2309 results about "Pixel pair" patented technology

In implementation of techniques for three-dimensional reconstructions based on Gaussian primitives, a computing device implements a reconstruction system to receive a first digital image depicting an object from a first angle and a second digital image depicting the object from a second angle. The reconstruction system segments the first digital image and the second digital image into patches. The reconstruction system then generates, using a machine learning model, three-dimensional Gaussian primitives that predict parameters of points of the object in a three-dimensional space that correspond on a per-pixel basis to pixels of the patches. The reconstruction system then forms a three-dimensional reconstruction of the object for display in a user interface by merging the three-dimensional Gaussian primitives.

The invention relates to the technical field of traditional Chinese medicinal material defect detection, in particular to a traditional Chinese medicinal material defect detection method based on vision, which comprises the following steps: regularly acquiring a time sequence image of a traditional Chinese medicinal material sample, acquiring a multi-view image, carrying out pixel alignment on the time sequence image, and carrying out structured organization on the multi-view image according to a shooting direction to associate camera parameters; and generating a registration time sequence image sequence and a multi-view image set. According to the method, the time sequence images of the traditional Chinese medicine samples are collected regularly, pixel alignment is carried out, image offset caused by environment illumination fluctuation and equipment jitter is eliminated, and time-space consistency of dynamic variable quantity calculation is ensured. Structured organization is carried out on multi-view-angle images according to shooting directions, camera parameters are associated, a geometric constraint relation between view angles is established, and the problem that three-dimensional reconstruction precision is insufficient due to view angle isolation in a traditional method is solved.

The invention discloses a medical image super-resolution reconstruction method based on multi-level attention guidance, and the method comprises the following steps: S10, constructing a deep learning network model based on a generative adversarial network architecture, which comprises a generator and a discriminator; the generator is based on an improved U-Net architecture, a hierarchical attention module and a dual-path feature processing module are configured in an encoder and a decoder of the generator, the hierarchical attention module adopts different attention strategies according to network levels to consider structure and texture, and the dual-path feature processing module separates and processes low-frequency and high-frequency information; the generator further comprises a multi-level feature fusion module for integrating the multi-scale features of the decoder, and an attention guide up-sampling module for final enhancement and dimension raising. The discriminator adopts a spectrum normalization U-Net architecture and uses multi-scale features for matching; s20, training the network model by adopting a composite loss function comprising pixels, adversarial, perception and total variation loss; and S30, inputting the low-resolution image into the trained model, and outputting a high-resolution image. According to the method, through deep fusion of multi-level attention and multi-scale feature processing, the image restoration quality can be remarkably improved, the texture detail definition can be enhanced, the anatomical structure accuracy can be ensured, and the noise robustness can be improved.

A scene flow estimation method includes: inputting a frame pair into an artificial intelligence (AI) network, and obtaining therefrom a motion embedding feature and a non-occluded-category label embedding feature corresponding to a target pixel in the frame pair; and estimating a scene flow corresponding to the frame pair based on the motion embedding feature and the non-occluded-category label embedding feature, the frame pair includes a first frame and a second frame, the first frame including a first color image and a first depth image and the second frame including a second color image and a second depth image, the non-occluded-category label embedding feature includes category information of an object corresponding to a pixel pair in the frame pair, the pixel pair includes a first pixel of the first frame and a second pixel of the second frame, and the second pixel corresponds to the first pixel.

The invention discloses a bionic moving target sensing, storage and calculation integrated photoelectric detection system based on continuous photoconduction, and relates to the field of bionic intelligent vision and the field of brain-like calculation, and the system comprises a synaptic bionic photoelectric detector array and a brain-like memristor array which are interconnected. When the synaptic-like bionic photoelectric detector array detects visual information of a moving target, a fixed bias voltage is applied to a device in each pixel, space illumination information corresponding to each pixel is converted into a continuously attenuated photocurrent signal, illumination information of a multi-frame dynamic target is continuously received in the continuous attenuation process of the photocurrent, and the visual information of the moving target is detected by the synaptic-like bionic photoelectric detector array. After the last frame of illumination information is input, a light current signal of each pixel is read, and a feature map memorizing all historical moment information is determined; the brain-like memristor array is based on the Ohm's law and the Kirchhoff's law, performs deep neural network operation on the feature map, outputs a current corresponding to a task result label, identifies a dynamic target, and can meet the requirements of a deep neural network and a complex dynamic visual task.

The invention discloses a shock wave overpressure field global measurement method based on multi-view image fusion, belongs to the technical field of explosive shock wave measurement, and is suitable for weapon equipment power evaluation and blasting safety analysis. The method comprises the following steps of: synchronously acquiring a time sequence image of the whole explosion process through a distributed multi-view high-speed imaging system, preprocessing the image by adopting pixel-by-pixel comparison, square operation enhancement and normalization processing, and positioning a shock wave edge contour; the blasting center coordinate is positioned through binocular parallax, the shock wave initial radius is determined by combining spatial domain analysis, and the key feature points of the front and rear edges of the wavefront are detected by using a dynamic search window and a radial gradient field. Based on multi-view constraints, a three-dimensional point cloud is generated through direct linear triangulation, and a three-dimensional wave front form is reconstructed in combination with least square spherical fitting. And finally, constructing a wavefront radius-time evolution model, and deducing a quantitative relationship between the instantaneous propagation velocity and the peak overpressure in combination with a Ranki ne-Huton iot relationship, thereby realizing the global high-precision calculation of the overpressure field.

The invention discloses an implicit coding and geometric prior-based indoor three-dimensional scene reconstruction method and system. The method comprises the following steps of: performing data processing on image data to be reconstructed to obtain corresponding pre-processed data; calculating rays under a world coordinate system corresponding to pixels needing to be trained on the image data according to internal and external parameters of the camera; inputting a three-dimensional coordinate code and an observation angle code of a sampling point on a ray and an implicit appearance code of a corresponding image into the neural radiation field model to obtain a predicted density value and a predicted color value, and performing volume rendering prediction based on a symbolic distance function to obtain a pixel value so as to obtain a trained neural radiation field model; and obtaining a standard implicit appearance code based on the density value and the color value, and inputting data at least comprising the standard implicit appearance code into the trained neural radiation field model to obtain scene geometric information. According to the method, the dependence of the existing reconstruction scheme on the three-dimensional consistency of the acquired data can be reduced, and the robustness and representation capability of the algorithm in a complex scene can be improved.

The embodiment of the invention discloses a mobile phone glass cover plate defect detection method, system and equipment and a storage medium, which are used for improving the accuracy of detecting a mobile phone glass cover plate. The method provided by the embodiment of the invention comprises the following steps: acquiring a first grayscale image of a to-be-detected glass cover plate and a second grayscale image of a standard glass cover plate; aligning the second gray level image with the first gray level image pixel by pixel; calculating a gray value difference value of each pixel of the first gray image and the second gray image to obtain a residual image; calculating an optimal Otsu threshold value of the residual image; performing binarization processing on the residual image according to the optimal Otsu threshold to obtain a binary image; and determining the defect position of the to-be-detected glass cover plate according to the binary image.

The invention discloses a display panel and a driving method thereof, the display panel comprises a plurality of first switches and a plurality of second switches, N columns of pixels are arranged in a first display area, and M columns of pixels are arranged in a second display area; the first scanning line is connected with N columns of pixels in the first display area and M columns of pixels in the second display area, the second scanning line is connected with the N columns of pixels in the first display area, and each first switch is connected with two data lines corresponding to sub-pixels with the same color in the first display area and the second display area; each second switch is connected with a data driving chip and a data line corresponding to a pixel in the first display area; wherein N and M are natural numbers greater than or equal to 2. The two data lines corresponding to the sub-pixels with the same color in the first display area and the second display area respectively receive the same data signal or different data signals to realize display in different display modes by switching on or switching off the first switch and the second switch, so that insufficient charging under a high refresh rate is improved.

The invention discloses a seabed sonar image processing method based on semantic segmentation. The method comprises the following steps: S1, obtaining preprocessed sonar image blocks; s2, generating a submarine landform boundary priori map corresponding to the preprocessed sonar image blocks pixel by pixel; s3, an improved U-Net double-branch semantic segmentation network is constructed, the improved U-Net double-branch semantic segmentation network comprises a semantic branch and a boundary branch, and a landform mutual feedback gating module is configured in each scale jumper connection layer; s4, obtaining a fusion feature map; s5, deploying to the autonomous underwater vehicle to form an onboard reasoning model; and S6, performing online semantic segmentation reasoning on the preprocessed sonar image block acquired in real time by using the onboard reasoning model, and outputting a semantic category diagram and a boundary vector diagram corresponding to the preprocessed sonar image block. According to the method, the average mIoU index of the model is improved in a small-target and multi-category hybrid scene.

The invention discloses a one-way door control method, system and equipment based on image recognition, and belongs to the field of images. The method comprises the following steps: synchronously obtaining a polarized light image, binocular depth point cloud data and a face image of a target area; the polarized light image and the depth point cloud three-dimensional feature are fused through spatial registration, and a point cloud polarization joint feature is constructed; performing pixel alignment on the joint feature and the face image, and generating a dynamic mask by using polarization degree data of the polarized light image; under the constraint of a dynamic mask, performing cross-modal fusion on the face image and the spatial feature map by adopting a multi-scale attention mechanism, and extracting a multi-modal feature vector containing a face spatial structure, a material attribute and a texture feature; performing similarity comparison on the vector and a preset database, and outputting verification information through a classification network; and triggering a self-adaptive control instruction of the one-way access control system according to the mapping relationship between the verification information and the control strategy. The one-way door can be accurately and effectively controlled in a complex reflective scene.

The invention discloses a high-load data steganography method, system and device and a medium, and the method comprises the steps: obtaining secret data of a power system, and converting the secret data into a novenary sequence; a quantum key distribution system is utilized to generate a shared key, the novenary sequence is encrypted through an encryption key, encrypted information is obtained, and the shared key comprises the encryption key and a mapping key; determining a corresponding first unit matrix from a preset mapping library based on the mapping key, and performing non-overlapping tiling on the first unit matrix in a two-dimensional plane to obtain a second-order reference matrix; and scanning a preset carrier image into a non-overlapping first pixel pair sequence according to a Hilbert curve, and embedding the encrypted information into the first pixel pair sequence by using the second-order reference matrix to obtain a steganographic image, so that the data transmission security can be improved.

The invention relates to the technical field of cholesteric liquid crystal electronic paper display, and discloses a cholesteric liquid crystal electronic paper display panel and a driving method thereof, and the cholesteric liquid crystal electronic paper display panel comprises an upper substrate, a lower substrate, a grid scanning line and a signal line, the electrode strip blocks continuously extend in the direction of a grid scanning line, each electrode strip block corresponds to one row of pixels, a single pixel corresponds to one pixel electrode, the electrode strip block corresponding to each row of pixels is conducted with the ITO electrode in the corresponding row through a conductive spacer, the ITO electrode is located on the periphery of a display area of the display panel, and in the direction of a signal line, the ITO electrode is connected with the electrode strip block in the corresponding row. The size of a single pixel is larger than the width L of each electrode strip block and larger than or equal to the width L'of a pixel electrode, and the distance S between every two adjacent electrode strip blocks is larger than or equal to the width W of a grid scanning line. According to the display panel and the driving method thereof, by optimizing the common electrode structure and the driving time sequence, the display performance and the refreshing efficiency of cholesteric liquid crystal electronic paper are remarkably improved.

The invention provides an interior trim part defect detection method and system, a computer and a storage medium. The method comprises the following steps: acquiring image data of a vehicle interior trim part; edge features of the front view image are extracted, and detection coordinates corresponding to the welding area are obtained; selecting a target sub-image from the front view image based on the detection coordinates, calculating an average gray value corresponding to each pixel in the target sub-image, and judging whether the average gray value exceeds a preset range; if the average gray value exceeds the preset range, it is judged that the interior trim part has the welding defect. Whether welding appearance abnormity exists or not is judged by calculating the average gray value of the target sub-image; the feature extraction is further performed on the abnormal image through the SURF algorithm, and then the feature is matched with the template features of different defect types through the RANSAC algorithm to identify the defect types, so that compared with a global image feature extraction detection mode, the detection efficiency can be effectively improved.

A grayscale compensation method, apparatus and system, a display driving method apparatus and system, and a chip and a medium, which belong to the technical field of displaying, and aim to improve the uniformity of the display brightness of a display picture. The grayscale compensation method comprises: determining compensation coefficients respectively corresponding to a plurality of sub-pixels in a display panel, wherein the compensation coefficients are used for representing compensation modes and compensation intensities of brightness compensation performed on the sub-pixels, and the compensation modes comprise a mode of increasing the brightness and a mode of reducing the brightness; on the basis of the compensation coefficients and grayscale values to be displayed corresponding to the sub-pixels, determining correction coefficients respectively corresponding to the plurality of sub-pixels, wherein different grayscales to be displayed correspond to correction coefficients of different sizes, and the correction coefficients are used for correcting the compensation intensities; and on the basis of the compensation coefficients and the correction coefficients, compensating for the grayscale values to be displayed of the plurality of sub-pixels, so as to reduce the difference in the luminous brightness of the plurality of sub-pixels.

The invention discloses a brightness compensation method and electronic equipment, and relates to the technical field of electronic equipment, and the method comprises the steps that the electronic equipment obtains corresponding gray scale compensation tables of a display screen in different mura forms; and when the display screen is in the target mura form, the electronic equipment compensates the gray-scale value of each pixel in the display screen according to the compensation gray-scale value corresponding to each pixel in the gray-scale compensation table corresponding to the target mura form, and controls the pixels of the display screen to be displayed based on the gray-scale value compensated by the compensation gray-scale value. In the application, the electronic device can obtain the gray scale compensation table of the display screen in different mura forms, the operation of brightness compensation can act on various mura scenes appearing on the display screen, and the demura operation is performed for the various mura forms, so that the demura effect of the display screen is effectively improved, and the display screen can realize a uniform display effect in different scenes.

The invention provides a self-supervised end-to-end vision reconstruction method and system, and relates to the technical field of computer vision processing. The method comprises the following steps: firstly, acquiring multi-camera parameters and different-view-angle image data, including focal length, lens distortion and main coordinate point data of each camera, and pixel point coordinates of a current frame of a first camera and a reference frame of a second camera, then constructing an end-to-end training model, and calculating a re-projection error of pixel points of the current frame and the reference frame to obtain a multi-view-angle image; and solving the parameter update quantity by using a Gaussian Newton iteration method, iteratively optimizing the camera pose, the pixel corresponding relation and the depth data, and reconstructing a three-dimensional coordinate by combining the obtained data set after the re-projection error is converged, and converting and splicing to realize three-dimensional scene reconstruction. By implementing the scheme, end-to-end self-supervised training can be realized under the condition of not depending on manual annotation, so that three-dimensional visual reproduction is realized.

A driving method for a display device, including: inputting a source image, wherein the source image includes depth information; searching in the source image by utilizing the depth information to determine first image pixel points matched with light-emitting sub-pixels of the display device; assigning pixel gray values of the first image pixel points to the light-emitting sub-pixels of the display device; and controlling the light-emitting sub-pixels to emit light according to assigned pixel gray values, so that the display device displays a multi-viewpoint image corresponding to the source image. The driving method of the display device provided in the embodiment of the present application directly searches for the first image pixel that satisfies the preset parallax condition on the source image based on the depth information, renders the display device, saves hardware storage resources, and improves the processing efficiency of multi-viewpoint naked-eye 3D display.

The invention provides a scene semantic occupancy prediction method based on probabilistic three-dimensional Gaussian, and the method comprises the steps: processing a multi-view camera image through employing a two-dimensional backbone network, generating multi-view and multi-scale image features, predicting the three-dimensional occupancy distribution aligned with pixels through the image features, and initializing the position attribute of the three-dimensional Gaussian according to the distribution; the method comprises the following steps: taking three-dimensional gauss as point cloud data, generating sparse voxel data through voxelization, and processing the sparse voxel data by using three-dimensional sparse convolution to complete feature interaction between the three-dimensional gauss; by using a cross attention mechanism, updating three-dimensional Gaussian features by image features; performing layer-by-layer iterative optimization on each attribute in the three-dimensional Gaussian features by using a Gaussian attribute fine tuning model; and aggregating geometric attributes and semantic attributes of the three-dimensional Gaussian by using a probability Gaussian superposition method, and generating a three-dimensional semantic occupation prediction result of the scene. According to the invention, the problem of efficient adaptive three-dimensional semantic occupancy prediction of an automatic driving scene can be solved.

The invention is suitable for the technical field of image annotation and intelligent transportation, and provides an intelligent image data annotation method and system based on multi-modal semantic fusion, and the method comprises the following steps: obtaining image data, point cloud data, voice data and environment information of the same time and space; performing semantic segmentation on the image data to obtain image segmentation information; performing alignment processing on the point cloud data and the image data to generate a 3D point cloud aligned with image pixels; according to the 3D point cloud, deducing the geometric contour of the shielded target to obtain point cloud geometric information; fusing the image segmentation information and the point cloud geometric information in combination with environment information to obtain image fusion information; and in combination with the voice data, marking visible targets and shielded targets in the image data with behavior semantics according to the image fusion information. According to the method, multi-modal data such as images, voices and point clouds are fused, so that semantic annotation can be accurately carried out on the occluded target in a complex scene.

The invention discloses a semi-supervised medical image segmentation method based on suspicious pixel mutual correction. The method comprises the following steps: 1, preprocessing and dividing a data set; 2, a semi-supervised medical image segmentation network model SPMC is constructed and comprises a network A, a network B, a CDPS module and an SPLAT module, and the network A and the network B adopt the same encoder and decoders with different up-sampling modes; the CDPS screens annotation data prediction results output by the network A and the network B, identifies common trapping pixels which are difficult to predict accurately, and corrects the common trapping pixels by referring to existing tags; the SPLAT screens unlabeled data prediction results output by the network A and the network B, performs pseudo supervision on suspicious pixels in the network B by using credible pixels generated by the network A, calculates total loss and updates parameters; 3, training a network model SPMC; 4, evaluating SPMC performance and optimizing parameters; and 5, inputting a medical image to be segmented into the trained SPMC, and outputting a segmentation result. According to the invention, the prediction accuracy of complex and fuzzy difficult areas in the image is improved.

The invention provides an ultrasonic sound velocity image generation method, ultrasonic equipment and a storage medium. The generation method comprises the following steps: transmitting ultrasonic waves through a probe of ultrasonic equipment, and receiving echoes of the ultrasonic waves to generate echo signals; based on the receiving delay of each preset sound velocity, performing beam forming processing on the echo signal to obtain an enhanced signal corresponding to each preset sound velocity; and based on the pixel value of each first pixel in the respective sound velocity image of each preset sound velocity, determining the pixel value of a second pixel corresponding to the first pixel in the target image to obtain the target image. On one hand, the sound velocity image obtained by the method is high in representativeness, so that the representativeness of the target image can be improved, and the accuracy of the detection result of the detected object can be improved; and on the other hand, the target image can be generated based on different sound velocity images, so that the information breadth of the target image can be improved, and the display flexibility is higher.

The invention relates to the technical field of image processing, in particular to a food production line cleanliness monitoring method and system based on machine vision. The method comprises the steps of obtaining a Lab image of a to-be-detected conveyor belt of a food production line, determining an abnormal score value of a target pixel point in the Lab image, and constructing a data item edge weight of an energy function according to the abnormal score value of the pixel point in the Lab image; determining a periodic displacement amount of a periodic structure of the Lab image in the motion direction of the conveyor belt to be detected, and determining a periodic consistency factor of adjacent pixel pairs according to the periodic displacement amount; according to the method, a data item edge weight of an energy function is established, a smooth item edge weight of the energy function is established according to texture features and a period consistency factor, the energy function is established by using the data item edge weight and the smooth item edge weight, segmentation enabling the energy function to be minimized is performed on the Lab image, and a cleanliness monitoring result can be determined according to the minimum value of the energy function. According to the technical scheme, the cleanliness of the conveying belt of the food production line can be automatically monitored.

A system comprises a tool and a catheter sized to receive the tool. The system also comprises a processor and a memory having computer readable instructions stored thereon. The computer readable instructions, when executed by the processor, cause the system to receive, via an image sensor of the tool, a plurality of images from within a lumen of the catheter. Each image comprises a plurality of pixels. For each image, a subset of the plurality of pixels associated with a viewable feature is identified. For each image, a likely orientation for the viewable feature is determined based on the identified subset of the plurality of pixels. A composite orientation for the viewable feature is determined based on a combination of the likely orientations of the viewable feature for each image. A rotational offset of the tool relative to the catheter is determined based on the determined composite orientation.

To reduce an amount of data to be transferred, while maintaining the number of gradations and graininess.SOLUTION: An information processing apparatus acquires a first image with first resolution, creates a second image with second resolution on the basis of the first image with first resolution, wherein the gradation value of one pixel of the second image indicates the number of dots when an area of the first image corresponding to one pixel is converted to have third resolution higher than the first resolution and the second resolution and is binarized by using a first threshold matrix according to the third resolution, and transmits the second image to an image forming apparatus. The image forming apparatus converts the received second image into a third image that is a binarized image with the third resolution, and creates the third image according to information indicating an arrangement of dots in the third image corresponding to the gradation value for one pixel of the second image, wherein the arrangement of dots corresponding to one pixel of the third image is determined on the basis of a relationship between magnitudes of thresholds in an area corresponding to one pixel in the first threshold matrix.SELECTED DRAWING: Figure 2

A method includes: obtaining original image information, where the original image information includes an original image and brightness level information; obtaining, based on the brightness level information included in the original image information and a brightness threshold, a thresholding segmentation result corresponding to the original image, where the thresholding segmentation result includes area class labels corresponding to pixels in the original image, the area class labels are used to indicate classes of areas corresponding to the pixels in the original image, the areas include a standard dynamic range area and an extended dynamic range area; and performing dynamic range extension on the extended dynamic range area in the original image based on the thresholding segmentation result corresponding to the original image, to obtain an extended dynamic range image corresponding to the original image.

Examples are described for processing images to mask dynamic objects out of images to improve feature tracking between images. A device receives an image of an environment captured by an image sensor. The image depicts at least a static portion of the environment and a dynamic object in the environment. The device identifies a portion of the image that includes a depiction of the dynamic object. For example, the device can detect a bounding box around the dynamic object, or can detect which pixels in the image correspond to the dynamic object. The device generates a masked image at least by masking the portion of the image. The device identifies features in the masked image, and uses the features from the masked image for feature tracking from other images of the environment, masked or otherwise. The device can use this feature tracking for mapping, localization, and / or relocation.

The invention discloses a multi-scale super-pixel-level ground feature classification and recognition method and device, a medium and a product, and relates to the field of image processing. The method comprises the following steps: respectively carrying out principal component analysis on optical and radar remote sensing images to extract principal component feature images; normalizing the principal component feature image, and determining a super-pixel clustering target criterion; performing super-pixel segmentation on the principal component feature image based on a super-pixel clustering target criterion by adopting a plurality of segmentation sizes; the features of texture, polarization and the like are extracted based on the preprocessed optical and radar remote sensing images, the pixel mean value of feature images corresponding to all pixels included in each segmented superpixel is taken as the feature value of each superpixel on different feature images, the extracted multi-dimensional feature value is taken as the input of a machine learning model, and ground feature classification is identified. The adaptability and robustness of superpixel segmentation can be enhanced, diversified classification tasks and complex scenes can be better served, and therefore the precision and efficiency of ground feature classification and recognition are remarkably improved.

The invention discloses a method and device for extracting a centerline of a quasi-cylinder by using a point cloud, and the method comprises the steps: obtaining a quasi-circular target of which the centerline is to be extracted, and carrying out the three-dimensional point cloud scanning, and obtaining the three-dimensional point cloud data; point clouds of a target in the three-dimensional point cloud data are cut at equal intervals in the direction of the center line, and a plurality of annular point clouds with noise are formed; projecting the annular point cloud to a two-dimensional plane in a direction perpendicular to the center line; calculating distances from all point positions in the annular two-dimensional image to external points on the projection image; the probability of occurrence of each distance is calculated, and the position corresponding to the pixel when the probability is the maximum value is obtained as the plane center point of the annular point cloud corresponding to the two-dimensional image; associating z values of the annular point cloud during cutting in the direction of the center line to form three-dimensional coordinates of the center point; and connecting the three-dimensional coordinates of all the central points of the equally-spaced cutting annular point clouds to obtain the central line of the cylinder-like body. According to the invention, the extraction precision of the centerline of the quasi-cylinder and the extraction reliability of the centerline are improved.

The invention discloses a farmland area determination method based on boundary determination and semantic segmentation of a G-Lap algorithm, and the method comprises the steps: collecting images of an agricultural region in multiple scenes through an unmanned plane, and constructing a multi-scene farmland image set; through a G-Lap algorithm, farmland edges in the multi-scene farmland image set are further enhanced to construct a multi-scene farmland training image set; using a multi-scene farmland training image set to carry out complete training of freezing and unfreezing trunks on the improved Deeplab model, and training for multiple rounds of convergence to obtain a final segmentation model; segmenting and extracting farmland area pixels in the test image set by using the trained segmentation model, counting the number of the pixels segmented by the segmentation model through an image binarization method, and obtaining the farmland area according to the number of the pixels in combination with the actual area corresponding to each pixel at different aerial photography heights. The method effectively solves the technical problems that an existing traditional farmland region extraction method is long in consumed time and low in efficiency, and a calculation method is lack of universality.