91 results about "Inverse projection" patented technology

The invention relates to the technical field of oceanographic engineering surveying and mapping, and provides an oceanographic engineering surveying and mapping deep sea terrain reconstruction method, which comprises the steps of multi-modal data synchronous acquisition, frequency domain feature cross enhancement, dynamic convolution feature fusion, inverse projection terrain reconstruction and terrain credibility verification. According to the invention, an acousto-optic multi-mode fusion technology is adopted, and the advantages of acoustic penetrability and optical high resolution are combined, so that the complex terrain detail capturing capability is improved; the dynamic convolution network adaptively adjusts convolution kernel parameters through frequency domain energy, the multi-scale feature fusion effect is enhanced, and the accuracy of terrain texture expression is improved; according to the motion distortion compensation method, carrier attitude drift is corrected in real time through inertial navigation data, and the geometric fidelity of a reconstruction model is guaranteed. The method is suitable for a deep sea complex environment, and provides high-precision terrain reconstruction method support for ocean engineering surveying and mapping and ocean resource exploration.

The embodiment of the invention discloses a high-precision three-dimensional surface reconstruction method and device based on polarization information enhancement. A specific embodiment of the method comprises the following steps: acquiring multi-view polarization images by using a polarization camera to obtain a polarization image set; performing structured light initialization processing on the target image to generate initialized target image information; inversely projecting the accurate depth information set into a three-dimensional space to obtain enhanced Gaussian point cloud data; gaussian point cloud de-reflection processing is carried out on the initialized target image information to generate de-reflection Gaussian point cloud data; generating Gaussian point cloud data based on the enhanced Gaussian point cloud data and the de-reflected Gaussian point cloud data; and performing data optimization and reconstruction processing on the Gaussian point cloud data to generate a three-dimensional triangular patch. According to the embodiment, the precision and efficiency of three-dimensional surface reconstruction can be improved.

The invention discloses a micro-sensing trapezoidal method, device and equipment based on Aruco recognition and a medium, and relates to the field of computer vision. The method comprises the following steps: S1, projecting an Aruco grid and collecting an image, and extracting marked angular points to construct a position set; s2, calculating a trapezoid tension factor based on the trapezoid area and the opposite side distance, and quantifying local deformation; s3, fusing multi-labeled spatial parallax and tension information, constructing a continuous tension equilibrium domain, and generating a final mapping position through gradient correction basic inverse projection mapping and boundary disturbance compensation; and S4, outputting an inverse perspective matrix based on a fusion result, and estimating a microscopic pose offset parameter for fine adjustment of the projector. And through tension factor and spatial parallax modeling, adaptive correction of a distortion region is realized, the edge consistency and overall symmetry of the image are improved, and the stability and precision of the projected image are enhanced.

The invention discloses a steel bar spacing measurement method and system based on multi-view vision, and relates to the field of distance measurement, in the method, based on remapping parameters, a stereo correction module carries out stereo correction on left and right images to obtain corrected left and right images; based on the global features of the corrected left and right images, a reinforcement depth information calculation module calculates a feature matching constraint matrix; based on the corrected local feature points of the left and right images, a reinforcing steel bar depth information calculation module determines matched feature point pairs; based on the matched feature points, a steel bar depth information calculation module calculates a parallax value and calculates depth information of a steel bar intersection point according to the parallax value and a preset camera calibration parameter; based on the corrected left and right images, a reinforcing steel bar intersection point three-dimensional coordinate reconstruction module detects and reconstructs pixel coordinates of reinforcing steel bar intersection points; and based on the depth information and the pixel coordinates, the physical spacing calculation module calculates the spacing of the reinforcing steel bars according to an inverse projection formula. The method is used for improving the accuracy of steel bar spacing measurement.

The present invention discloses a multi-source heterogeneous large-scene remote sensing image matching method for global optimal solution, including: S1, preprocessing the input image information; S2, extracting the reference image point group features by constructing a sub-pixel point group feature detection algorithm; S3, constructing a multi-mode matching algorithm based on wavelet analysis for sub-pixel matching; S4, constructing a point group global registration algorithm for forward and inverse projection, hierarchical elimination, and obtaining the global optimal solution; S5, judging whether the sub-pixel requirement is met. If it is judged that the requirement is not met, return to step S4, otherwise output the result. According to the present invention, a high-precision sub-pixel is obtained, the consistency of image intensity and phase is ensured, an accurate matching result is obtained, and an algorithm system with good compatibility is provided.

The invention discloses an edge SLAM dense map construction method based on Kirsch combined high and low thresholds. The method comprises the following steps: inputting an RGB-D picture into an improved visual odometer, extracting Kirsch edge features of a color image, performing instance segmentation on the color image of a key frame by using a BlendMASK deep neural network, and fusing semantic information into the color image. And distance transformation is constructed according to the binary image of the Kirsch edge and is used for calculating a residual error of 3D-2D registration of the key frame and the current frame. And then, carrying out pose estimation on the current frame by using 3D-2D registration of the key frame and the current frame segmented by the fusion instance. A sliding window local BA optimization algorithm based on a foreground edge is used for carrying out local optimization on the pose, and on this basis, a closed-loop detection algorithm based on a random fern descriptor is used for carrying out global pose graph optimization. And finally, based on the camera projection model and the inverse projection function, constructing a dense point cloud map. According to the method, an edge detection algorithm combining Kirsch with high and low thresholds, a BlendMASK deep neural network, a sliding window local BA optimization algorithm based on foreground edges and a closed-loop detection algorithm based on random fern descriptors are combined, so that the mapping accuracy can be improved, and a remarkable effect is achieved in the aspect of realizing an accurate dense point cloud map.

The present invention discloses a nonlinear calibration method and device for a line structured light 3D camera laser surface, including: performing reverse projection according to the camera's intrinsic parameter matrix, lens distortion parameters and extrinsic parameter matrix to obtain the three-dimensional coordinates of the center point of the laser line in the two-dimensional image captured by the line structured light 3D camera in the camera coordinate system; fitting the surface using a trained neural network model according to the three-dimensional coordinates of the center point of the laser line to obtain the surface model of the line structured light 3D camera laser surface in the camera coordinate system; obtaining the ray corresponding to each pixel according to the three-dimensional coordinates of each pixel on the center contour of the laser line in the two-dimensional image captured by the line structured light 3D camera; calculating the intersection of the ray corresponding to each pixel and the surface model of the line structured light 3D camera laser surface to complete the three-dimensional reconstruction of the line structured light 3D camera laser surface. The present invention can effectively improve the accuracy of laser surface calibration and the accuracy of three-dimensional reconstruction results.

The invention discloses an external parameter calibration method of a vehicle-mounted around-view camera and a storage medium, and the method comprises the steps: projecting an image point on an imaging plane of a camera to the ground through employing a reverse projection process, and calculating a reverse projection error between two adjacent cameras through employing a reverse projection point obtained in the reverse projection process; an objective function is constructed based on the inverse projection error, and finally, the objective function is utilized to adjust the external parameters of the vehicle-mounted around-view camera to complete calibration. The world coordinate of each image point does not need to be calculated in the constructed target function, so that the calculation amount in the calibration process is greatly reduced, and the camera calibration efficiency is improved; and when the inverse projection process is utilized, because the image points are uniformly projected to the ground, the influence of the camera height arrangement can be avoided, the problem that the calibration effect is poor due to the fact that the calculation of the re-projection error in the classical BA algorithm is influenced by the camera arrangement height is solved, and the calibration effect of the vehicle-mounted around-view camera can be effectively improved.

The invention relates to a large language model long context cache compression method, system and device and a medium, and the method comprises the steps: obtaining a long text data set as verification data, inputting a large language model, calculating key value cache, executing Fourier projection and expansion operation, and calculating data reconstruction loss of each feature dimension, identifying key value dimensions insensitive to long text context information; in a text data pre-filling reasoning stage, finite-order Fourier projection compression is performed on key value cache data in a GPU memory, and marked data with insensitive dimensions are compressed from an original storage mode of linearly increasing along with the text length to a fixed-length frequency spectrum for representation; in the decoding reasoning stage, inverse projection expansion is directly carried out on the compressed Fourier projection coefficient in a GPU memory to restore the insensitive dimension by customizing a kernel operator in the GPU, and the attention weight is calculated. Compared with the prior art, the method has the advantages of reducing the storage overhead of the large language model and the like.

The invention discloses a hyperspectral image classification method and device, equipment, a medium and a product, and relates to the field of hyperspectral image classification, and the method comprises the steps: constructing a full connection graph with a global receptive field; constructing a graph sparse network based on a dynamic sparse graph learning framework; learning the optimal edge connection weight of the graph sparse network by using a label propagation algorithm; after the updated fully connected graph is subjected to graph inverse projection, determining a classification label of each node in the fully connected graph after graph inverse projection based on a graph convolutional neural network, and obtaining a feature graph with the classification labels; and fusing the original features of the hyperspectral image and the feature map with the classification label, inputting the fused feature map into a pixel-level map neural network, and determining a final classification result of the hyperspectral image. According to the invention, the hyperspectral image classification precision can be improved.

The invention discloses a geometric perception multi-view consistency image generation method based on a diffusion model. The geometric perception multi-view consistency image generation method comprises the following steps: S1, multi-modal condition coding: uniformly coding conditions such as texts, reference images and camera postures into fusion features and inputting the fusion features into U-Net; and S2, gating multi-path attention fusion: dynamically fusing the multi-source features in the U-Net through a gating multi-path attention mechanism, and enhancing the multi-view two-dimensional features. And S3, voxel feedback closed-loop refinement: inversely projecting the two-dimensional features to a three-dimensional voxel space, and re-projecting the two-dimensional features back to two dimensions after three-dimensional volume accumulation so as to form closed-loop geometric feedback. And S4, residual error correction autoregressive sampling: generating an image according to autoregressive of a view angle sequence, reprojecting and correcting a current sampling track by using front view angle information, and inhibiting error accumulation. And S5, composite supervision collaborative optimization: carrying out collaborative optimization on the model in combination with multiple loss functions, and freezing a backbone network so as to realize multi-view generation with high geometric consistency and high quality.

The invention relates to a vehicle visual positioning optimization method in a structured scene, and the method comprises the steps: calibrating a camera of an unmanned vehicle, obtaining a distortion-removed image, converting the distortion-removed image into a bird's-eye view through the inverse projection, updating the initial depth estimation result of a foresight camera through the bird's-eye view, and obtaining reliable depth estimation. And performing vanishing point estimation based on space line segment cosine similarity by using depth information in reliable depth estimation, estimating the attitude of the unmanned vehicle and the sub-tratta coordinate system, obtaining an attitude estimation result, updating an attitude extended Kalman filter of the unmanned vehicle by using the attitude estimation result, and determining attitude information of the unmanned vehicle. A sub-graph optimization objective function of the sliding window is further set, and a cost function is introduced to update the position information of the unmanned vehicle; and a key frame is extracted based on the attitude information and the position information, when the key frame detects that the unmanned vehicle returns to the intersection where the unmanned vehicle passes historically, a loopback frame is searched in an uncertainty ellipse, and a historical track is optimized after successful matching.

The present application relates to computer vision and three-dimensional reconstruction technical field, disclose a kind of based on image recognition's building decoration component three-dimensional modeling method, method includes the following steps: receiving near-axis active illumination image sequence, solving camera pose and equivalent point light source trajectory;Inverse projection two-dimensional foreground mask generates maximum base envelope grid;Combining brightness gradient and texture high-frequency edge tensor field removes false shadow, extract effective sweep shadow mask;Tracking shadow residual optical flow displacement and combining light source trajectory analysis absolute depth, generate local three-dimensional point cloud;Surface reconstruction is executed to point cloud and opening plugging constructs negative space primitive entity grid;Extract the Laplace curvature energy of interface to generate normal shrinkage constraint, control entity grid after shrinkage and envelope grid execute Boolean difference set operation output model.The present application extracts absolute depth in featureless recessed area, and eliminates the boundary interference and distortion of grid Boolean operation.

The application provides a kind of plane panel surface coating profile size detection method, comprising the following steps: arranging positioning target in the uncoated area of the plane panel to be measured and collecting the image of the plane panel to be measured, extracting the panel outer contour and coating profile from the image;High-precision three-dimensional reconstruction is carried out on the positioning target to obtain three-dimensional point cloud data, and the spatial plane expression of the plane panel is fitted;The coating profile and the panel outer contour are calculated by inverse projection to the plane panel surface, and the three-dimensional point cloud data of the coating profile and the panel outer contour is obtained, and the two-dimensional point cloud data is obtained by dimension reduction on the three-dimensional point cloud data;Read the design profile of the plane panel, calculate the coordinate conversion relationship with the outer contour as the registration object;The two-dimensional point cloud data of the coating profile is registered with the design profile by the coordinate conversion relationship, and the detection value of the coating profile size error is obtained.The application can be suitable for surface coating size detection of plane panels with different sizes, sizes and coating shapes.

This invention relates to the field of point cloud computing technology, specifically a hatch positioning method based on multimodal lidar point clouds. First, a multi-dimensional point cloud containing three-dimensional coordinates and reflection intensity is acquired. The coordinate system of the multi-source data is unified and fused through radar extrinsic parameter calibration or point cloud fusion algorithms. Then, the deck area point cloud is segmented based on normal vector filtering and DBSCAN clustering, and converted into a two-dimensional image through orthogonal projection. Next, edges are extracted using an improved Canny operator, and the hatch planar position is determined by least-squares rectangle fitting. Finally, the three-dimensional coordinates of the hatch are restored through inverse projection mapping. This invention overcomes the detection limitations of a single radar, effectively improving the accuracy and stability of hatch positioning under complex conditions. It can provide reliable positional information for the automated control of bulk carrier unloading operations, and has significant industrial application value.

An imaging system (10, 100, 200) includes a housing assembly (20) having a head unit (18) configured to be at least partially directed toward a target site (16, 116, 216). An optical coherence tomography (OCT) module (14, 114, 214) and a visualization module (12, 112, 212) are located in the housing assembly (20) and are configured to obtain OCT data and visualization data of a target site (16, 116, 216), respectively. The system (10, 100, 200) includes a controller (C) configured to generate a scan pattern for a calibration region selected in a calibration target. And synchronously acquiring OCT data of the calibration area. The controller (C) is configured to obtain, based on the OCT data, a projected two-dimensional OCT image of the calibration region as an average intensity inverse projection. The controller (C) is configured to register the projected two-dimensional OCT image to a corresponding view extracted from the visualization data via a cascaded image registration process having a coarse registration phase and a fine registration phase.

A point cloud semantic segmentation method based on a cross-modal Transformer belongs to the field of semantic segmentation technology. The method focuses on guiding the dense visual information of the camera image into the point cloud semantic segmentation task to complete the point cloud semantic segmentation task. First, the three-dimensional point cloud is unified to the camera image coordinate system according to the coordinate system transformation relationship, and then the perspective projection is used to obtain the two-dimensional representation of the three-dimensional point cloud. Then, the multi-scale feature map is calculated and extracted, and then cross-modal attention fusion is performed. The camera image and the projected point cloud image are fused at the feature level. Finally, the image is upsampled and classified. The classification result is projected onto the three-dimensional point cloud according to the inverse projection transformation relationship to complete the point cloud semantic segmentation task. The Transformer self-attention mechanism is used to establish cross-modal feature dependencies. The feature information of the two modalities of image and point cloud is combined to enhance the feature expression ability of the model. A boundary loss function is designed to emphasize the boundaries of semantic objects, thereby improving the boundary accuracy of object segmentation.

The invention discloses a three-dimensional point cloud reconstruction method and system based on ultra-wideband signal multi-view fusion. A commercial pulse wireless ultra-wideband radar is adopted to collect signals of an object under multi-view; background reflection reduction and effective distance box positioning are carried out on the reflection signals; the radar signal is subjected to data enhancement through virtual antenna offset and phase rotation; designing a signal multi-view depth reconstruction network model, extracting three-dimensional structure information of an object from a radar signal, and generating a multi-view object depth map; and performing inverse projection on the multi-view depth map, converting the depth map into local point cloud, and generating global three-dimensional reconstruction point cloud of the target object through point cloud fusion. The method provided by the invention still has relatively high robustness in a scene in which the object is shielded, and can be widely applied to the fields of logistics sorting, security check and the like.

The present application relates to a kind of radar scattering cross section test methods based on deep learning and ISAR imaging inversion, the method includes using ISAR imaging test mode to obtain target and calibration body broadband radar scattering echo signal, calibration processing and applying filter-inverse projection imaging algorithm to carry out ISAR imaging, application completes the deep learning model of accurate extraction ISAR image scattering signal area, application FFT transform-interpolation integral ISAR imaging inversion algorithm high-precision inversion reconstruction RCS value, change ISAR imaging aperture center azimuth angle traversal acquisition target all azimuth angle RCS data.The radar scattering cross section test method based on deep learning and ISAR imaging inversion provided by the present application solves the accurate extraction ISAR image scattering area and ISAR image high-precision inversion reconstruction RCS and other test practical problems, expands RCS data acquisition method.

The invention relates to a method for fusing a monitoring video and a 3D virtual scene, and belongs to the technical field of image processing and computer vision. The method comprises the following steps: firstly, constructing a three-dimensional digital scene and completing space calibration of a camera; then, camera offset is detected and corrected through real-time video analysis, and meanwhile, a target in a picture is recognized and tracked; secondly, the 2D pixel coordinates of the target are converted into space coordinates in a 3D scene through inverse projection calculation, and the positioning precision is improved through the multi-view intersection technology; and finally, based on the converted 3D coordinate data, performing fusion rendering on a video picture in a rendering engine, and outputting an interactive three-dimensional scene which can be used for situation awareness and behavior analysis. According to the method, the problems of inaccurate virtual and real space registration, dynamic error accumulation and unnatural rendering in the prior art are effectively solved, and high-precision and dynamic fusion of the monitoring video and the virtual scene is realized.

The invention provides a transition state searching method and system for complex biomacromolecule conformational change and application. The method comprises the steps that an input data set related to molecular dynamics simulation trajectory data is acquired; dimensionality reduction is carried out on an input data set through a reversible dimensionality reduction method, trajectory data is projected from a high-dimensional space to a low-dimensional space on the basis of reserving transition state information, and a low-dimensional free energy surface is generated; searching a low-dimensional candidate transition state sample on the low-dimensional free energy surface by adopting a low-dimensional search method; inversely projecting the low-dimensional candidate transition state sample back to a high-dimensional space through a reversible dimension reduction method to obtain a complete candidate transition state structure; and verifying the candidate transition state structure, and screening out a real transition state structure according to a verification result. According to the method, high-dimensional transition state structure search with high calculation efficiency and high accuracy is realized through reversible dimension reduction and low-dimensional search based on generative artificial intelligence, and the method is high in expandability, high in physical reliability and suitable for a complex biomolecular system.

The invention provides a gastrointestinal three-dimensional dynamic imaging analysis method based on conductivity distribution modeling, and the method comprises the steps: applying at least three composite electric signals with different frequencies to a gastrointestinal region, collecting voltage response signals, and constructing a dynamic conductivity parameter matrix based on frequency domain feature separation. And then, spatial discretization processing is carried out by using a nonlinear interpolation algorithm, a three-dimensional conductivity distribution map is generated, dynamic modulation and demodulation processing is carried out on the map, and a frequency band weight coefficient is adjusted. And mapping the modulated data to a dynamic imaging coordinate system through an inverse projection algorithm to generate a three-dimensional dynamic imaging result. According to the space coordinates and the change rate of the conductivity gradient mutation area, the position and the activity state of the gastrointestinal abnormal area are judged, and a self-adaptive feedback mechanism is adopted to update boundary condition parameters of the initial conductivity distribution model. The accuracy and efficiency of gastrointestinal disease diagnosis can be improved.

The invention relates to a radar cross section test method based on deep learning and ISAR imaging inversion, and the method comprises the steps: employing an ISAR imaging test mode to obtain broadband radar scattering echo signals of a target and a calibration body, carrying out the calibration processing, and carrying out the ISAR imaging through employing a filtering-inverse projection imaging algorithm, and accurately extracting an ISAR image scattering signal region by using the trained deep learning model, inverting and reconstructing an RCS value with high precision by using an FFT (Fast Fourier Transform)-interpolation integral ISAR imaging inversion algorithm, and traversing by changing an ISAR imaging aperture center azimuth angle to obtain target omni-directional angle RCS data. According to the radar cross section test method based on deep learning and ISAR imaging inversion provided by the invention, the test practice problems of accurate extraction of an ISAR image scattering region, high-precision inversion reconstruction of the RCS by the ISAR image and the like are solved, and an RCS data acquisition method is expanded.

The present application relates to the field of road nondestructive testing, and discloses a three-dimensional ground penetrating radar imaging method based on interface constraint coherent weighting and a radar. The method comprises the following steps: discretizing a road section to be detected into a three-dimensional voxel grid, and calculating the two-way travel time of electromagnetic waves from the ground surface to the grid points based on the electromagnetic wave propagation speed. The signal amplitude of the radar data is extracted based on the two-way travel time, and the stacking energy and the coherence factor of the grid points are calculated. The interface penalty weight of the sudden interface to the grid points is calculated and weighted with the stacking energy and the coherence factor. After traversing all the grid points, the ground penetrating radar data volume is generated. Through the deep integration of the continuous steps of first arrival wave correction, layered velocity modeling, weighted inverse projection and customized physical penalty weight, the divergent hyperbolic curve characteristics in the original radar image can be converged and focused into the real disease geometry, the lateral resolution is significantly improved, the disease boundary can be accurately identified, and the depth positioning error is reduced.

The present application belongs to the technical field of face attribute fidelity editing, and relates to a super-resolution face attribute lossless editing method, device and readable storage medium; an original super-high resolution source image is input into a local perception network, a core topological anchor point set of a target face is output, a forward spatial affine transformation mapping operator is solved to perform affine clipping and normalized projection on a region of interest containing the target face in the original super-high resolution source image, and a local feature sub-manifold image is obtained; the local feature sub-manifold image is input into a generative adversarial network to output an evolved feature manifold image; a matrix inversion operation is performed on the forward spatial affine transformation mapping operator to obtain an inverse spatial mapping operator to perform topological inverse projection transformation on the evolved feature manifold image, and output a low-dimensional evolved image; the low-dimensional evolved image is enlarged and rotated to a physical coordinate system of the original super-high resolution source image by using a high-order resampling interpolation algorithm, and a high-dimensional registered evolved image is obtained.

The application discloses a mesh texture simplification algorithm suitable for three-dimensional reconstruction, and comprises the following steps: S1, reference three-dimensional model scene construction; using the recovered scene structure in three-dimensional reconstruction and the calibrated image, performing texture mapping on the original fine three-dimensional mesh to complete the reconstruction of the reference three-dimensional model scene; S2, reference three-dimensional model scene image acquisition; according to the three-dimensional space relative attitude relationship between the reference three-dimensional model scene and the view internal and external parameters, using the inverse projection principle to perform three-dimensional mesh to two-dimensional image rasterization calculation pixel by pixel and view by view, and completing the acquisition of the reference image set; S3, mesh and texture simplification; using the QEM algorithm to simplify the mesh, using the reference image set as a data source and using the texture reconstruction algorithm to perform texture remapping and simplification. The application has the advantages that the algorithm can support mesh and texture simplification together, under different texture simplification parameters, the texture almost has no distortion and can significantly reduce the texture data amount.

This invention belongs to the field of industrial automation control technology, specifically relating to an intelligent optimization system for process parameters in automotive steel wire production. The system includes: a data acquisition module, a deviation calculation module, a prediction module, and a control module. The data acquisition module constructs a Lagrange coordinate system with material particles as the core, collects time-series process data throughout the entire process, and maps it to the material particles. The deviation calculation module calculates the historical deviation modulus of the material particles based on the arc-length integral of the state deviation trajectory. The prediction module constructs a dissipative damping model based on the historical deviation modulus and calculates the performance loss of the material particles. The control module generates process parameter compensation quantities through inverse projection based on a preset historical regression gradient vector and performance loss, and implements feedforward control of the equipment. This invention achieves closed-loop compensation for historical accumulated damage, improving the consistency of automotive steel wire product performance.

The invention relates to a method for automatic segmentation of a dental arch that comprises acquiring a three-dimensional surface of the dental arch, in order to obtain a three-dimensional representation comprising a set of vertices, generating virtual views from the three-dimensional representation, projecting the three-dimensional representation onto each two-dimensional virtual view, in order to obtain an image representing each vertex on the virtual view, processing each image by means of a deep learning network, carrying out inverse projection of each image in order to assign, to each vertex of the three-dimensional representation, one or more pixels of the images in which the vertex appears and to which it corresponds, and assigning one or more probability vectors to each vertex, determining the class of dental tissue to which each vertex most probably belongs based on the probability vector or vectors.

Disclosed are a method, device, and system for X-ray CT contrast imaging using magnetic nanoparticles. The method includes: rotating a target object containing magnetic nanoparticles as contrast agent for one circle relative to X-rays, recording multiple consecutive transmission projection images of the target object at all angles; for multiple transmission projection images at each angle, after conducting image post-processing operations, performing inverse projection transform on magnetic nanoparticle imaging projection images at all angles to obtain a tomographic imaging result. The image post-processing operation includes: time-frequency transform: converting the third dimension of the three-dimensional time domain matrix to frequency domain to obtain a three-dimensional frequency domain matrix; feature extraction: selecting a two-dimensional matrix with frequency of kf along the third dimension of the three-dimensional frequency domain image matrix; k is a positive integer; f is the frequency of the excitation magnetic field.