by using a backbone network in combination with a low-light feature fusion module, and enhancing detail expression in a low-light and low-contrast scene; 2) sparse modeling is carried out on deep features through a dynamic sparse attention module, key attention connection is reserved, and the global semantic ability is improved while the calculation amount is reduced; 3) inputting the deep enhanced features and the shallow features into a double-frequency feature enhancement module, and respectively modeling low-frequency background and high-frequency details to realize foreground highlighting and background suppression; according to the unmanned aerial vehicle aerial image target detection method, the problems of low illumination, complex background and small target detection are effectively solved, and the precision and robustness of unmanned aerial vehicle aerial image target detection are remarkably improved.

The invention discloses a code nail stamping defect detection method and system based on machine vision, and belongs to the technical field of industrial automation quality control. The method comprises the following steps: acquiring multi-modal visual data such as a two-dimensional bright field image, a two-dimensional dark field image and three-dimensional contour data of a to-be-detected code nail; preprocessing the multi-modal visual data to obtain bright field, dark field and three-dimensional image data adaptive to a neural network model; inputting the three kinds of image data into a preset neural network model for defect identification processing, and outputting a defect segmentation mask and defect category information; and finally, performing quantitative analysis on the defects based on the output defect information, and judging whether the product is qualified or not according to an engineering specification threshold value. According to the method, multi-dimensional and complementary visual information is fused, and a specially designed neural network model is adopted for analysis, so that the problem of insufficient detection capability of tiny, low-contrast and three-dimensional geometric defects in the prior art can be effectively solved.

The invention discloses a free-form surface non-zero interference detection device and method based on thin film deformable mirror wavefront compensation. The device comprises a laser source, a polarization regulation and control module, a film deformation mirror surface type control and monitoring module, an interference measurement module and a data processing and control module. The control and monitoring module performs closed-loop adjustment on the deformed mirror surface type to realize high-precision wavefront compensation; in the interference measurement module, reference light is generated by laser collimation, measurement light forms interference with the reference light after being modulated by a deformable mirror and a free-form surface to be measured, four frames of phase-shifting interferograms with 45-degree phase difference are extracted, and profile data after compensation are obtained by combining an algorithm; and carrying out point-to-point subtraction on the data and deformed mirror surface type data to obtain the surface type of the free-form surface to be detected, and completing high-precision detection. The problems that interference fringes are low in contrast ratio and cannot be distinguished due to aberration in free-form surface detection are effectively solved, the large-curvature and large-inclination free-form surface detection precision and efficiency are improved, and the method is suitable for high-precision free-form surface optical element detection scenes.

The invention discloses a medical image tumor segmentation model based on a TransUNet framework and a construction method and a segmentation method thereof, the construction method comprises the steps that a segmentation model is constructed based on TransUNet, an encoder comprises a CNN module and a user-defined VSS module which are connected in sequence, the output end of each convolution block of the CNN module is input into a decoder of the segmentation model through a jump connection layer, and the output end of each convolution block of the user-defined VSS module is input into the decoder of the segmentation model through a jump connection layer; the user-defined VSS module comprises a plurality of novel conversion layers which are stacked and connected in series based on a Mama module and a focusing linear attention module, and the user-defined VSS module performs multi-level feature interaction and global modeling on a first coding feature output by the CNN module to obtain a second coding feature; the decoder performs feature fusion and up-sampling on the second encoding feature and the output of the jump connection layer to obtain a segmented image for the sample image. According to the method, the segmentation robustness and precision of the breast ultrasound image with low contrast, high noise and fuzzy boundary can be remarkably improved.