14results about How to "Increased computational burden" patented technology

The invention relates to the technical field of scene recognition, in particular to a real-time semantic segmentation method and system for a multi-shape pyramid in a traffic scene, and the method comprises the steps: initializing an original to-be-segmented image, and obtaining an initial segmentation feature map; channel splicing is carried out on the initial segmentation feature map and an output feature map obtained after first downsampling is carried out on the original to-be-segmented image, downsampling is carried out after a fused feature map is obtained, detail information is extracted, and a detail information feature map is obtained; down-sampling the detail information feature map and extracting semantic information; the method comprises the following steps: extracting transverse and longitudinal receptive field information of a semantic information feature map, performing expansion convolution to obtain feature maps of different receptive fields, obtaining a multi-scale and multi-shape context information feature map, performing up-sampling, performing point-by-point convolution dimension reduction on a detail information feature map, and performing channel splicing on the two features to obtain a fusion feature map containing semantic and detail information. And adjusting the channel number into a prediction category number, and normalizing to obtain a prediction segmentation result. The method has good segmentation precision and reasoning speed.

The invention discloses a circuit board welding defect detection method based on edge detection and target identification, belongs to the technical field of welding defect detection, and solves the problems of missing detection, false detection and inaccurate positioning caused by defect superposition, shielding and complex production environment interference in a circuit board welding defect detection process. According to the invention, through circuit board image edge extraction based on adaptive median filtering, the identification and retention effects of welding defect edge information are enhanced; a multi-level cross feature fusion network for anti-occlusion defect feature extraction is constructed, and through an FFE module and context information extraction based on RFB-Net and cavity convolution, the anti-interference analysis ability of the model to defect features is enhanced; according to the detection frame positioning optimization method based on the weighted repulsive force loss function, the reliability and efficiency of circuit board welding detection are improved while the calculation burden is reduced; the method is suitable for circuit board welding defect detection in multi-process continuous production, and can realize high-precision defect positioning and identification in a complex environment.

The invention discloses a dynamic backlight regulation and control method and system for a display device, and the method comprises the steps: forming a grayscale image according to the feature value of each pixel point in a current to-be-displayed original input image; inputting the grayscale image into a preset convolutional neural network encoder model to obtain a compressed feature map; inputting the compressed feature map into a preset coupling diffusion algorithm decoder model to obtain an updated image; adjusting and controlling the backlight of the display based on the gray value of each pixel point of the updated image; according to the method, the neural network method is adopted to extract the compression features of the image backlight, the adaptability of the algorithm to different display scenes is improved, the operation in the expansion stage is not too complex through the coupling of bilinear interpolation and nearest neighbor interpolation in the decoder stage, the relatively reliable and smooth image effect is provided, and the method is suitable for large-scale popularization and application. A ping-pong design and a pipeline architecture design are adopted for video stream processing, the degree of parallelism and the data throughput rate are increased, and the data processing efficiency is improved.

The invention relates to a remote sensing image building extraction method based on deep learning and contour regularization, belongs to the technical field of remote sensing image data mining based on deep learning and a building post-processing algorithm, and aims to solve the problem that a building contour extracted by deep learning automation is greatly different from an actual contour. Compared with the prior art that the building contour is directly extracted from a deep learning model, the building contour extraction method has the advantages that subsequent post-processing operation is added, and the technical problem that the extracted building contour is greatly different from the actual contour is solved; according to the method, building distribution vectors with regular outlines and high applicability can be obtained under different image scales, and the method has the advantages of high efficiency, accuracy, high robustness and the like.

The embodiment of the invention discloses a video coding method and device, electronic equipment and a computer readable medium. A specific embodiment of the method comprises the following steps: acquiring a video; for each frame of image, determining an average gradient value; in response to the fact that the average gradient value is not lower than a first threshold value, determining a motion vector precision mode as a first mode; in response to the condition not higher than the first threshold and not lower than a second threshold, determining that the motion vector precision mode is a second mode; in response to not higher than the second threshold, determining the motion vector precision mode as a third mode; obtaining each coding unit; determining a self-adaptive motion vector precision tool; setting a motion vector precision range; performing encoding processing on each encoding unit to generate encoding information; and determining each piece of coding information as video coding information. According to the embodiment, the overall characteristics of the image are utilized, so that the selection range of the motion vector precision is more reasonable, the coding efficiency is improved, the consumption of computing resources is reduced, and the overall performance of video coding is improved.

The present invention relates to a SOC estimation method that periodically considers battery capacity decay, and the method includes: establishing a battery model, determining battery model parameters based on battery aging experimental tests and data analysis, battery aging feature extraction and correlation analysis; battery capacity and model parameter estimation based on a PSO-LSTM algorithm, SOC estimation based on a CHIF algorithm, and online co-estimation of battery capacity and SOC using a joint estimation framework; the present invention adopts the CHIF algorithm to improve estimation accuracy and adapt to nonlinear systems. By transferring state volume points in a nonlinear system, CHIF can suppress non-Gaussian noise to the greatest extent. The PSO de-optimization of LSTM hyperparameters used in the present invention can effectively improve prediction accuracy. The present invention considers battery aging factors through the PSO-LSTM algorithm, updates the battery capacity in real time, and then estimates the SOC of the battery in combination with the CHIF algorithm, which can effectively improve the estimation accuracy of the battery SOC.

The invention belongs to the technical field of power distribution network operation management, and particularly provides a scene reduction method and system for a multi-dimensional operation scene of an active power distribution network, a terminal and a storage medium. Respectively generating a corresponding load power scene, a photovoltaic output power scene and a wind power output power scene; distributing a corresponding occurrence probability for each generated scene; combining the scenes generated by different factors based on the occurrence probability of each scene to obtain an initial scene set; after a large number of initial scenes are generated, clustering analysis is performed on the scenes by using a preset clustering algorithm, so that typical output working conditions capable of representing original data features are screened out to serve as a reduced scene set. According to the process, the number of scenes is effectively reduced, the complexity of subsequent optimization calculation is reduced, and the calculation efficiency is improved.

The invention discloses an insulator fault detection method and device based on deep learning and adaptive image processing, and the method comprises the steps: firstly constructing an insulator data set, and training a DetectoRS target detection network and a Bi SeNet semantic segmentation network on a training set; inputting the test set image into a DetectoRS network to generate an insulator detection frame; extracting each insulator according to the detection frame, and inputting an extraction region into a Bi SeNet network to generate an insulator mask; post-processing the image in the semantic segmentation result image to obtain the outer contour of the insulator; calculating the minimum external rotation rectangle of the outer contour of the insulator, and rotating the image to be horizontal according to the rotation angle of the minimum external rotation rectangle; constructing a quantization curve of the shape structure of the insulator according to the horizontal insulator image; and finally, the curve is smoothed, a self-adaptive threshold value required by fault judgment is constructed, the existence state and the position of the insulator fault are judged, and the accuracy and the universality of fault detection are effectively improved.

The invention relates to a safe homomorphic k-means clustering method based on a Paillier cryptosystem, a system applied by the method comprises a cloud server, a computing service provider and a plurality of user sides, each user side is registered on the cloud server and is in communication connection with the cloud server, and the cloud server is in communication connection with the computing service provider; the method comprises a data preprocessing process, a safe K-means clustering process, an iterative updating process and a result decryption process. The invention provides a practical privacy protection data clustering solution, and promotes the development of a privacy protection data mining technology.

The invention discloses an electron microscope image drift distance rapid calculation method based on Fourier transform, and the method comprises the steps: obtaining two electron microscope images with the same size, and naming the images as sum; respectively carrying out Fourier transform on the image sum to obtain a phase difference between the image and the frequency domain; fourier transform is carried out on the phase difference, and the drift distance of the image is determined according to the distance between the diffraction points in the frequency domain. According to the method, through the Fourier transform method, the problems of high calculation complexity and low processing speed in the prior art are effectively solved, and the calculation efficiency can be remarkably improved while the precision is ensured.

The application discloses a monocular hand-eye robot vision servo method for three-dimensional space dynamic depth estimation, comprising establishing a parameterized vision servo system model, constructing an estimation error driven adaptive law and executing position-based vision servo control. The application establishes a parameterized model of the robot-camera-target unification, estimates the depth changes caused by the target motion and the robot motion in the same framework, which enables the application to effectively deal with the more common and more complex working conditions that both the robot and the target are moving. The application solves the problem that the nonlinear time-varying depth is difficult to measure in the monocular camera vision servo without prior geometric knowledge of the observed object and under the condition that both the target and the robot are moving, and realizes the rapid convergence of the control error and the estimation error in the hand-eye robot system.

The present invention discloses a reliable target tracking method and system in a rapid convergence process. The method constructs a tracking model of the rapid convergence process between a drone and a target, uses the tracking model to obtain a rough estimate of the target scale and position, and fine-tunes the rough estimate of the target scale to obtain the final target scale and position, thereby realizing reliable tracking of the target in the rapid convergence process, and enabling the drone to accurately perform tasks. The method provided by the present invention solves the problem that when a drone quickly approaches a target, the target scale rapidly expands in the drone's field of view, causing tracking failure, and can accurately track the target and perform landing or strike tasks. The method provided by the present invention is flexible to use, has strong versatility, and is fast to calculate.

A meshless large deformation simulation method using displacement-pore pressure non-uniform support domain interpolation is disclosed, belonging to the field of numerical simulation technology for civil engineering and hydraulic engineering. The steps are as follows: S1, read in the node coordinates, background mesh information, material information, and load information of the meshless model; S2, generate Gaussian points within the background mesh, calculate the displacement support domain and pore pressure support domain of the Gaussian points, and search for meshless nodes within each support domain; S3, calculate the displacement shape function and pore pressure shape function of the Gaussian points based on the nodes within the displacement and pore pressure support domains, and determine the displacement and pore pressure at each Gaussian point; S4, calculate the matrices required in the meshless governing equations; S5, conduct meshless large deformation analysis based on the shape function matrix and strain matrix. This invention improves simulation accuracy by calculating displacement and pore pressure shape functions through nodes within different support domains, and is of great significance for simulating large deformation of soil weakening and liquefaction in practical engineering.

The invention discloses an unmanned aerial vehicle image detection confrontation sample generation method and system based on a black box normal form. The method comprises the steps of S1, candidate region generation: extracting multi-scale features from an input unmanned aerial vehicle image, and generating a candidate region set containing potential target frame coordinates through feature fusion and transformation; s2, target category retrieval and allocation: allocating a target category with the highest attack efficiency to each candidate region in the candidate region set based on pre-constructed object category correlation prior information to form a target category set; and S3, adversarial sample generation: combining the candidate region set with the target category set, generating a virtual instance set, and adding the virtual instance set into the original input image to obtain a final adversarial sample image. According to the method, efficient black box attacks are realized, calculation delay and energy consumption are innovatively induced, the attack pertinence is high, the concealment is high, the attack success rate is improved, the adaptability is high, and a new view angle is provided for security assessment and defense.