1553 results about "Color image" patented technology

The invention provides an aluminum alloy surface oxidation spot defect identification method and device based on machine vision, and relates to the field of intelligent manufacturing and industrial automation, and the method comprises the steps: obtaining an aluminum alloy surface color image, and carrying out the preprocessing of the image, so as to extract a brightness component image; self-adaptive threshold segmentation of local contrast enhancement is carried out on the brightness component image, a defect area binary mask is generated, morphological connected domains are extracted according to the mask, and three basic feature indexes of the area pixel value, the contour Fourier descriptor complexity and the area gray scale standard deviation contrast of each connected domain are calculated; and extracting and marking a connected domain boundary, verifying a boundary closed topological structure, and dynamically generating a curvature-driven self-adaptive sampling point through multi-scale B-spline curvature extreme value detection. Through optical-algorithm-process three-level collaborative innovation, the curved surface reflection false alarm rate is reduced, the pinhole detection rate is increased, and the boundary precision is + / -0.2 pixel.

The invention provides a self-adaptive nonlinear image enhancement method and system for a low-light scene of a mobile terminal, and relates to the technical field of image enhancement, and the method comprises the steps: carrying out the image preprocessing and noise reduction, and carrying out the graying and noise suppression of an input color image through a local variance self-adaptive algorithm; adaptive down-sampling is carried out, and the down-sampling proportion is dynamically adjusted according to the image resolution and the content complexity, so that the processing efficiency is improved; brightness adaptive enhancement is carried out, and the overall brightness of the image is rapidly improved by adopting an Otsu method and a lookup table; contrast nonlinear enhancement: enhancing image details and contrast in combination with a Laplace operator and local mean adjustment; and color restoration: restoring the resolution through bilinear interpolation and performing weighted fusion to realize natural color reconstruction. And finally, a high-quality image of which the brightness, the contrast ratio and the color are remarkably improved is output. According to the invention, the recognition accuracy and processing efficiency of the low-illumination image are improved.

The invention relates to the technical field of foamed silicone rubber surface detection, and discloses a foamed silicone rubber surface detection method based on image visual identification, which comprises the following steps: acquiring a color image of a foamed silicone rubber surface, carrying out graying and normalization processing, and generating a diffuse reflection effective detection area by using a polarization filter difference technology; color abnormity is detected based on an image gridding and gray statistical method, meanwhile, pore and microbubble defects are detected in combination with minimum value seed point extraction and an eight-neighborhood region growing algorithm, finally, various masks are fused, the defect area and number are counted, and a quality inspection report is generated; the method can effectively inhibit highlight interference and accurately identify uneven colors and pore microbubbles, the detection process does not need external training, calculation is efficient, and the method is suitable for rapid detection and quality control of surface defects of foamed silicone rubber.

The invention relates to the technical field of automatic spraying, and particularly discloses a spraying system based on multi-modal vision and artificial intelligence self-correction, comprising a multi-modal vision acquisition module for acquiring color image information, depth information and infrared feature information; the control module is used for generating a spraying sensing model according to the information of the multi-modal visual acquisition module; the control module carries out spraying area identification, track planning and spraying parameter decision making based on the spraying sensing model; the spraying execution module is used for spraying; and the feedback self-correction module iterates the spraying parameter decision in the control module based on a reinforcement learning algorithm according to the difference between the actual coating state information and the expected state. The control method based on multi-modal vision and artificial intelligence self-correction is applied to a spraying system based on multi-modal vision and artificial intelligence self-correction. The scheme is used for solving the problems that the spraying quality of complex workpieces is not high due to the single sensing dimension of an existing spraying system, and the production flexibility is poor due to the rigid control mode.

The invention discloses an electronic cabin assembly quality detection method based on three-dimensional vision. The method comprises the following steps: firstly, generating data for comparison during detection for a standard electronic cabin; shooting all to-be-detected objects in the to-be-detected electronic cabin to obtain a depth image and a color image of each to-be-detected object; according to the depth image and the color image of the to-be-detected object, combining with the data of the standard electronic cabin for comparison during detection; and whether screws of the electronic cabin to be detected are neglected and not installed in place, whether cable plugs are neglected and not installed in place, whether connectors and cable plugs in the wiring unit are wrongly matched, and whether the bending radius of cables is too small are detected in sequence. Through fusion of depth information, three-dimensional quantitative detection of the assembly quality is realized, and the detection precision, reliability and automation level are significantly improved.

The invention discloses an uncertainty mapping method and device based on selective learnable depreciation, and the method comprises the steps: S1, collecting color images and depth maps at different time points, and projecting the color images and the depth maps to a three-dimensional coordinate system; s2, predicting an evidence vector for each pixel by using an evidence generation network; s3, calculating total evidence intensity based on the evidence vector, and defining credibility and uncertainty based on the total evidence intensity; s4, constructing a noise mask to represent the observation quality, and predicting a damage coefficient by using a selective damage network; s5, the evidence intensity is adjusted in a zooming and depreciation mode; and S6, outputting a final probability, carrying out basic probability distribution of multi-frame fusion observation on the three-dimensional voxel grid, and constructing a semantic map containing uncertainty estimation. According to the method, the conflict rate and uncertainty in the semantic map fusion process can be effectively reduced, and the accuracy and credibility of the map are remarkably improved.

The invention discloses an AI-based child story video generation method and system, and relates to the technical field of artificial intelligence and multimedia crossing, and the method comprises the steps: generating a script from an original text input by a user through constructing an AI model fusing an emotion modeling capability; constructing an image generation combination model, defining a joint loss function, calculating an edge intensity graph of the contour image by using a Sobel edge detection algorithm, calculating an optical flow field of frame change by using a block matching algorithm, and performing color image dynamic frame alignment; a fine tuning WaveNet model is used to generate audio; through constructing an image generation combination model, combining a StyleGAN3-T model and an LDM model, defining a joint loss function, and using a Sobel edge detection algorithm and a block matching algorithm to calculate an edge intensity graph and an optical flow field, dynamic frame alignment of a color image is realized, and inter-frame continuity of a generated video is improved.

The invention relates to an intelligent inspection data processing method and system. The intelligent inspection data processing method comprises the following steps: analyzing an infrared image in an inspection scene to obtain an image resolution, a temperature matrix and a pseudo-color image of the infrared image; synchronizing annotation information of the infrared image and the visible light image in the inspection scene to obtain multi-modal annotation data; based on preset reference object information, performing defect quantitative analysis on the multi-modal labeling data, and generating a real physical size quantitative result of the defect; and carrying out associative storage on the image resolution, the temperature matrix, the pseudo-color image and the quantification result. According to the invention, the whole-process optimization of the inspection data can be realized, and the processing stability, the defect detection accuracy and the system practicability are effectively improved.

The invention discloses a LiDAR-IMU-camera tight coupling positioning and mapping method and device for a mobile platform, and belongs to the technical field of robot SLAM. Synchronously triggering the color camera, the laser radar and the IMU through hardware, and unifying timestamps; performing compensation and distortion removal on the laser point cloud motion by the IMU data; based on curvature, intensity and density self-adaptive downsampling, local plane fitting errors are used for distributing observation weights; the IMU pre-integration pose is used as an initial value, point-to-surface registration of the weighted point cloud and the local map is carried out, and a laser-IMU tight coupling odometer factor is obtained; the color image and the laser intensity graph are fused into a multi-mode loopback descriptor, and loopback factors are generated through global retrieval and geometric verification; and inputting an odometer, an IMU, a laser loopback factor and a visual loopback factor into an increment factor graph optimizer, jointly solving a global optimal key frame pose, and outputting a dense laser point cloud map and a color visual point cloud map. The method can be operated in real time on an embedded platform, effectively inhibits drifting, and realizes centimeter-level global consistent positioning and mapping.

The invention relates to the technical field of liquid crystal display, and provides liquid crystal display equipment based on a field order and a color image display method. An image to be displayed can be decomposed into three sub-frames based on a field sequential display technology, each sub-frame corresponds to a color field, when a liquid crystal panel is driven according to a main color gray scale value of the color field corresponding to each sub-frame, at least one backlight source corresponding to each color field is firstly driven to emit backlight of a first duration within a time period when liquid crystal molecules reach a steady state, and then the backlight source corresponding to each color field is driven to emit backlight of a second duration within a time period when the liquid crystal molecules reach the steady state. The white light of the second duration is emitted after the first duration, so that the white light is added in the backlight lightening stage of each color field, the color gamut range of the color field corresponding to each sub-frame is effectively reduced, and the color separation phenomenon is reduced. In addition, each subframe corresponds to a single color field or a mixed color field, so that the color separation phenomenon in various algorithms can be improved.

The invention discloses an indoor scene three-dimensional reconstruction method based on deep fusion and confidence modeling, and belongs to the technical field of computer vision. According to the method, composite data frames such as a color image, a depth map, an IMU (Inertial Measurement Unit) and a camera attitude are comprehensively utilized to carry out regional three-dimensional reconstruction on an indoor scene: firstly, the scene is divided into a smooth region (such as a wall, a ground, a ceiling, a glass plane, a mirror surface, a blackboard and other planes) and a complex curved surface region; aiming at the smooth area, adopting geometric prior guide plane fitting provided by a visual large model, and combining sensor attitude information to quickly reconstruct a regular plane model; for a complex curved surface area, a multi-frame point cloud fusion strategy is adopted to accumulate different view angle information, and a deep residual error refining network is utilized to recover curved surface details, so that a high-precision curved surface model is obtained. According to the method, the three-dimensional structure of the indoor scene can be efficiently reconstructed, the global framework of the smooth area is reserved, and the details of the surface of a complex object are depicted in detail.

The invention discloses an object three-dimensional reconstruction method, device and system based on deep learning. The reconstruction method comprises the following steps: acquiring a multi-view color image of an object through a controllable image acquisition device; reconstructing a sparse three-dimensional point cloud by using a motion recovery structure method and obtaining a camera pose; initializing parameters of the three-dimensional Gaussian sputtering model based on the sparse point cloud and performing training optimization; a target object semantic segmentation data set is constructed, and a low-rank adaptive technology is adopted to finely segment all models; generating prompts through an open vocabulary detection model at each view angle, obtaining an accurate segmentation mask, and optimizing a three-dimensional segmentation weight by adopting a joint loss function fusing color consistency loss and edge perception loss; and finally outputting the color three-dimensional point cloud of the target object. According to the method, the original image is segmented, so that the influence of the quality of the rendered image is avoided; the segmentation precision of the model in a specific scene is improved through field adaptive fine tuning; and the accuracy of the segmentation boundary is ensured by adopting a double-loss joint optimization mechanism.

The invention discloses a digital human rendering method based on Gaussian splashing and multi-scale characteristic field distillation, and belongs to the field of three-dimensional human body digital reconstruction. According to the method, feature extraction is carried out through a Vision Transformer encoder, based on an SMPL model, through a cross-modal parameter estimation module and dynamic human body modeling of three-dimensional Gaussian splashing and semantic feature rendering, three-dimensional Gaussian is projected to a two-dimensional image plane to calculate a covariance matrix and color mixing, and after a rendered color image and an initial feature field image are output, a three-dimensional image is obtained. A student feature map is obtained through a convolution acceleration module, a teacher feature map is obtained after feature extraction is carried out through a two-dimensional basic model, the constructed model is trained, and optimization training is completed through comprehensive total loss function calculation; according to the method, the problems of fuzzy semantics, detail missing, low rendering efficiency and inaccurate human body-scene separation in the existing method are effectively solved, so that more efficient, fine and robust monocular or multi-view human body three-dimensional reconstruction is realized.

The invention provides a steel plate surface defect detection method and system based on multi-source data fusion, and relates to the technical field of steel production automatic detection.The method comprises the steps that water stain pretreatment is conducted on the surface of a steel plate, the residual humidity is reduced to be below a preset low humidity threshold value, and a dry surface is obtained; based on the dry surface, collecting multi-source data through a 3D laser line scanning camera and a 2D industrial line scanning camera which are synchronously triggered, obtaining line scanning laser point cloud data, a line scanning reflectivity gray level image, a line scanning depth gray level image and an area array reflectivity color image, and establishing a space corresponding relation among the multi-source data; and forming a multi-source data set based on the spatial correspondence among the multi-source data. According to the method, a high-precision and high-robustness steel plate surface defect detection system covering the whole detection process is constructed by eliminating water stain interference, unifying a multi-source data reference, optimizing data quality, accurately identifying defects and realizing automatic feedback and data closed loop.

The invention discloses a motor fault diagnosis method and system based on color image fusion symmetric point mode, and the method comprises the steps: converting a vibration signal and an electromagnetic signal of a motor into symmetric point mode images, and generating a color signal image fusing feature information; respectively abstracting the color signal images fused with the feature information into nodes and edges in a high-dimensional semantic space so as to construct graph structure data; and performing diagnosis classification on the graph structure data of the vibration signals and the electromagnetic signals by using respective capsule graph network models, and fusing diagnosis classification results of the vibration signals and the electromagnetic signals through a voting mechanism to obtain a final diagnosis classification result. According to the method, multi-channel time domain signals are converted into image expressions with dense information and consistent geometry, and unified feature modeling is carried out on the images based on a depth map structure network with topology perception capability, so that motor fault diagnosis with high diagnosis precision and strong robustness is realized.

The invention provides a focusing method, device and equipment. The method comprises the following steps: determining pixel offset between a red channel image and a green channel image of a tissue slice; the red channel image and the green channel image are a red channel image and a green channel image of a tissue slice in a ghosting image captured by the color image sensor when a red light filter, a green light filter and double optical wedges in the focusing module are modulated at the same time; determining an out-of-focus distance according to the pixel offset and a preset corresponding relation; the preset corresponding relation is the corresponding relation between the pixel offset and the defocus distance; and focusing according to the out-of-focus distance. According to the method provided by the embodiment of the invention, the problem that positive and negative defocusing directions cannot be distinguished through a single-frame image in a phase shift detection automatic focusing technology based on image monochromatic double-hole modulation is solved, and the problem of time waste caused by high-speed switching of an illumination light source in an automatic focusing technology based on red-green double-LED illumination is solved; and meanwhile, the problem that a phase shift detection automatic focusing technology based on pupil segmentation images is relatively small in focusing range is solved.

The invention discloses an intelligent robot navigation system with a body based on a multi-modal large model, and the system comprises an RGB-D camera module which captures a color image and a depth image of an environment in real time, and carries out the preprocessing of the color image and the depth image, and outputs the preprocessed image; the laser radar SLAM module provides a basis for subsequent coordinate calculation and navigation path planning; the multi-modal VLM module is used for calculating the three-dimensional coordinates of the target location in the map; and the navigation control module controls the robot to move. The beneficial effects of the invention lie in that the system realizes efficient environmental perception and natural language understanding through multi-modal fusion, and can accurately identify a target location, calculate a three-dimensional coordinate and plan a safe path, thereby improving the autonomous navigation precision and reliability of the robot, and being suitable for intelligent movement control in a complex scene.

The invention relates to the technical field of computer vision, and discloses a jade defect intelligent detection method and system based on machine vision and deep learning, and the method comprises the following steps: S1, based on a high-resolution industrial camera and a laser three-dimensional scanner, adopting a multi-mode synchronous collection strategy, and rotating a jade sample through a precise motion control system, a jade surface high-resolution two-dimensional color image and high-precision three-dimensional point cloud data are respectively obtained, and a jade multi-mode original data set is generated. A high-resolution two-dimensional color image and high-precision three-dimensional point cloud data are integrated through a multi-modal synchronous acquisition strategy, and multi-dimensional feature expression under unified coordinates is constructed, so that the limitation of a single data source is effectively overcome; an image registration algorithm and a feature pyramid network are combined with a point cloud network to perform multi-modal feature fusion, complementarity of color texture and geometric morphology information is enhanced, and image quality is optimized based on adaptive histogram equalization and non-local mean filtering.

The invention discloses an SAR image orientation sensitive target detection method based on sub-aperture image saturation characteristics, and belongs to the technical field of synthetic aperture radar image target detection. The azimuth sensitive target detection is realized through the following steps: 1) sub-aperture image generation: dividing an azimuth frequency spectrum into a plurality of sub-bands, and generating a plurality of sub-aperture images through inverse Fourier transform; 2) color synthesis and feature extraction: allocating different hues to each sub-aperture image by using an HSV color space, synthesizing an RGB color image, converting the RGB color image to the HSV color space, and extracting a saturation channel in the RGB color space as an azimuth sensitivity feature map; and 3) target detection: carrying out threshold segmentation on the saturation feature map, and identifying a pixel region with a high saturation value, namely, an orientation sensitive target. According to the method, the color saturation change caused by the scattering difference of the target in different sub-apertures is utilized, effective detection of azimuth sensitive targets such as artificial buildings and vehicles is achieved, and the method has the advantages of being simple in calculation and high in robustness.

The invention relates to the technical field of image processing, and discloses a pitaya fruit pose estimation method based on point cloud three-dimensional reconstruction. The pitaya fruit pose estimation method comprises the steps that RGB color images and depth images of pitaya fruits are collected, expanded and marked; the DeepLabV3 + network model is improved and trained, and then an RGB color image is segmented; segmenting the depth image by using the semantic segmentation mask image, then matching the depth image with the RGB color image to obtain a pitaya tree point cloud, and preprocessing the pitaya tree point cloud; registering the pitaya fruit tree point cloud by using an improved point cloud registration algorithm to obtain a fruit tree three-dimensional point cloud model, and segmenting fruit local point clouds from the fruit tree three-dimensional point cloud model; establishing a pitaya fruit point cloud coordinate system by using a PCA principal component analysis method; and performing spherical fitting and ellipsoid fitting on the pitaya fruit point cloud coordinate system to obtain an ellipsoid model with unique pose parameters. According to the invention, a fruit tree three-dimensional point cloud model and accurate fruit three-dimensional space pose information can be provided for a pitaya fruit picking system.

An image processor for a medical observation device includes a color conversion function, The image processor is configured to retrieve an input pixel of a digital input color image and a location of the input pixel in the input color image, and apply the color conversion function to the input pixel to generate an output pixel in a digital output color image. The color conversion function depends on the location of the input pixel.

The invention provides an array image demosaicing method based on dynamic convolution and adaptive coding, which relates to the technical field of image processing, and comprises the following steps: acquiring single-channel original image data and a corresponding color filtering array arrangement type identifier; converting the arrangement type identifier into a multi-dimensional physical feature vector, and inputting the multi-dimensional physical feature vector into a feature processor of a neural network model to generate a weighting coefficient vector; carrying out weighted combination on the plurality of special arrangement transformation matrixes through a weighting coefficient vector to obtain a transformation component, adding the transformation component and a basic convolution kernel parameter to obtain a dynamic convolution kernel parameter, and carrying out directional modulation on a specific spatial position of the dynamic convolution kernel parameter based on a direction weight component in a multi-dimensional physical feature vector; and performing convolution operation on the original image data by using the dynamic convolution kernel parameters, extracting multi-scale features, reconstructing image features, and outputting multi-channel color image data, so that the method can be adaptive to different color filter array types, and the demosaicing precision and generalization capability are improved.

The invention discloses an electronic component packaging defect detection method and system based on image acquisition, and relates to the field of image analysis, and the method comprises the steps: collecting a multi-mode image of a to-be-detected electronic component package; pixel alignment is carried out on the multi-modal image, the multi-modal image after pixel alignment is used as an R channel, a G channel and a B channel to be stacked, and a three-channel pseudo-color image is generated; inputting the three-channel pseudo-color image into a pre-trained convolutional auto-encoder model to obtain a reconstructed three-channel image; analyzing the difference between the three-channel pseudo-color image and the reconstructed three-channel image to obtain an analysis result, and generating a three-channel residual image according to the analysis result; generating a single-channel defect saliency map based on the three-channel residual image; and when a connected region of which the pixel value exceeds a preset pixel threshold value exists in the single-channel defect saliency map, judging that the connected region is a real defect. The detection false alarm rate can be effectively reduced, and the production efficiency is improved.

The invention discloses a landslide susceptibility evaluation method based on a time sequence InSAR, and the method comprises the steps: obtaining a surface deformation rate through the inversion of an SAR image time sequence of a target monitoring region, and rendering the surface deformation rate into an RGB color image; inputting the color image into a trained landslide hidden danger identification model to identify a landslide hidden danger area; if the landslide hidden danger area is detected, continuously carrying out deformation monitoring on the area; for each grid of the target monitoring area, calculating the total information amount formed by the multi-source influence factors; wherein the multi-source influence factor comprises a deformation rate and a plurality of static influence factors; and inputting the total information amount of the multi-source influence factors of each grid of the target monitoring area into a landslide susceptibility evaluation model to obtain a susceptibility grade corresponding to each grid of the target monitoring area. According to the method, the identification efficiency of the landslide hidden danger and the accuracy of landslide susceptibility evaluation can be remarkably improved, and the method is suitable for monitoring and evaluation of geological disasters such as mountain landslide, urban settlement and mining area surface deformation.

The present inventive concept relates to a method and a device for training a machine learning model to construct a digital color image depicting a sample. The method comprising: acquiring a training set of digital images of a training sample by: illuminating, by a plurality of white light emitting diodes, the training sample with a plurality of illumination patterns, and capturing, for each illumination pattern of the plurality of illumination patterns, a digital image of the training sample; receiving a ground truth comprising a high-resolution digital color image of the training sample, wherein a resolution of the high-resolution digital color image is relatively higher than a resolution of at least one digital image of the training set of digital images; and training the machine learning model to construct the digital color image depicting a sample using the training set of digital images and the ground truth. The present inventive concept further relates to a microscope system and a method for constructing a digital color image depicting a sample.

The invention provides a three-dimensional digital human reconstruction method, and the method comprises the steps: estimating a parameterized human body model corresponding to an input single human body reference image based on the input single human body reference image, and obtaining a multi-view human body semantic segmentation image of the parameterized human body model according to a plurality of preset camera poses and rendering parameters; taking the multi-view human body semantic segmentation image as a condition signal, taking the single human body reference image as input, and generating a multi-view human body color image based on a pre-trained first video diffusion model; taking the multi-view human body color image as a condition signal, taking a human body normal image extracted from a single human body reference image as input, and generating a multi-view human body normal image based on a pre-trained second video diffusion model; and performing 3D Gaussian splashing based on the multi-view human body color image and the multi-view human body normal image so as to reconstruct a three-dimensional digital human body corresponding to the single human body reference image.

The invention discloses an anomaly detection method and system for a plate splicing machine based on visual intelligence, and relates to the technical field of anomaly detection of the plate splicing machine, and the method comprises the steps: collecting a color image sequence and a structured light graph, synchronously extracting a glue spraying control instruction, calculating a disturbance difference graph between adjacent frames, and carrying out the three-layer discrete wavelet decomposition of the disturbance graph, a frequency domain disturbance feature tensor is extracted in combination with a channel attention mechanism; using a MobileNetv1-YOLOv4 model to identify an abutted seam area and generate a mask, enabling the mask area to act on the structured light graph to reconstruct a local three-dimensional point cloud, and identifying structural abnormality through PointNet + +; and extracting a joint center path skeleton based on a mask, sampling a disturbance tensor according to a path sliding window to construct a time sequence, and inputting a Bayesian variable point detection model to identify a dynamic abnormal point. According to the invention, the visual intelligent detection capability of the board splicing machine under high-speed and complex working conditions is obviously improved.

The invention relates to the technical field of three-dimensional scene reconstruction methods, in particular to a multi-modal data three-dimensional scene reconstruction method and device, and the method specifically comprises the following steps: 1, carrying out the data collection of a target scene through a plurality of sensors, comprising a color image sequence acquired by an optical camera, point cloud data acquired by a laser radar and a depth image sequence acquired by a depth camera; 2, denoising processing is carried out on the collected color image sequence, and a median filtering algorithm is adopted; 3, rough alignment is carried out on preprocessed multi-modal data by adopting a method based on feature points; according to the method, the alignment precision is remarkably improved, accurate matching of different modal data in space is ensured, the geometric structure of a reconstruction model is accurate, and the relative position of an object fits an actual scene.

The invention relates to the technical field of navigation, in particular to an end-to-end visual language navigation method and device based on occupancy network awareness and computer equipment, and the method comprises the following steps: receiving a language instruction, and obtaining a color image and a depth map collected by an RGB camera and a depth camera; performing feature extraction on the color image and the depth image through a ViT model to generate image features and depth features, and performing rotation and translation transformation on the depth features through a camera model to generate a blank 3D voxel space; mapping the image features to a blank 3D voxel space to generate 3D voxel features, and constructing the image features into a topological graph; inputting the 3D voxel features, the topological graph and the language information into a BERT model to obtain text features; and inputting the text features, the 3D voxel features and the topological graph into a cross-modal model, and predicting a navigation path. The method and the device are helpful for improving the accuracy of navigation decision.

The invention relates to the technical field of industrial automation, in particular to an industrial robot posture recognition method. Aiming at the problems of weak workpiece texture, strong surface reflection, serious stacking and shielding and the like in an industrial production line, a pixel-level dense feature fusion and self-attention mechanism is introduced, and semantic texture information of a color image and spatial geometric information of a depth image are integrated in a feature extraction stage; the problem that the posture of a rotationally symmetrical workpiece is fuzzy is solved through an asymmetric loss function, meanwhile, a comprehensive grabbing scoring model containing force sealing stability, environment collision risk probability, mechanical arm kinematics reachability and visual uncertainty is further established, and through the mode of combining off-line grabbing candidate construction and on-line real-time evaluation, the grabbing accuracy of the mechanical arm is improved. And the globally optimal grabbing pose is screened out. According to the method, the recognition precision and robustness of the industrial robot in the unstructured environment are remarkably improved, and the success rate and safety of grabbing operation are improved.