131 results about "Image fusion algorithm" patented technology

The invention discloses an infrared and visible light image fusion method based on a conditional diffusion model, and belongs to the technical field of image fusion. Compared with a current popular infrared and visible light image fusion algorithm, the algorithm is obviously improved in the aspects of color fidelity, texture detail maintenance and fusion quality, and feature extraction and fusion tasks are completed through combination of a conditional denoising diffusion implicit model and a multi-channel high and low frequency parallel fusion model; the problem that the denoising diffusion implicit model lacks a reference true value of a fusion result when executing an image fusion task is relieved, the unique advantages of the denoising diffusion implicit model in image processing are further shown, and the method is excellent in the aspects of high-quality image generation and mathematical interpretability. The optimization of the existing method and the upgrading of the existing technology are realized.

The invention provides a power transmission and transformation project ecological influence space-time boundary determination method and system based on multi-source remote sensing information deep fusion interpretation, and belongs to the technical field of power transmission and transformation project management. The method comprises the steps of data acquisition, data preprocessing and multi-source data integration, multi-scale space-time analysis, ecological influence index construction, remote sensing ecological index model construction, ecological influence space-time boundary determination, dynamic monitoring and early warning system platform establishment and long-term monitoring and evaluation. By integrating satellite remote sensing, unmanned aerial vehicle aerial photography and ground monitoring data and combining an image fusion algorithm, a principal component analysis method and a random forest algorithm technology, the method is beneficial to promoting accurate evaluation of the influence of the power transmission and transformation project on the ecological environment, visualizes the space-time boundary of the ecological influence, breaks through the limitation of a traditional evaluation method on time and space coverage, and improves the evaluation efficiency of the power transmission and transformation project. And a scientific method system for analyzing and evaluating the ecological influence of the power transmission and transformation project is further improved.

The invention relates to the technical field of photovoltaic power station operation and maintenance, and discloses a photovoltaic power station intelligent operation and maintenance and fault early warning system. A multi-band image fusion algorithm and a temperature gradient enhancement algorithm are applied to identify composite defects; establishing a defect interaction influence model to analyze the synergistic effect of different defects; constructing a defect evolution prediction algorithm to predict a defect development trend; and developing a defect-performance mapping model and a processing priority algorithm to optimize a maintenance decision. According to the method, the defect detection sensitivity and accuracy are improved, accurate identification and evolution prediction of composite defects are realized, a prospective maintenance decision is effectively supported, the operation and maintenance efficiency of a photovoltaic power station is remarkably improved, the service life of the power station is prolonged, and the safety risk is reduced.

The invention relates to the field of image comparison, and discloses a multi-layer PCB alignment deviation detection system and method based on image comparison, and the method comprises the steps: obtaining high-resolution image data before and after lamination of a multi-layer PCB, carrying out the unified registration preprocessing of an original image through the combination of a multi-mode image fusion algorithm and a geometric distortion correction technology, and obtaining a multi-layer PCB alignment deviation detection result; constructing a standardized image registration input set; key alignment feature extraction is carried out on the image registration input set, and a multilayer structure graph model is constructed based on a graph neural network; in combination with the alignment reference map, dynamically adjusting an image comparison window and a search region by adopting a region attention mechanism and a local adaptive matching algorithm, and constructing an alignment deviation mapping map; constructing a deviation evolution model by using a time sequence behavior recognition network according to the constructed alignment deviation mapping graph and historical process deviation data; and performing comprehensive evaluation on the image registration input set based on the deviation evolution model and the generated early warning information. The method has the advantage of improving the accurate detection level.

The invention discloses a multi-target and long-distance field domain temperature measurement method based on dual-spectrum imaging. A visible light camera module and an infrared camera module are included. Obtaining internal parameters and external parameters in the visible light camera module and the infrared camera module, and optimizing the internal parameters and the external parameters; performing quick image feature search and image feature stereo matching on the acquired visible light image and infrared image to realize identification and positioning during image measurement; registering and fusing the visible light image and the infrared image by adopting a double-spectrum image fusion algorithm to obtain a fused image; carrying out image intersection measurement positioning and temperature measurement area stereo matching on the image; performing systematic edge calculation on the image by adopting a data-based intelligent analysis algorithm, and analyzing a temperature change trend according to the obtained image data to obtain the temperature of the measured object; the current running state of the detected power equipment is evaluated, the equipment fault is diagnosed, and the application end is used for storing, displaying and processing the detection result; and the temperature detection accuracy of the power equipment is effectively improved.

The invention discloses an infrared and visible light image fusion method, electronic equipment and a storage medium, and belongs to the technical field of image processing. The method comprises the following steps: acquiring a weak visible light image and an infrared image corresponding to the weak visible light image; an image fusion network is constructed, the image fusion network is supervised by using a supervised label generated by a pre-trained brightness adjustment network, and the quality of a fused image is constrained by using a pre-trained cue word generation network; and inputting the weak visible light image and the infrared image corresponding to the weak visible light image into the image fusion network for fusion to obtain a fused image. According to the technical scheme, the problem of daytime overexposure existing in a traditional night image fusion algorithm is effectively restrained, and meanwhile the problem of night point light source overexposure is effectively restrained.

The invention discloses a semiconductor packaging defect detection system based on machine vision, and relates to the technical field of semiconductor packaging defect detection, and the system comprises a multispectral collection module which uses a multispectral camera and a light source to collect multiband images of a packaging surface from different angles, and carries out the preprocessing of the multiband images; the image fusion module is used for fusing the preprocessed wave band images into a comprehensive image by adopting an image fusion algorithm; the illumination adjusting module is used for carrying out illumination quality analysis on the comprehensive image, adjusting light source parameters based on a quality analysis result and re-collecting a multi-band image; the method has the beneficial effects that through the comprehensiveness of a multispectral image, the detail enhancement capability of image fusion, the stability of dynamic illumination adjustment, the feature extraction capability of deep learning and the anomaly detection capability of an auto-encoder, the robustness of the image fusion is improved; the defects of unstable image quality, insufficient feature extraction capability, inaccurate error judgment and the like in the prior art are comprehensively overcome.

The invention discloses a territorial space dynamic remote sensing monitoring system and method based on spectral analysis, and relates to the technical field of remote sensing monitoring. A characteristic wave band extraction module; the image fusion and ground object monitoring module is used for fusing the feature data of each wave band by using an image fusion algorithm, identifying the fused ground object type and monitoring the change of the ground object type along with time; and a change monitoring and visualization module. According to the method, the remote sensing images of different spectrum sections are acquired and corrected, so that the accuracy and consistency of data are remarkably improved, the accuracy of a monitoring result is ensured, and a solid data basis is provided for subsequent analysis.

The invention relates to the technical field of remote sensing image processing, in particular to a remote sensing image strip denoising method and system based on a generative adversarial network, and provides the following scheme: collecting remote sensing image data; constructing an image detection model, extracting spatial distribution characteristics of the stripe noise through the first processing network to generate an abnormal node marking graph, and extracting time dynamic characteristics of the stripe noise through the second processing network to generate a time characteristic mapping graph; a generative adversarial network is constructed, a generator performs joint optimization according to space and time characteristics to generate a de-noised image, and a discriminator performs authenticity verification on the de-noised image and feeds back parameters of the optimization generator; and finally, through an image fusion algorithm, carrying out fusion processing on the de-noised image and the original noiseless area to generate an optimized remote sensing image. The method can efficiently remove wide strip, thin strip and dynamic strip noise, retains image details and spectral information, and is suitable for multi-scene remote sensing image processing.

The invention discloses a tumor ablation puncture operation planning method and system, and relates to the technical field of medical image processing, and the method comprises the steps: obtaining multi-modal image data of a tumor region, carrying out the preprocessing, carrying out the fusion processing of the preprocessed multi-modal image data through an image fusion algorithm, generating a fusion image, and carrying out the planning of a tumor ablation puncture operation based on the fusion image. A tumor area is automatically marked by using an image segmentation technology, a doctor manually corrects a segmentation result, an optimal puncture path is selected through design and simulation according to a fusion image with a tumor mark, a model and a tumor entity position are accurately aligned by using a space positioning technology based on the fusion image and augmented reality equipment, and an AR view is generated. And under the guidance of the AR view, the puncture needle carries out ablation operation according to the optimal puncture path. Through high-quality images and three-dimensional guidance, high precision and visualization of the whole process from image acquisition to operation implementation are achieved, and the safety and success rate of the operation are remarkably improved.

The invention relates to a vehicle in-the-loop test method based on visual augmented reality. Firstly, a positioning and motion information acquisition system of a real vehicle is established, and high-precision motion information of the vehicle can be obtained; then image information and motion information of a vehicle-mounted camera are synchronously collected, and the purpose is to obtain camera output and vehicle motion information under the same timestamp; secondly, calculating the pose of the camera at each moment according to the internal participation initial external parameter information of the camera and the synchronously acquired motion information; synchronously inputting the calculated pose into simulation software, and obtaining virtual camera output consistent with the real camera in time and space; taking a really acquired camera image as a background, taking a traffic participant provided by a virtual simulation camera as a foreground, and obtaining fused image information of an augmented reality view angle through an image fusion algorithm; and finally, the fused image is input to the sensing layer, a sensing result is settled and then input to the decision control layer, a decision instruction is output to an execution mechanism of the vehicle, the vehicle motion information is changed and then is acquired by the high-precision information acquisition system, and thus a round of information circulation is completed.

The invention relates to the technical field of data processing, in particular to a digital twin scene efficient rendering method based on dynamic and static combination and a vertex shader, which comprises the steps of early-stage preparation and scene analysis, three-dimensional modeling and material mapping, vertex shader design and optimization, dynamic and static combination rendering process design, post-processing and optimization, and deployment and integration. The method has the beneficial effects that the static scene is quickly rendered by using the pre-rendered cache data, the dynamic element is dynamically rendered by combining the real-time data source and the vertex shader, and the dynamic element and the real-time data source are seamlessly combined by using the image fusion algorithm, so that the rendering speed of the digital twinning scene is remarkably improved, and the rendering efficiency of the digital twinning scene is improved. The real-time performance and interactivity of dynamic elements are ensured, natural fusion of static and dynamic scenes is realized through fine regulation and control of weight factors, a high-quality and vivid digital twinning rendering image is finally presented, and the user experience and the application effect are greatly enhanced.

The invention discloses a robot automatic grabbing path planning method based on visual identification, and relates to the technical field of intelligent grabbing. The method comprises the following steps: acquiring multi-view visual data of a target scene, and generating a scene three-dimensional compact reconstruction model through a multi-modal image fusion algorithm; performing target detection and feature extraction on the model, and screening an optimal capture point in combination with a visual attention mechanism; constructing a dynamic environment obstacle probability map, updating an obstacle state through time sequence visual tracking, and quantifying an interference weight; an initial grabbing path is planned based on an improved fast expansion random tree algorithm, path smoothness constraints and robot joint movement limit parameters are introduced, and path nodes are optimized through a Bezier curve; visual servo feedback and path deviation prediction are fused, path parameters are corrected in real time, and a continuous movement track is generated. The method effectively adapts to the dynamic environment, gives consideration to path safety, smoothness and mechanical arm motion characteristics, and remarkably improves the grabbing success rate and operation reliability.

The invention discloses a rock and ore microscopic image splicing method and system based on deep learning, and relates to the field of image processing and the technical field of microscopes, and the method comprises the steps: obtaining a local rock and ore microscopic image of a rock and ore slice, carrying out the preprocessing of the local rock and ore microscopic image, and carrying out the overlapping region coarse registration of the preprocessed local rock and ore microscopic image through a phase correlation method; based on an improved image feature detection model, basic features and description features in the local rock and ore microscopic image after coarse registration are extracted, and local image features of the local rock and ore microscopic image are obtained; performing feature matching on the local image features of the two groups of local rock and mineral microscopic images by using an image feature matching model to obtain a matching corresponding relation of the local image features; and based on an image fusion algorithm of a homography matrix and a partial differential equation, splicing and optimizing the local rock and ore microscopic images in combination with a matching corresponding relation of local image features, and generating a large-view-field rock and ore microscopic image. According to the method, the complex transformation between the images can be better processed.

The invention belongs to the field of application of an image processing technology and an automatic detection technology, provides a nodular cast iron pipe white sprayed code integrity detection method and system, and aims to solve the problems of low nodular cast iron pipe white sprayed code recognition accuracy, poor environmental adaptability, low processing speed, image splicing and the like in industrial production. In combination with an optimized image preprocessing technology, an advanced feature point detection and matching technology and an image fusion algorithm are applied to solve the problem of image seamless splicing. By adopting the method provided by the invention, the error and time cost in a manual identification process can be greatly reduced through prediction, and the automation and intelligence level of a production line is improved. The production efficiency can be improved, the product quality can be ensured, and the increasing industrial automation requirement is met. Finally, the system becomes an important tool for improving the product quality and the production efficiency of nodular cast iron pipe production enterprises.

The invention discloses a sparse view angle surface reconstruction method based on cross-view angle information complementation. The method comprises a training stage and an inference stage. The training stage comprises the following steps of: firstly, inputting an input sparse view angle image into a network to obtain a multi-scale feature and construct a cost body; inputting the multi-view-angle cost body into a cross-view-angle geometric complementation module to obtain an optimized cost body; inputting the optimized cost body into a cross-view-angle geometry Transform, obtaining an image and a depth map of a new view angle through a body rendering technology, and using a truth value depth to supervise the training of a network; in the inference stage, the multi-view-angle images and the corresponding poses are input into the trained cross-view-angle information complementary sparse view angle surface reconstruction network, the depth of each view angle is obtained, and a final surface is obtained by using a depth image fusion algorithm. According to the invention, through explicitly constructing a cross-view complementary information fusion mode, the scene reconstruction accuracy and robustness under large view change are improved.

The invention provides a flue-cured tobacco leaf image distortion correction enhancement method, and belongs to the technical field of flue-cured tobaccos. Distortion segmentation point detection is carried out on an obtained flue-cured tobacco leaf image, and an illumination change area caused by unevenness of a tobacco leaf surface is identified by using an illumination homogenization algorithm based on a Poisson equation; secondly, constructing a distortion jump point matrix to mark gray abrupt change positions caused by shadows and light reflection between adjacent areas, then decomposing and reconstructing multi-scale image features by adopting a Laplacian pyramid image fusion algorithm, and eliminating optical interference generated by surface fluctuation through a designed fusion rule; and finishing processing when the accuracy reaches a set range, otherwise, carrying out correction and optimization again, thereby solving the technical problem that the image quality and the subsequent analysis accuracy are influenced by shadow and light reflection generated during image acquisition due to unevenness of the tobacco leaves in the prior art.

The invention is suitable for the field of intelligent jet cleaning, and provides a dirt recognition cleaning method, device and equipment, and the dirt recognition cleaning method comprises the steps: obtaining an original image corresponding to target dirt and a sound reflection signal corresponding to the target dirt; converting the sound reflection signal corresponding to the target dirt into an image representation; obtaining a fused image corresponding to the target dirt according to the original image corresponding to the target dirt, the image representation and a preset multi-channel image fusion algorithm; inputting a fusion image corresponding to the target dirt into the improved deep neural network to obtain segmentation mask data and size data of the target dirt; adjusting the pose of the dirt cleaning equipment according to the segmentation mask data and a preset coordinate conversion algorithm; according to the size data of the target dirt, dirt cleaning parameters are adjusted, and dirt cleaning equipment is controlled to execute cleaning operation. According to the method, the dirt recognition and cleaning precision and efficiency are remarkably improved, and various cleaning requirements and operation environments can be better met.

The invention discloses a power line identification method, device and equipment based on image fusion and a medium, and the method comprises the steps: obtaining an infrared image and a visible light image of a power line of a power distribution network, and carrying out the registration of the infrared image and the visible light image through a preset image registration algorithm, and generating a first image; wherein the infrared image and the visible light image are derived from the same picture; performing feature extraction and feature fusion on the first image and the visible light image through a preset image fusion algorithm to generate a second image; and performing power line identification on the second image through a preset power line identification algorithm. According to the method, the infrared image and the visible light image which are derived from the same power line picture are fused to obtain the fused image which contains more source information and has more obvious power line characteristics, so that power line recognition is performed, the difficulty of power line recognition is greatly reduced, and the accuracy of power line recognition of the power distribution network is further improved.

The invention discloses a target detection method and device based on multi-modal fusion, electronic equipment and a medium, and the method comprises the steps: obtaining a first image set and a second image set, the first image set comprises a plurality of visible light images, and the second image set comprises a plurality of infrared images; performing data alignment operation on the first image set and the second image set to obtain a first feature image set corresponding to the first image set and a second feature image set corresponding to the second image set; performing feature fusion on the first feature image set and the second feature image set through a preset dual-scale image fusion algorithm to obtain a fused image set; and performing target detection on the fused image in the fused image set through a preset deep learning model to obtain the object category of the to-be-detected object and mark the bounding box of the to-be-detected object. In the embodiment of the invention, the accuracy of target detection can be improved, and the detection precision is further improved.

The invention relates to the technical field of liquid crystal display screen detection, in particular to a liquid crystal display screen defect intelligent detection method and analysis system based on visual perception, and the method comprises the following steps: obtaining multi-angle image data of a liquid crystal display screen; generating a fused image based on the multi-angle image data; extracting defect features according to the fused image; determining a defect category based on the defect features; according to the method, multi-angle image acquisition and an innovative image fusion algorithm are adopted, so that the visibility of the defects and the comprehensiveness of detection are greatly improved. According to the method, different types of defects can be effectively captured, especially small defects which are difficult to observe at a single angle, and the omission ratio is remarkably reduced. By fusing multi-angle information, a more comprehensive and clear defect image can be generated, and high-quality input data is provided for subsequent analysis.

The invention provides a welding spot quality detection method and system, a binocular camera is combined with an image fusion algorithm to form a three-dimensional image of a welding spot, welding spot feature information is obtained through calculation and is automatically compared with technical standard requirements, the quality of the welding spot is judged, and cause feedback of disqualification is given. The method solves the problem of large manual misjudgment caused by the fact that the welding spot cannot be shot by 360 degrees in all directions in the re-inspection operation after AOI misjudgment, the understanding of the geometric shape of the welding spot can be improved by using an image fusion algorithm, and potential defects can be highlighted. And meanwhile, the welding spot characteristics are automatically compared, so that the detection speed, the detection efficiency and the detection accuracy are improved, and subjective errors possibly caused by manual detection are reduced. In addition, due to the reason feedback function, an inspector and the quality control process are greatly facilitated, the inspector is helped to rapidly position problems and take correction measures, and automation and accuracy of welding spot detection are improved.

The invention provides a power transmission line bird damage risk assessment and early warning method and system, and relates to the technical field of intelligent power grid environment perception and risk early warning, and the method comprises the steps: data collection and multi-modal fusion: collecting image data, voiceprint data, meteorological parameters and ecological environment information along a power transmission line to form multi-modal data, the multi-modal data is uniformly coded by synchronizing timestamps and geographic coordinates, an original bird activity sample set is generated, and for the problems that the monitoring means is single and the recognition precision is insufficient, image, voiceprint, weather and ecological information are fused through a multi-modal data acquisition module, so that the recognition precision is improved. The acousto-optic image fusion algorithm and the two-channel convolutional neural network are adopted, all-weather bird type and behavior accurate recognition is achieved, the recognition accuracy is improved, key behaviors such as nesting and dung pulling can be stably captured even at night or in a foggy environment, and potential risk missing detection is remarkably reduced.

The invention relates to a medical image tuning method, and belongs to the technical field of image processing. According to the invention, a medical image tuning device is adopted; a core board fusion algorithm module automatically identifies the number and pixels of connected cameras, automatically matches a built-in image processing algorithm module, and performs preprocessing through an exposure algorithm module, a denoising algorithm module and a contrast enhancement algorithm module to obtain a single or multiple images as a state 1; extracting image features of the state 1, automatically selecting a single-image algorithm module or a multi-image fusion algorithm module, and obtaining a state 2 through the fusion algorithm module; low-delay and high-pixel transmission is met, image information is completely reserved, a camera host can better complete image processing, and the final image quality is guaranteed; the problem of image information loss caused by image processing at the camera module end is reduced, and the problems of module temperature rise, component damage, influence on imaging quality and the like caused by image processing at the camera module end are solved.

The invention relates to the technical field of image fusion, and provides an infrared and visible light image fusion algorithm based on a hybrid expert system, and the infrared and visible light image fusion algorithm comprises the following steps: S1, carrying out the cross feature extraction of an input infrared image and an input visible light image; s2, inputting a cross feature extraction result into a gating module composed of a full connection layer; S3, selecting two environment categories with the maximum probability value, and executing differential fusion processing in an expert model; S31, splitting an infrared image and a visible light image into local windows, then through a wavelet adaptive transformation module, performing multi-band decomposition on each window and applying an extra weight coefficient to generate a preliminary fusion image; s32, performing edge feature extraction, spot feature extraction and feature fusion processing on the preliminary fusion image to obtain a final fusion image; and S4, performing weighted fusion on the final fusion image through the trainable weight coefficient to obtain a final output image.

The invention provides a transformer fire detection method based on infrared and visible light image fusion, and the system comprises an infrared camera, a visible light camera, an image fusion and analysis host, a large monitoring screen, and an alarm device. The use method of the detection system comprises the steps that infrared images and visible light images captured in real time are transferred into the image fusion and analysis host through the infrared camera and the visible light camera, and after the image fusion and analysis host receives the image flow, the infrared images and the visible light images are fused through an image fusion algorithm; a multi-modal image is formed, then a pre-trained deep learning model is utilized to analyze the fused image, an early occurrence sign of a fire disaster is identified, a fire hazard part is found in time, after identification, the fire hazard part is presented to a large monitoring screen, and meanwhile, a sound-light alarm device is triggered so as to timely remind related operators on duty to rapidly take measures. Early-stage fire disasters are extinguished, and operation safety of the transformer is guaranteed.

The invention discloses an image processing method based on an adaptive weight multi-image fusion algorithm, which relates to the technical field of image processing and comprises the following specific steps of: extracting a plurality of images; constructing an adaptive weight multi-image fusion algorithm to carry out fusion processing to obtain a fused image; the adaptive weight multi-image fusion algorithm specifically comprises the following steps: inputting an image sequence; convolution processing is carried out; performing position alignment on the image; screening overexposure and underexposure areas of the source image for shielding to obtain a convolution result image; carrying out merging processing to obtain a convolution image; performing normalization processing on the convolution image; and calculating the distance standard deviation from each pixel of the source image to the average gray value, and fusing a new image. According to the method, multiple images can be fused into one image with a soft surface and a luxurious edge, detail information in the multiple images is reserved, the noise of the images is reduced, and the local contrast of different areas in the images is enhanced, so that the quality of the images is improved.

The invention provides a soil humidity prediction method and device, electronic equipment and a storage medium, and relates to the technical field of soil humidity prediction.The method comprises the steps that current meteorological data and predicted meteorological data are obtained, the predicted meteorological data comprise meteorological data of N days in the future, and N is larger than or equal to 1; acquiring Landsat remote sensing image data and MODIS remote sensing image data; based on a GEE image fusion algorithm, estimating the Landsat remote sensing image data through the MODIS remote sensing image data to obtain fused remote sensing image data; fusing remote sensing image data to obtain a fused spectral index; and inputting the current meteorological data, the predicted meteorological data and the fused spectral index into a hierarchical soil humidity model to obtain a predicted soil humidity data set. According to the invention, the soil humidity prediction precision is improved.

The invention relates to the technical field of image fusion, in particular to an image processing system based on tumor imageomics, which improves an existing medical image fusion algorithm based on multi-scale characteristic frequency domain decomposition filtering, and after a fusion layering with a better representation effect on graphic information and focus information after first fusion is determined, the fusion layering is carried out on the image information and the focus information; and according to the image feature expression of the fusion layer and the focus feature expression, adaptively determining the fusion weight of image fusion of different feature image layers, thereby performing adaptive image fusion based on the adaptive fusion weight. The problem that a medical image fusion algorithm based on multi-scale characteristic frequency domain decomposition filtering is poor in universality and flexibility is solved, and the universality and flexibility of the obtained final fusion image are better and more accurate.

The invention relates to the technical field of aircraft detection, in particular to a cross-modal aircraft fuselage surface fault detection method based on multi-task learning, which comprises the following steps: collecting a visible light image and an infrared image of an aircraft fuselage surface and carrying out image preprocessing; feature extraction is carried out for target identification, and infrared and visible light image fusion is realized; carrying out fault detection on the extracted features by adopting a multi-task learning framework; parameters of the model are continuously adjusted, the trained model is applied to actual aircraft fuselage surface fault detection, and when a new image is input, the model can quickly provide accurate decision support; visible light and infrared images are creatively fused, and an efficient and accurate aircraft fuselage surface fault detection model is constructed by using a multi-task learning strategy; an image fusion algorithm, a feature extraction method and a task division and cooperation mode in a multi-task learning framework are key technologies.