46 results about "Light-field camera" patented technology

This invention discloses a method, device, storage medium, and program product for calibrating and aligning a light field camera and an RGB camera. The deviation value is calculated by matching the images from the two cameras and fed back to the engineer, allowing for more precise fine-tuning. The deviation value provided by this invention is accurate and objective, unaffected by human factors. With objective guidance, engineers can significantly improve the efficiency of fine-tuning camera positions, thereby enhancing the efficiency of this task in actual production assembly.

The invention belongs to the technical field of computer vision and light field imaging, and discloses a large-area highlight detection and repair method and system. The method comprises the following steps: imaging by using a light field camera array to obtain a multi-view-angle light field image, and calculating an initial parallax; the method comprises the following steps: obtaining a multi-view highlight area mask through highlight detection, removing the highlight area mask to obtain a light field image without a highlight area, carrying out light field refocusing to obtain a non-highlight refocusing image stack, applying the highlight area mask obtained through central view angle detection to an initial parallax image, and constructing a reliable parallax set to realize parallax layering; according to parallax layer information of a non-highlight area pixel adjacent to a highlight pixel space in the highlight area, carrying out adaptation on a parallax layer of the highlight pixel; and constructing a hierarchical restoration strategy based on observation conditions of different viewpoints when no-highlight refocusing image of the highlight pixel in the corresponding parallax layer is superposed and averaged at the pixel position, thereby realizing restoration of the highlight area. According to the method, high-quality and large-area highlight repair can be realized.

The application provides a robot tactile and visual fusion double-ring coupling adaptive grasping control method, relates to the technical field of robot grasping control, and comprises the following steps: acquiring surface characteristic data through a light field camera array, collecting dynamic contact information and extracting stability features by a sensor array, constructing cross-modal coupling and fusing tactile and visual information based on Tucker decomposition, establishing mapping parameters and constructing an adaptive time window to optimize grasping parameters, and adjusting grasping posture and clamping force in combination with Lyapunov stability criterion and a dynamic compensation mechanism. The method can improve the stability and robustness of robot grasping in a complex environment.

The invention discloses a tea bud picking obstacle avoidance path planning method and system based on a multi-modal information fusion diffusion model. A tea tree canopy image is acquired through a light field camera, a dense three-dimensional point cloud is constructed, tea bud target detection and positioning are realized in combination with an improved YOLOV8 and a PointNet + + network, a path generation method based on a U-net architecture condition denoising diffusion model is provided, path planning is modeled as a condition track generation problem, and the problem is solved. According to the method, multi-modal condition information input and a target guiding steering function are combined, multi-step denoising iteration is executed, a continuous, smooth, collision-free, dynamic and feasible mechanical arm picking path is directly generated, the sawtooth-shaped defect and post-optimization dependence of a path generated through traditional sampling algorithms such as RRT are avoided, and the path quality and execution stability are improved. The system comprises a light field acquisition module, a three-dimensional reconstruction module, a visual positioning module, a path planning module, an execution control module and the like, realizes accurate obstacle-avoiding picking of tea buds, is suitable for path planning of picking operation of tea leaves and other fruits and vegetables, and provides technical support for agricultural picking and harvesting.

The invention discloses a naked-eye 3D imaging method, and relates to the technical field of naked-eye 3D, and the method comprises the steps: employing a binocular camera or a light field camera to obtain multi-angle data, enabling the distance between cameras to change according to a target distance through dynamic baseline adjustment, increasing a baseline in a long distance to improve the depth-of-field resolution, and reducing the baseline in a short distance to enhance the local stereoscopic impression; meanwhile, in combination with an AI depth-of-field prediction model, a low-light environment and a high-reflectivity surface are subjected to deep optimization treatment, and errors caused by light interference are reduced; in the depth calculation process, an optical flow analysis method is adopted, a pixel motion vector field is calculated through a Horn-Schunck optical flow, the stability of optical flow calculation is optimized in combination with Kalman filtering, parallax matching or optical field reconstruction is performed based on an optical flow calculation result, for a binocular camera mode, a block matching method is adopted to calculate parallax, and the optical field is reconstructed. And in the light field camera mode, depth reconstruction is carried out by using an EPI polar plane parallax estimation algorithm.

The present application belongs to the technical field of camera calibration, and discloses a focusing light field camera geometric parameter calibration method, which comprises the following steps: using a focusing light field camera to collect and acquire a white image and a corresponding checkerboard original light field image under different object distances; determining the focal plane of the camera based on the original light field image; using an edge diffusion function information fitting-based corner point detection algorithm to calculate the checkerboard corner point coordinate information in the original light field image on the focal plane; calculating the center coordinate information of each microlens in the microlens array of the camera and the size information of a single microlens in the microlens array based on the white image; calculating the pixel block size in the microlens sub-image taken by the center sub-aperture image according to the above information; and calculating the distance from the virtual imaging surface to the microlens array and the distance from the microlens array to the plane where the photosensitive element is located based on the pixel block size and in combination with a Gaussian imaging formula. The present application can improve the calibration accuracy of the geometric parameters of the focusing light field camera.

The invention discloses a spectrum optimization-based double-layer medium wide spectrum parameter simultaneous reconstruction method, relates to the technical field of spectral imaging, and aims to solve the problems that when wide optical parameter reconstruction is carried out based on a multi-layer medium model in the prior art, detection radiation optical signals are limited by a traditional contact type measurement mode, and the reconstruction precision is low. On the basis of a traditional participating medium optical parameter measurement and reconstruction method, a light field camera is introduced to more accurately detect a reflected radiation measurement signal of a medium boundary; a reconstruction optimization algorithm of searching an initial point by a steepest descent method and searching an optimal solution by a conjugate gradient method is combined, so that the efficiency of accurate reconstruction of the double-layer medium is greatly improved, the limitation of a traditional contact measurement mode is broken through, and the reconstruction precision is improved; in view of the problem that a radiation measurement signal is extremely low in sensitivity to a reduced scattering coefficient of a bottom-layer medium, prior information of a spectrum dimension is introduced, reconstruction results of an absorption coefficient and a reduced scattering coefficient of a double-layer medium, especially the bottom-layer medium, are improved, and simultaneous reconstruction of a double-layer medium absorption spectrum and a reduced scattering spectrum is realized.

The application discloses a kind of light field imaging parameter control method and device of underwater light field camera, the method includes: S1, the underwater light field camera including light field camera and waterproof spherical shell is built;S2, the physical parameter of light field camera is acquired, and the light propagation model of the waterproof spherical shell is built;S3, according to light propagation analysis, the equivalent object distance of light field camera is acquired;S4, based on the disparity between the microlens array of light field camera and equivalent object distance, the shortest distance between the center of the waterproof spherical shell of underwater light field camera and the lens optical center of light field camera is calibrated;S5, based on equivalent object distance and the shortest distance between the center of the waterproof spherical shell of underwater light field camera and the lens optical center of light field camera, the light field imaging parameter of underwater light field camera is optimized;The application can improve the light field imaging definition of underwater light field camera;Improve the field of view angle of underwater light field camera, visual effect is better, improve the light field imaging quality of underwater light field camera.

To address the shortcomings of existing methods for detecting surface texture defects in complex curved parts, such as low detection efficiency, poor stability and recognition, and difficulty in rapidly and accurately quantitatively characterizing sub-millimeter-level surface texture defects in three dimensions, this invention provides a rapid three-dimensional quantitative characterization method for surface defects in curved parts. This invention utilizes a light field camera to simultaneously acquire two-dimensional images and three-dimensional point cloud data, improving the consistency and stability of subsequent three-dimensional quantitative characterization of defects. It employs a deep learning semantic segmentation model to extract two-dimensional contour information of defects, providing precise semantic guidance for subsequent three-dimensional spatial mapping. It establishes a coordinate mapping relationship between pixels in the two-dimensional image and physical points in the three-dimensional point cloud coordinate system, enabling precise selection of subsets of the defect's three-dimensional point cloud. Finally, it uses principal component analysis (PCA) to establish a local coordinate system, and within this coordinate system, employs the minimum bounding box method to calculate the three-dimensional geometric dimensions of the defects, achieving rapid and accurate three-dimensional quantitative characterization of the defects.

The invention discloses a double-lens cooperative light field camera system and a local light field analysis method, and relates to the technical field of light field cameras, the system comprises a main lens, a spectroscope, a micro lens array, a single lens, a main CMOS sensor, a preview CMOS sensor and a processor which are modularly designed; wherein the main lens is used for focusing scene light to capture full-light-field information; the spectroscope distributes light output by the main lens to a main light path and a preview light path according to a preset splitting ratio; the micro-lens array is used for decomposing transmission light into light field four-dimensional information containing direction and intensity; the main CMOS sensor is used for recording full-light-field data after the micro-lens array is decomposed; the single lens processes the reflected light to generate a preview image; the preview CMOS sensor is used for recording a preview image output by the single lens; and the processor extracts corresponding full-light-field data recorded by the main CMOS sensor according to the target point coordinates in the preview image, and analyzes the full-light-field data to generate a high-definition image. The calculation amount and the hardware cost can be remarkably reduced.

The present application relates to the technical field of light field imaging measurement, and particularly relates to an absolute depth measurement method based on a parallax angle of a polarized microlens light field image. The method comprises the following steps: 1) image pixel level classification based on multi-view sub-aperture depth maps, which provides a benchmark for optimization of a parallax map; 2) constructing a matching cost equation based on pixel classification results, so as to realize global optimization of a relative depth map from coarse to fine; 3) establishing a linear mapping model from the relative depth map to the absolute depth map based on the hardware structure of a microlens light field camera, so as to realize depth calibration; 4) combining the depth calibration with internal parameters of the light field camera, so as to calculate absolute depth physical parameters of an observed target. Thus, physical measurement and characterization of three-dimensional absolute size of a real observed object are realized, and absolute depth information of a target to be measured in a light field image can be obtained, in particular, absolute depth information of a weak texture and a blocked target can be obtained.

This invention relates to a continuous three-dimensional particle field reconstruction method based on single-light-field imaging using EM deconvolution, belonging to the field of optical imaging technology. This method employs a light field camera to capture light field images of a discretized three-dimensional flow field, acquiring the light intensity information of scattered light emitted by particles on each pixel of a CCD detector. Based on the light intensity information, a weight matrix for each voxel is calculated, and the entire weight matrix is ​​loaded only once. The nth frame of the light field image is loaded, and the non-zero pixels in the nth frame are identified. The non-zero pixel sequence formed by the scattered light from tracer particles is extracted. An EM deconvolution algorithm is used to iteratively calculate the non-zero pixel sequence, updating the intensity of each voxel, thereby reconstructing the three-dimensional particle field of a single frame. The above steps are repeated to complete the reconstruction of the three-dimensional particle field from multiple frames of high temporal resolution light field images, resulting in higher reconstruction accuracy and consistency, and less time consumption.

This invention provides a method and system for multispectral temperature measurement based on an infrared sensor. The method includes: calculating the optical parameters of a microlens array and a filter array according to optical formulas; then fabricating the microlens array through photolithography, development, pressing, and atomic force scanning; next, constructing an infrared multispectral radiation temperature measurement system; calibrating the system using a blackbody furnace to obtain the correspondence between the camera output signal intensity and the standard temperature; decoupling the multispectral light field image based on the imaging characteristics of the light field camera; and finally, using the decoupled data to sequentially calculate the true temperature of points on the high-temperature component, obtaining the two-dimensional true temperature field of the high-temperature component. This invention overcomes the limitations of visible light multispectral temperature measurement by combining multispectral light field measurement with infrared imaging technology to achieve two-dimensional temperature field measurement, effectively overcoming the shortcomings of existing optical temperature measurement systems, such as complex structure and inability to simultaneously achieve both high measurement accuracy and wide range.

This invention discloses a modular light field microscope system based on a Lytro light field camera and its debugging method, belonging to the field of optical imaging and microscopy technology. The system includes a microscope objective, a first lens, a second lens, and a Lytro light field camera arranged sequentially along the optical path. The focal length of the first lens is shorter than that of the second lens, forming a relay optical path. The system also includes a modular mechanical support assembly composed of a cage rod and a cage plate, and each optical component can be slidably adjusted along the axial direction of the cage rod. This invention also discloses a step-by-step debugging method for the system. This invention utilizes the advantages of the commercial Lytro light field camera—small size and low cost—and overcomes the technical bottleneck of Lytro cameras being unsuitable for microscopic observation through a focal length gradient dual-lens relay optical path design. The system is compact, portable, and easy to debug, enabling high-quality four-dimensional light field imaging of microscopic samples. It is suitable for scenarios such as three-dimensional velocimetry of fluid particles, biological fluorescence microscopy, and in-situ microscale detection.

This invention belongs to the field of logistics management technology, and specifically relates to an automatic weighing system and method for ore loading vehicles based on light field technology. The system includes a light field intelligent detection device, a weighing system, a loading vehicle, and a barrier control system. It acquires three-dimensional vehicle data through a light field camera array, combines this data with a deep learning algorithm to calculate the material volume and weight in real time, and performs dual verification with weighbridge data to achieve high-precision load measurement. An adaptive grid algorithm and density deviation detection are employed to effectively identify theft or overloading. This system can operate stably even in dusty environments, achieving a volume measurement accuracy of ±0.8% and a weight verification error of ≤±1%, significantly improving the efficiency and safety of mine weighing and reducing manual intervention.

This invention provides a dual-modal real-time microscopic imaging device, comprising: an illumination component configured to illuminate a sample; an imaging component configured to receive light signals from the sample and emit scattered light signals and fluorescence signals via different optical paths; a full-field camera configured to receive scattered light signals emitted from the imaging component; a fluorescence imaging two-dimensional array detector and a GISC camera configured to switchably receive fluorescence signals emitted from the imaging component; and an image processing processor electrically connected to the full-field camera, the fluorescence imaging two-dimensional array detector, and the GISC camera. The dual-modal real-time microscopic imaging device and method of this invention solve the problems of low spatial resolution, inability to simultaneously perform real-time dual-modal imaging, and difficulties in multi-color real-time imaging in existing dual-modal phase and fluorescence imaging methods.

This application discloses a color difference detection method, apparatus, device, and storage medium. The method includes: acquiring images to be calibrated obtained by a light field camera taking pictures of a black and white calibration object with no filter and different color filters respectively; performing scale calibration processing on the images to be calibrated to obtain a color calibration matrix for each color; acquiring an image to be detected taken by the light field camera, in which an extended reality device displays an image of the color calibration object; and detecting the color difference of the extended reality device based on the color calibration matrix and the image to be detected for the corresponding color.

The invention provides a flame three-dimensional multi-physics field reconstruction method based on a light field focus stack, which comprises the following steps of: imaging blackbody radiation sources at different temperatures by using a light field camera, and establishing and outputting a fitting relationship between response values of color channels of a light field refocusing image and corresponding spectral radiation intensities according to a Planck law; acquiring a flame light field focus stack image sequence of the flame light field original image by adopting a super-resolution reconstruction method based on a sub-aperture image; and outputting temperature distribution and soot concentration distribution of the corresponding flame slice image by using the three-dimensional convolutional neural network model constrained by the fused radiation transmission physical information. According to the method, the implicit correlation model between the flame focus slice and the three-dimensional physical field of the flame focus slice is directly constructed by fusing the three-dimensional convolutional neural network model constrained by the forward radiation transmission physical information, and the generalization ability and the interpretability of the model can be improved, so that the joint reconstruction work of the physical field is more effectively realized.

The invention relates to the technical field of image detection, in particular to a glass surface flatness intelligent evaluation method based on deep learning, which comprises the following steps: acquiring four-dimensional light field data of a glass surface through a micro-lens array light field camera; performing digital refocusing processing on the four-dimensional light field data to generate a multi-focal plane image stack; and inputting the multi-focal plane image stack into a refraction compensation parallax network, performing ray tracing compensation by combining glass refractive index priori knowledge, generating a refraction correction depth map, and outputting a flatness evaluation result based on the refraction correction depth map. According to the method, a refractive index prior and differentiable ray tracing module is introduced in a depth estimation link, the problems of geometric misjudgment and depth distortion generally existing in transparent body detection in a traditional method are solved, and the physical credibility and calculation precision of a detection result are improved.

The invention provides an efficient end-to-end sparse-to-dense light field reconstruction method based on space-angle multi-dimensional interaction and a guide residual network, and the method comprises the steps: enabling a sparse view collected by a light field camera to serve as the input of a network, and generating a dense light field sub-view through a designed space-angle multi-dimensional interaction module; and performing view optimization on the generated dense light field sub-view through a guide residual refining module, and finally realizing dense light field reconstruction. The method comprises the specific steps that firstly, an input sparse light field view is converted into a YCrCb format from an RGB format; secondly, space-angle information of a Y channel view is extracted; then a space-angle multi-dimensional interaction module generates dense sub-views according to information extracted from the Y channel view; performing texture detail optimization on the generated dense sub-views through a guide residual refining module; and finally, orderly arranging the optimized dense sub-views and the original view to realize dense light field reconstruction.

The application discloses a kind of fast light field depth estimation methods based on multiple parallax scales.The specific steps are as follows: (1) constructing database;(2) input light field image into neural network, neural network extracts feature vector, interacts with edge enhancement module to enhance edge characteristics, and constructs matching cost of different parallax scales according to the obtained feature vector, carries out matching cost regularization through cost aggregation module, and finally obtains light field parallax map through parallax regression;(3) construct the loss function of neural network, the loss function includes absolute value loss function and weighted binary classification cross-entropy loss function;(4) according to the light field camera parameter, the camera focal length and baseline length corresponding to adjacent light field image are obtained, and then according to the light field parallax map, it is converted into light field depth map;(5) input the light field image of test set into the trained neural network, and obtain the light field depth map.The method of the application can directly obtain the accurate depth map of the light field image.

This invention proposes a large baseline light field video depth estimation method, belonging to the field of visual image processing. Specifically, it involves: First, acquiring images using a light field camera, sequentially reading in the original single-viewpoint video and splitting it into frames, grouping sub-aperture video frames from different viewpoints at the same time. For each group of images, horizontal and vertical EPI images are generated using the row and column pixels of each frame. Then, gradient map template detection and graph-based image segmentation are performed on the center row / column to obtain macro-pixel block segmentation reference results for each EPI frame. Next, the target pixel position is located in each EPI frame as a preliminary depth estimation result, and the preliminary estimation result is mapped using triangular coordinates to determine the depth information of the target pixel, i.e., the depth information of the original pixels contained in the horizontal / vertical EPI images. This invention improves the coordinate matching accuracy of macro-pixel blocks and also enhances the accuracy of the depth estimation results.

This invention discloses a parameter design method for a long-wave infrared focusing light field camera, belonging to the field of computational imaging technology. The method includes: determining the parameters of the main lens and the pixel size of the detector based on the imaging object distance, while satisfying the long-wave infrared diffraction limit constraint; employing a high-fill-rate microlens array structure; calculating the axial distance between the microlens array, the detector, and the main lens based on pupil matching and multi-view depth recovery requirements; determining the allowable circle of confusion according to the long-wave infrared diffraction limit, and designing a multi-focal-length microlens array to achieve continuous depth-of-field coverage; finally, optimizing the microlens aperture using a depth resolution model to balance depth resolution and depth-of-field range. This invention, through the collaborative design and overall optimization of the parameters of the main lens, microlens array, and detector, achieves long-distance, large depth-of-field, and high depth-resolution light field imaging in the long-wave infrared band, providing effective design guidance for infrared three-dimensional sensing systems.

The application discloses a kind of vector light beam inseparability measuring device, comprising: laser, first beam splitter, vector light beam generation module, quarter-wave plate, first collimator, polarizer, second beam splitter, polarization beam splitter, light field camera module and calculation module;The present application uses light field camera to build measuring device, which can reconstruct the light field complex amplitude of two polarization channels of vector light beam by measuring the off-axis interference hologram under two orthogonal polarization channels of vector light beam, and then completes the inseparability test of vector light beam.The present application does not depend on phase hologram loading device in the process of decoding and calculating the inseparability of vector light beam, so that the inseparability measurement rate of vector light beam and the complexity of optical path are no longer limited by demodulation device, the fast, efficient and low-cost measurement of the inseparability of vector light beam can be realized, and the problem that the inseparability of multi-order vector light beam superposition state cannot be simultaneously measured in the prior art is solved.

Systems, methods, and apparatuses are provided herein for changing the positions and / or shapes of microlenses of a light field camera to generate light field images with enhanced depth of field and / or dynamic range. This may be accomplished by a light field camera determining a plurality of focus measurements for a plurality of microlenses, wherein one or more of the plurality of microlenses vary in distance from a main lens of the light field camera. The light field camera may use the plurality of focus measurements to determine a microlens of the plurality of microlenses that captures information that is the most focused. The light field camera can then determine defocus functions for the microlenses that are not capturing information that is the most focused. The light field camera can then generate a light field image using the determined defocus functions and the information captured by the plurality of microlenses.

The invention provides a flame light field imaging measurement uncertainty quantitative evaluation method, which comprises the following steps of: acquiring a light field original white image of a light field camera under an actual error condition and an ideal condition, and determining a central pixel block of each micro-lens sub-image area, the method comprises the following steps: acquiring an original light field image of a light field camera for imaging a target scene under an actual error condition, dividing a reference image and a to-be-measured image into corresponding sub-image pairs, sequentially calculating a local mean square error and a local structural similarity of each sub-image pair, sequentially calculating an overall mean square error and a structural similarity of a sub-aperture image under each view angle, and obtaining a sub-aperture image; according to the method, a quality evaluation system of a local sub-image and a multi-view sub-aperture image is established for a flame light field image from the angles of image intensity and structural distortion; therefore, the quantitative evaluation of the response uncertainty of the flame light field imaging measurement system under the given error input is realized.

The application discloses a kind of mine area obstacle detection method, system, electronic equipment and readable storage medium, comprising: obtaining time-synchronized 4D light field camera point cloud data and laser radar point cloud data, according to the different position information in two sensors, solve out calibration parameter;Find each the near neighbor point of the laser radar point cloud data in 4D light field point cloud data, and whether it is abnormal point according to adjacent point number judgment;Then respectively obtain 4D light field camera and laser radar's obstacle point cloud, are re-registered to obtain new calibration parameter;Corresponding is carried out to update boundary box information, and the number of abnormal points is higher than the threshold value Obstacle point cloud filtering;Using 4D light field camera and laser radar detection characteristics reach high accuracy mine area target detection, using 4D light field camera can penetrate the characteristics of dust barrier, to the obstacle finally detected, to improve the range and precision of mine area obstacle detection Abnormal point number judgment.

The application relates to a welding workpiece surface spatter detection and removal method based on light field deep learning, which comprises the following steps: acquiring an original lens white image of a welding workpiece surface through a light field camera, and correcting the distortion of the original lens white image; extracting sub-aperture from the corrected lens white image to obtain a sub-aperture image array; calculating a light field polar plane feature map based on the sub-aperture image array; building a depth estimation network, inputting the light field polar plane feature map into the depth estimation network, and generating a scene depth map; screening an optimal sub-aperture image from the sub-aperture image array, and performing weighted fusion on the optimal sub-aperture image and the scene depth map to obtain an imaging image; inputting the imaging image into a U-Net network to identify a spatter area in the original lens white image; converting a pixel coordinate system of the original lens white image into a coordinate system of an end execution mechanism, and driving the end execution mechanism to the spatter area to remove the spatters.

The application relates to the technical field of image processing, in particular to a spatial position calibration method and device of a light field camera, equipment and a storage medium, wherein the method comprises the following steps: acquiring a light field image captured by the light field camera; rearranging image units in the light field image to obtain a montage image composed of different view angles; adjusting a current position of the light field image to a target position in a target coordinate system; when detecting that the light brightness distribution of the montage image composed of different view angles is uniform, obtaining a calibration position of the light field camera in the target coordinate system; when detecting that the light brightness distribution of the montage image composed of different view angles is not uniform, readjusting the current position of the light field image until the light brightness distribution of the montage image composed of different view angles is uniform. Thus, the embodiment of the application can calibrate the spatial position of the light field camera based on the montage image, the operation is simple, the error rate is low, the calibration precision is improved, and the simultaneous calibration of multiple cameras can be realized, thereby improving the calibration efficiency.

The invention discloses a small baseline light field depth estimation method based on observability driving. The method comprises the following steps: firstly, constructing a large / medium / small three-gear equivalent baseline set on the same light field data according to an angular step length; calculating observability scores related to imaging geometry in each file, selecting sub-view angles according to the observability scores, and constructing a weighted matching body; taking the large base line as a teacher, taking the medium / small base line as a student, and adopting curriculum type distillation training to realize migration of stable geometric prior from the large base line to the small base line; in the inference stage, small baseline matching bodies are input into a trained student model, body aggregation, probabilization and expectation regression are completed to obtain parallax, and parallax-depth mapping is performed according to baselines and focal lengths. According to the method, on the premise that external data is not introduced, the precision, boundary fidelity and robustness under the small baseline condition are improved, the view angle and the calculation overhead are controllable, and the method is suitable for three-dimensional perception application such as a light field camera, a multi-camera array and AR / VR.