315 results about "Region selection" patented technology

The invention discloses a texture perception state space modeling method for an image restoration task, and the method comprises the steps: 1, constructing a region selection mechanism based on texture complexity, and enabling the region selection mechanism to be used for distinguishing a flat region and a high-texture region in an image; 2, introducing a texture modulation mechanism, and performing explicit adjustment on a state transition matrix in the state space model; 3, enhancing the context modeling capability of the model through a multi-direction sensing module; and 4, by combining position embedding and a sequence modeling structure, the capability of the model in the aspects of image structure understanding and spatial information maintenance is improved. The method can effectively alleviate the problem of information loss when a traditional image restoration method processes texture details, improves the structure restoration capability of a complex region, gives consideration to the restoration quality and the calculation efficiency, is suitable for multiple image restoration scenes such as image super-resolution, image rain removal, low-light image enhancement and the like, and improves the image restoration efficiency. And the method has good engineering adaptability and actual deployment value.

The invention discloses a non-matching image super-resolution reconstruction method based on multi-scale Hash self-adaption, and the method comprises the steps: firstly obtaining a low-resolution LR image, carrying out the up-sampling of the LR image to a high-resolution LR image, and enabling the LR image and a high-resolution reference image Ref image to serve as a training sample; then constructing a convolutional neural network based on image super-resolution, inputting the training sample into the convolutional neural network to perform image multi-scale feature enhancement, feature decomposition, matching of different-scale LR images and Ref images, feature extraction, multi-scale feature interaction and feature fusion, and obtaining a reconstructed image; and finally, designing a loss function of the convolutional neural network. And inputting the LR image to be subjected to super-resolution reconstruction and the high-resolution Ref image into the optimized convolutional neural network based on image super-resolution to obtain a super-resolution reconstructed image. According to the method and the device, the problem of inaccurate selection of LR and Ref matching corresponding regions in the prior art is solved.

The invention provides a fiber diameter batch processing method. The method comprises the steps of image preprocessing and scale region determination, filtering and region screening, morphological processing and region selection, final processing and radius calculation, and fiber diameter calculation. According to the invention, the scale region is accurately determined in image preprocessing, and the accuracy is improved by multi-feature screening of scales; in the filtering link, median and mean filtering is combined, areas are accurately divided based on filtering differences, noise is effectively suppressed, and meanwhile key information is reserved; during morphological processing, through mutual cooperation of multiple corrosion, expansion and screening operations, a fiber area is gradually and accurately removed, and fibers are accurately stripped from a complex image background; in radius calculation, circle fitting and iterative optimization are carried out through a least square method, and it is ensured that the radius of an inscribed circle is accurate; and finally, according to diameter calculation, pixels and physical sizes are unified through a rigorous formula, batch high-precision processing is achieved, solid data support is provided for fiber characteristic research, quality control and the like, the processing efficiency and precision are greatly improved, and the industry competitiveness is enhanced.

The invention discloses an image processing method, computing equipment, electronic equipment and a storage medium, and relates to the field of image generation and image processing. The method comprises the steps that a region selection instruction acting on a target region on an initial scene image is responded, a mask matrix of the target region and a target description text of a target object are acquired, and the target description text is used for describing the target object; inputting the mask matrix, the target description text and the initial scene image into an image generation model, and generating an initial object scene image by using the image generation model; the initial scene image and the initial object scene image are fused, a target object scene image is obtained, and a target area in the target object scene image comprises a target object. According to the invention, the technical problems of high cost, complex user operation process and poor fusion effect of the target object and the scene when the image for fusing the target object into the scene is generated in the prior art are solved.

The invention discloses an equipment micro-vibration real-time monitoring method and system based on visual enhancement, and belongs to the field of equipment micro-vibration monitoring, and the method comprises the following steps: S1, collecting a high-frame-rate video, and recording collection parameters in real time; s2, constructing a Gaussian pyramid, constructing a Laplacian pyramid through adjacent layer image difference of the Gaussian pyramid, and decomposing the collected video into different spatial frequency bands; s3, designing a band-pass filter to extract a specific frequency band where the equipment vibration signal is located, and enhancing and reconstructing the micro vibration signal; s4, carrying out region selection from the enhanced image frame, calculating an optical flow in a selected region through an optical flow method, and extracting a characteristic frequency of vibration; and S5, performing data visualization, and generating a vibration signal image in real time to display the motion characteristics of the target area. According to the invention, the problems of insufficient sensitivity and poor real-time performance of small vibration detection of equipment in a traditional vibration monitoring method can be solved.

The invention is suitable for the technical field of terminals, and particularly relates to a region selection method, electronic equipment and a computer readable storage medium. In the method, a first image feature of a target image and interaction information for the target image can be acquired, and the first image feature and the interaction information are input into a first area mask generation model for processing to obtain a first mask result corresponding to the interaction information. Besides, the target image and the first mask result can be input into a second area mask generation model for processing, a second mask result corresponding to the interaction information is obtained, and a target area corresponding to the interaction information is determined according to the second mask result. Wherein the second area mask generation model can optimize the first mask result according to the target image and perfect the edge of the first mask result and / or the missing area in the first mask result, so that the target area which the user wants to select can be accurately determined according to the second mask result, and the edge of the determined target area is smooth.

The present invention discloses an infrared and visible light image fusion method based on multi-modal features, including: 1. An encoder-decoder network for extracting multi-modal features; 2. Measuring the multi-modal features using entropy, gradient, and saliency and designing an adaptive loss function; 3. Constructing a fusion weight learning model incorporating an embedded Transformer fusion strategy; 4. Using the saliency map of the infrared image as a label and adding this label as the region selection for optimizing the fusion network; 5. Cascade the fusion weight learning model with the embedded Transformer fusion strategy and the encoder-decoder to construct an infrared and visible light image fusion network and train it. This method utilizes the Transformer to capture the global correlation of multi-scale features, taking into account both local and global information to enhance the overall visual effect of the fused image; and uses a multi-modal adaptive fusion strategy to retain the multi-modal feature information of the image and improve the quality of the fused image.

The invention discloses a power grid tower cross-domain adaptive type selection method, system and device and a medium. The method comprises the steps that tower data and multi-domain environment data are collected; inputting the tower data and the multi-region environment data into a domain adversarial neural network, and constructing a tower-region calculation model; based on the multi-region environment data, obtaining environment parameters of a target region through an LSTM prediction model, and obtaining tower parameters through a multi-objective evolutionary algorithm; and according to the environmental parameters and the tower parameters, generating a three-dimensional tower model, and performing safety analysis on the three-dimensional tower model to generate an optimal cross-domain adaptive type selection scheme. According to the method, multi-region data cross-domain sharing and privacy protection are realized through federated learning and homomorphic encryption, the domain adversarial neural network, LSTM prediction and a multi-objective evolutionary algorithm are combined, and a BIM modeling and knowledge graph reinforcement scheme is matched, so that the cross-region type selection suitability and precision of the power grid tower are improved, the type selection is safe and efficient, and the method is suitable for large-scale popularization and application. The problems that traditional model selection data are limited, and dynamic optimization is weak are solved.

The invention discloses a video signal de-interlacing method and device and electronic equipment, and the method comprises the steps: S100, obtaining a directional correlation interpolation result P directional interpolation of a to-be-interpolated pixel point in a current field, and obtaining a median filtering result P median filtering of the to-be-interpolated pixel point in the current field; and S200, acquiring a motion coefficient M of a to-be-interpolated pixel point in the current field, and selecting an interpolation mode and an interpolation result of the to-be-interpolated pixel point in the current field according to an interval in which the motion coefficient M is located. According to the method, interpolation results of direction correlation interpolation and median filtering are calculated respectively, then motion detection is carried out, a motion area of a pixel point to be interpolated is divided, and then a proper interpolation algorithm is selected according to the divided area, namely, an optimal mode to be interpolated is selected through motion detection. Therefore, the final interpolation suitable for various motion scenes is obtained, and the interpolation mode and the interpolation effect during de-interlacing are greatly optimized.

The present invention relates to the technical field of image recognition, and discloses a multi-commodity automatic calibration recognition method and an intelligent self-service cash register. The steps of the method include: extracting the front image of a commodity from the commodity image; identifying the label area and selecting associated feature points for each label area; segmenting the commodity area from the front image of the commodity; determining the associated commodity area of each label area; extracting the side image of the commodity and determining the associated commodity area of each label area in the side image of the commodity; reconstructing the three-dimensional model of the commodity; scanning the label area to obtain commodity information, and verifying the commodity information based on the gravity sensor and the three-dimensional model of the commodity. The present invention not only improves the accuracy and settlement efficiency of commodity recognition, but also improves the user experience, enhances the security and reliability of the system, and provides a more efficient, convenient and reliable self-service cash register solution.

The invention belongs to the technical field of sleep health monitoring, and particularly relates to an optimal area selection chest / abdomen respiration signal extraction method based on human body pressure distribution. The method comprises the steps that data signals of human body pressure distribution changing along with time during sleeping are collected; the collected original pressure data are preprocessed, and clear and accurate pressure distribution data are obtained; for the preprocessed data, based on calculation of parameters such as the frequency, the power and the signal-to-noise ratio of the chest and abdomen area, an optimal breathing area selection algorithm is used for calculating respective band-shaped areas, most suitable for breathing signal extraction, of the chest / abdomen area; for the selected area, respiratory waveforms are extracted through a pressure signal fusion method based on weighted summation, and thoracic / abdominal respiratory signal monitoring is achieved; the sleep breathing monitoring device can accurately monitor sleep breathing in the form of a non-contact mattress under the condition that sleep quality is not interfered, and has a wide clinical application prospect.

The embodiment of the invention relates to a video generation method and device, equipment and a medium, and the method comprises the steps: responding to a region selection operation for a target reference image, and determining a target region based on the region selection operation; acquiring target track information of the target area moving in the target reference image; wherein the target track information is track information represented by three-dimensional space information; based on the target area, the target reference image and the target trajectory information, generating a motion video of a target object by using a preset target network model; wherein the target object is the target area or an object contained in the target area. According to the embodiment of the invention, the user demand can be well met, the image video effect is enriched, and the video infection is enhanced.

The invention relates to the technical field of slope monitoring, in particular to a loess slope multi-parameter environment coupling monitoring system, which comprises a data acquisition unit; the state judging unit is used for determining the state of the slope body according to the regeneration index of the slope body and the slope angle fluctuation coefficient; the region division unit is used for determining a slope region division mode according to the slope state; the area selection unit is used for monitoring and analyzing each slope body area to determine whether the slope body area is used as a monitoring area or not; the monitoring judgment unit is used for determining a monitoring point setting mode of the monitoring area according to setting conditions; the first setting unit is used for determining an association state of the abnormal sub-region according to the abnormal variation coefficient and an association abnormal region reference value, and determining a processing mode according to the association state of the abnormal sub-region; the second setting unit is used for determining an optimization mode according to the abnormal aggregation region type; according to the invention, the accuracy degree of loess slope data monitoring can be improved.

The invention provides a gene expression analysis method and device, electronic equipment and a storage medium. The gene expression analysis method comprises the following steps: clustering a plurality of capture sites according to spatial transcriptome data of a target slice to obtain a biological function region; generating a region distribution diagram according to the biological function region and the spatial position information of each capture site; receiving area selection operation to determine a target analysis area; determining a biological gradient direction, and dividing the target analysis region into sub-regions according to the biological gradient direction; according to the spatial transcriptome data, obtaining the region expression quantity of the target gene in each sub-region; performing gene expression analysis according to the biological gradient direction and the region expression quantity corresponding to each sub-region; the gene expression analysis is used for describing the variation trend of the gene expression of the target gene in the target analysis region along the biological gradient direction. According to the embodiment of the invention, gene expression analysis can be carried out in the target analysis area based on the gradient direction.

The invention discloses a full-closed-loop control method for a humanoid robot joint module based on an FPGA (Field Programmable Gate Array), which relates to the technical field of robot control and comprises the following steps of: performing second-order active low-pass filtering and programmable gain amplification on analog signals of an encoder, a gyroscope and a force sensor to convert the analog signals into digital signals; a multi-dimensional feature sequence is constructed, temperature signals of a joint driver and an encoder are collected through a temperature sensor interface, a temperature drift model is constructed, and zero drift compensation parameters are obtained; further obtaining a joint elastic deformation error, and applying the joint elastic deformation error to an encoder signal for online offset correction; outputting a reconfiguration region selection signal based on the lightweight neural network; and triggering bit stream local refresh corresponding to the PR Region according to a region selection signal output by the lightweight neural network. The problem that a control loop is difficult to eliminate flexible errors due to the fact that a traditional rigid model or a constant parameter model is difficult to reflect deformation characteristics under the heat-lubrication-load combined action in real time is solved.

The invention discloses a multi-unmanned ship area coverage path planning method considering dynamic collision avoidance, belongs to the field of unmanned surface ship path planning, and aims to solve the problems that a coverage dead zone exists, backtracking cannot be adapted and dynamic collision avoidance is not considered when a traditional DARP area coverage algorithm is adopted to plan a multi-unmanned ship path. The method comprises the following steps: step 1, loading electronic chart data, and extracting static obstacle information in the electronic chart data; rasterization modeling is carried out; 2, according to the initial positions of the m unmanned ships, a task area is allocated to each unmanned ship based on a DARP algorithm; 3, for each task area, selecting a heuristic fusion algorithm with the minimum moving step number and action cost to obtain a static path, and completing full coverage of areas outside static obstacles; and step 4, introducing a DWA algorithm, performing real-time dynamic collision avoidance path planning on the basis of the static path, and when the angle of the advancing direction and the connecting line of the target point is too large, performing re-planning by using an APF algorithm so as to prevent the unmanned ship from falling into local optimum.

The invention relates to the technical field of disinfection devices, and discloses a disinfection device control method, a disinfection device and a computer readable storage medium, the disinfection device control method comprises the following steps: receiving an area selection instruction, and determining a target processing area according to the area selection instruction; receiving a mode selection instruction, and determining a target processing mode according to the mode selection instruction; according to the invention, a user can autonomously select the target area according to actual requirements, so that energy waste caused by idling of a chamber which does not need to be treated is avoided, and the treatment requirements of different spaces can be met in a targeted manner; the target processing mode is accurately matched through the mode selection instruction, so that the disinfection device only executes the corresponding function in the selected area, it is ensured that the processing effect meets the actual requirement, the operation process is simplified, and on-demand processing can be achieved without complex setting. And the disinfection device can be flexibly adapted to diversified use scenes through the control logic of double autonomous selection.

The present application discloses a text recognition method, device, and electronic device, and belongs to the field of text recognition technology. The method includes: a feature extraction module of a pre-trained text recognition model performs deep multi-scale feature extraction on a text image to obtain multi-scale image features; an encoder of the text recognition model performs encoding processing on the multi-scale image features to obtain encoding features; an area selection module of the text recognition model performs text line area selection processing on the encoding features to obtain a first query vector corresponding to the text line; a decoder of the text recognition model performs layer-by-layer decoding processing based on the first query vector and the encoding features to obtain a second query vector; and a plurality of different first prediction heads of the text recognition model predict text-related information on the second query vector to obtain a text information prediction result. This method combines multimodal features such as vision and semantics for decoding, thereby improving the accuracy of text recognition.

The invention relates to the technical field of automobile multi-screen interaction, and discloses a multi-screen interaction method and system applied to an intelligent automobile, and the method comprises the steps: capturing frame images of all display screens, calculating the difference, and selecting an interaction region; information entropy evaluation complexity is calculated, a high-efficiency or low-power-consumption encoder is dynamically allocated based on the complexity, and a matched compression algorithm is selected for encoding; and decoding and updating the received compressed data on the target display screen to realize collaborative response and real-time feedback among multiple screens, and the system comprises an image acquisition module, an interaction area selection module, an information entropy calculation module, an encoder task allocation module, a compressed encoding module, a data transmission module and a synchronous display module. According to the invention, a multi-region dynamic coding scheduling technical scheme based on image difference analysis and information entropy evaluation is adopted, and refined coding processing of image regions with different complexity is realized.

Methods, systems, and devices for wireless communications are described. Generally, the described techniques provide for efficiently determining a region in which a user equipment (UE) is located. The UE may perform measurements on reference signals received from a first network entity to determine the region in which the UE is located. In some cases, the UE may provide the measurements or other information to a second network entity, and the second network entity may indicate the region in which the UE is located to the UE based on the measurements or the other information. The second network entity or the UE may also indicate the region in which the UE is located to the first network entity. The UE and the first network entity may then select suitable beams for communicating based on the region in which the UE is located.

Apparatuses, systems, and methods for determining a field in a path of travel of a vehicle for detection of objects in the path of travel. An apparatus includes a path identifier configured to identify a path of travel for a vehicle. At least one camera is disposed on the vehicle and is configured to capture image data of an area including the path of travel. A region of interest selector is configured to select a region of interest within the image data. A horizon identifier is configured to identify a horizon in the image data. A field determiner is configured to project the horizon onto the region of interest to isolate a field specified by the path of travel and the horizon, the field being evaluatable for the presence of objects in the path of travel of the vehicle.

The invention provides an image enhancement method and device, a storage medium and electronic equipment, and the method is applied to the technical field of computers, and comprises the steps: obtaining a first sample data set which comprises a first sample image, a first sample text and a sample region coordinate, performing model training on the initial region selection large model based on the first sample data set to obtain a region selection large model, and obtaining a second sample data set which comprises a second sample image, a second sample text and a sample clear image, and performing model training on the initial image enhancement large model based on the second sample data set to obtain an image enhancement large model, and performing image enhancement processing based on the region selection large model and the image enhancement large model. According to the method, the local image can be determined through the area selection large model, the image enhancement large model is adopted to perform enhancement processing on the local image, the pertinence of image enhancement is improved, and the operation amount of image enhancement is also reduced.

The invention discloses an energy consumption optimization path planning method and system for unmanned aerial vehicle routing inspection, and relates to the technical field of unmanned aerial vehicle routing inspection path planning, and the method comprises the steps: calculating an evaluation function during node expansion in each step: a heuristic function represents the accumulated cost from a starting point to a current node, and a heuristic function represents the estimated cost from the node to a target point; the heuristic function is composed of a distance component and an energy consumption component, and the expression of the heuristic function is as follows: the linear distance from the current node to the target node, the minimum additional energy consumption estimated according to the height change and the flight resistance, and the weight coefficient; the improved heuristic function can preferentially avoid a high-energy-consumption area, a flatter and smoother air route is selected, and accumulated energy consumption and flight time of a global path are remarkably reduced; meanwhile, due to the introduction of the multi-dimensional cost, the global planning is more accurate and reliable, and the search efficiency of the algorithm in a complex three-dimensional environment is improved.

The application relates to the technical field of data processing, and discloses a radar real-time data transmission processing classification identification method and system.The method comprises the following steps: performing track mode identification on narrowband data of a target radar to obtain a real-time motion mode category of the narrowband data; performing directional region selection on synchronous wideband echo data of the target radar based on the real-time motion mode category to obtain a signal segment of the synchronous wideband echo data; performing sparse decomposition on the signal segment to obtain compact features of the signal segment; performing motion mode compatibility quantification on the compact features based on the real-time motion mode category to obtain category scores of the compact features; performing feedback adjustment on the real-time motion mode category based on the category scores and the compact features to obtain an optimized motion mode of the real-time motion mode category; and performing confidence fusion judgment on the optimized motion mode to obtain a classification identification of the real-time motion mode category; and the application can improve the efficiency of radar real-time data transmission processing classification identification.

The invention discloses a wide-angle video correction method and device based on area selection, relates to the technical field of network wide-angle cameras, and mainly aims to improve the definition of key details in a video so as to enhance the safety of remote operation. According to the main technical scheme, an area video viewing request is obtained, the viewing request comprises view angle configuration parameters of an area, and the view angle configuration parameters specify the coordinates and the size of the area to be intercepted in an original video stream of a wide-angle camera; intercepting a to-be-intercepted area from an original video stream output by the wide-angle camera according to coordinates and sizes in the view angle configuration parameters to obtain a to-be-corrected video area; extracting a local mapping table corresponding to the to-be-corrected video area from a preset global distortion correction mapping table according to the view angle configuration parameters; performing real-time distortion correction on the to-be-corrected video region by using the local mapping table to generate a corrected video stream; and outputting the corrected video stream through a network interface of the wide-angle camera.

The present disclosure provides an image processing method, a system, an electronic device, a storage medium, and a program product. A method comprises: in combination with image content of a target image, performing point sampling on the target image to obtain sampling point information; performing full image segmentation on the target image on the basis of the sampling point information; displaying, on a client interface, the target image which has undergone full image segmentation, the target image which has undergone full image segmentation comprising multiple segmented selectable regions; and in response to a region selection operation triggered by a user for the multiple selectable regions, performing image matting of the target image on the basis of at least one selected region, to obtain a matted image. By using the present solution, the segmented target image can be adaptively changed on the basis of the image content of the target image, which can improve the full image region division effect of the target image, and balance the time consumption of functions; moreover, interaction between the target image which has undergone the full image segmentation and the user can be used to clarify an actual intention of the user on the basis of the region selected by the user, so that excessive matting or missing matting is avoided, and better image matting service can be provided for the user.

The embodiment of the invention discloses a virtual object control method and device, a terminal and a storage medium, and belongs to the technical field of computers. The method comprises the following steps: displaying a master control virtual object in a first orientation in a virtual scene; in response to a first trigger operation on the perspective skill control, displaying a perspective area selection control; in response to an area selection operation on the perspective area selection control, controlling the master control virtual object to change from the first orientation to a second orientation, the second orientation being the direction of the perspective area indicated by the area selection operation; and controlling the master control virtual object to release a perspective skill in the perspective area in the virtual scene, wherein the enemy virtual object in an invisible state in the perspective area has a perspective display effect. By adopting the scheme provided by the embodiment of the invention, the perspective skill release flexibility can be improved, and the display form of the virtual object in the virtual game is enriched.

The invention discloses a physiological signal extraction method based on self-adaptive ROI and ViM dual paths, which comprises the following steps: constructing a first quality perception space-time diagram of a first user and a second quality perception space-time diagram of a second user based on a self-adaptive region-of-interest selection mechanism and a low-rank tensor completion method; extracting a first time sequence prediction signal of the first quality perception space-time diagram and a second time sequence prediction signal of the second quality perception space-time diagram through a time path of the ViM dual-path network structure, and extracting a first frequency domain prediction signal of the first quality perception space-time diagram and a second frequency domain prediction signal of the second quality perception space-time diagram through a frequency path; and fusing the first time sequence prediction signal and the first frequency domain prediction signal to obtain a first physiological signal of the first user, and fusing the second time sequence prediction signal and the second frequency domain prediction signal to obtain a second physiological prediction signal of the second user. According to the method, the robustness, precision and efficiency of physiological signal extraction are improved, and different scene requirements can be considered.

1. Name of the Design Product: Graphical User Interface for Item Selection of an Electronic Device. 2. Use of the Design Product: For an electronic device. 3. Design Key Points of the Design Product: Lies in the content displayed on the user operation interface. 4. Picture or Photograph Best Illustrating the Design Key Points: The front view of Design 1. 5. The other views are of conventional designs, and the rear view, top view, bottom view, left view, and right view are omitted. 6. Design 1 is designated as the basic design. 7. Use of the Graphical User Interface: The graphical user interface is used to select items in a store. In the front view of Design 1, when the user obtains selectable items, it enters the first state change diagram of the Design 1 interface. In the first state change diagram of the Design 1 interface, click on the middle area to select an item and enter the second state change diagram of the Design 1 interface. In the second state change diagram of the Design 1 interface, click on the "Confirm Selection" button at the bottom to enter the third state change diagram of the Design 1 interface. In the third state change diagram of the Design 1 interface, click on the second button from left to right at the lower left corner to enter the fourth state change diagram of the Design 1 interface, and the selected item is displayed at the lower right corner. In the front view of Design 2, when the user obtains selectable items, it enters the first state change diagram of the Design 2 interface. In the first state change diagram of the Design 2 interface, click on the middle area to select an item and enter the second state change diagram of the Design 2 interface. In the second state change diagram of the Design 2 interface, click on the temporarily retract button at the lower right corner to enter the third state change diagram of the Design 2 interface, and an exclamation mark icon will appear for prompting.

A snowfield passenger flow statistical method based on deep learning relates to the technical field of crowd flow statistics in dense areas such as snowfields, and comprises the following steps: segmenting an image into different areas according to the difference of crowd density in the image, adopting a multi-branch structure, and carrying out crowd counting on the corresponding density area by each branch; for an area with relatively low density, accurate counting of the area can be realized if multi-scale information is extracted; a branch model aiming at low density is designed, the core of the branch model is that the receptive field range of neurons is expanded by using expansion convolution, and meanwhile, multi-scale information is obtained through aggregation by using Dense connection, so that accurate estimation of a multi-scale head is realized; for a high-density area, a branch model which is based on Transform and aims at high density is provided. According to the invention, the final density map is obtained by selecting the most suitable branch result for each area according to the local density information of the density map generated by double branches, so that high-accuracy counting is obtained.