2681results about How to "Reduce time complexity" patented technology

The invention discloses a robot indoor positioning and navigating method based on vision, belonging to the field of vision navigation. According to two-dimension codes, the method designs a manpower road sign which is simple and convenient, is easy to recognize, internally contains absolute position coordinates and has certain error correction capability; the road sign is arranged on a ceiling; a camera arranged on the robot, with the optical axis vertical to the ceiling, is used for photographing; the coordinate information included in the road sign is analyzed by road sign positioning of a series of steps such as threshold segmenting of images, extracting of communicating region, profile curve matching and recognizing of road sign characteristic; and the current absolute position and course angle of the robot can be finally obtained by the position estimation arithmetic of the robot. The method has the advantages of reducing the blocking and noise interference, and greatly reducing the processing time and the complexity of surrounding environments of the image processing method.

The invention provides an SAR image terrain classification method based on a depth RBF network. The method mainly solves the problem of the prior art that the accuracy of classification is low. The method comprises the steps of (1) extracting the texton features of an SAR image; (2) training the texton features of the SAR image through a first-layer RBF neural network of the depth RBF network to obtain the advanced features of the image; (3) training the advanced features through a second-layer sparse autocoder network SAE of the depth RBF network to obtain more advanced features of the image; (4) training the more advanced features through a third-layer RBF neural network of the depth RBF network to obtain the terrain classification features of the image; (5) comparing the terrain classification features of an image test sample with a test sample label, adjusting the parameters of each layer of the depth RBF network, and obtaining an optimal test classification accuracy. The method is high in classification accuracy and can be used for complicated image classification.

The invention discloses a video flame-detecting method based on multi-features fusion, which is characterized in that: a fast Gaussian hybrid model is established and a pixel counter is arranged when a computer reads a first frame image a monitoring site captured by a colored camera; when the computer reads new images of a second frame and the following frames, a fast Gaussian hybrid method is firstly used for carrying out motion detection so as to extract a moving object and then a flame color decision is carried out to the moving object to obtain an alternative flame area; and finally, flicker analysis is carried out on the alternative flame area so as to identify a real flame object. The video flame-detecting method carries out modeling to flame color characteristics, motion characteristics and flickering characteristics in flame areas; various flame detection interfering resources are gradually and rapidly eliminated in accordance with the recognition sequence of motion, color and flickering. The test results show that the method has relatively strong robustness; in addition, based on an AMD 2.04GHz processor, the processing speed can reach 22 frames per second for video images with pixels of 320 multiplied by 240.

The invention discloses a method and a system for joint resource allocation and computing unloading in a software defined vehicle-mounted edge network. The method comprises the following steps: establishing a mathematical model of a wireless communication theory according to data information, and modeling a joint resource allocation and calculation unloading problem as a mixed integer nonlinear programming problem (MINLP) according to the mathematical model; decomposing an original problem into a resource allocation sub-problem and a calculation unloading sub-problem through a Tammer decomposition method, and adopting function monotonicity definition, Lagrangian duality and KKT (Karush-Kahn-Tucker) to obtain optimal resource distribution, and obtaining an optimal unloading strategy by adopting a multi-stage low-complexity heuristic algorithm; wherein the server can allocate computing resources to the vehicle according to the scheme of the controller, and the vehicle selects the serverto unload the computing task according to the strategy of the controller. According to the invention, an optimal unloading strategy and an optimal resource allocation scheme are provided for vehicles,and the total processing delay of vehicle-mounted tasks in a system range is reduced.

The invention discloses a method for quick vehicle type recognition under video monitoring mode, which is implemented according to the following steps: setting up a road monitoring device and classifying the vehicles into cars, taxis marked by special color, microbuses, middle-sized vehicles, public buses and large trucks; step 1, initialization, conducting training to the video monitoring device; step2, extracting the area of the vehicle target area and the length and the width of the boundary rectangle of the area, constructing corresponding characteristics and roughly classifying the vehicles into minitype vehicles, middle-sized vehicles and large vehicles; step 3, extracting the main colors of bodyworks of a plurality of minitype vehicle targets to recognize the taxis, then extracting the characteristic parameter of window position of the minitype vehicles, and furthermore, determining the microbuses or cars; and step 4, extracting the characteristic parameters of roof brightness and roof texture to determine whether the large vehicles are public buses or not. Through the method, a variety of vehicles can be recognized quickly and accurately according to the classification characteristics.

The invention discloses a short text query expansion and indexing method based on a word vector. The short text query expansion and indexing method particularly comprises: A, pretreatment of corpus information of a short text; B, expression of every word in a corpus dictionary by the word vector through a training model; C, query extension; D, obtaining of a text candidate set through a query expansion word set and a BM25 index model; E, extraction of subject of the short text; F, calculation of the text vector of the short text; G, re-sequencing of the short text returned by a traditional indexing model. The short text query expansion and indexing method can more exactly and effectively satisfy the indexing demand of a user; moreover, the query expansion module can find out words capable of expressing user's intension according to the existing data so as to perform the query expansion.

The invention discloses a method for detecting and classifying all-network flow abnormity on line. The method comprises the following steps of: (I) acquiring network flow (NetFlow), namely, receiving a NetFlow data packet transmitted from a border router by adopting a NetFlow collector, resolving the data packet and aggregating data streams to form data suitable for statistical analysis, and transmitting the data to a central control board through network to store in a database; (II) building a flow matrix taking the entropy of flow characteristics as measure; (III) detecting the flow abnormity on line by adopting a main increment component analyzing method; and (IV) constructing sample points in four-dimensional space by utilizing residual vector acquired through on-line detection and classifying the flow abnormity on line by adopting an increment k-mean value clustering method. The method has the advantages of detecting the flow abnormity on line, classifying the flow abnormity on line in real time, meeting the requirement on the real-time detection and classification of the flow abnormity better and laying the technical foundation for subsequently defending against network attack, along with lower time complexity and storage expenditure.

The invention discloses a human face video processing-based heart rate detection method. The method is implemented by the steps of (1) controlling a camera to acquire a video that the camera is over against a human face for 10s; (2) performing human face characteristic point detection and tracking on the acquired video, and performing tilt correction to obtain a stable human face video; (3) performing color space conversion on the human face video; (4) extracting a clean BVP (Blood Volume Pulse) signal from the human face video by utilizing video color amplification; and (5) performing iterative frequency domain analysis on a BVP and computing the frequency of the BVP signal to estimate a heart rate. According to the method, the interferences of a background and a rigid head motion are eliminated by utilizing human face characteristic point detection and tracking, the clean BVP signal is extracted through the video color amplification, and the iterative heart rate estimation is carried out, so that the accuracy of heart rate detection is improved and very good robustness is ensured; and the method can be used for non-contact heart rate detection in a real scene under cooperative and non-cooperative conditions.

The invention discloses an intra-frame lossless compression coding method based on an HEVC (high efficiency video coding) frame, wherein the method is low in time complexity and high in prediction accuracy and increases compression ratio. By the aid of the HEVC frame, a coding unit can be adaptively divided according to image flatness, the best division scheme is determined, and the optimal prediction mode is selected. DPCM (differential pulse code modulation) differential codes are used for preliminary prediction, prediction errors are possibly large when only the differential codes are used for an area with complex texture, and accordingly, prediction values are corrected for the area by means of error compensation. By combining the two methods, reduction of time complexity and improvement of prediction accuracy can be effectively compromised. Distribution characteristics of residual errors depend on the prediction mode, so that the scanning sequence of the residual errors is determined according to the prediction mode, and entropy coding can be more effective.

The invention relates to a multi-document auto-abstracting method facing to inquiry, which comprises the following steps: performing preprocessing on the inquiry and documents; performing topic segmentation and semantic paragraph clustering on the preprocessed documents to obtain subtopics; expressing the inquiry and the sentences in each of the subtopics in the form of a word frequency vector, and calculating the correlation measurement of the inquiry and the subtopics; screening the subtopics according to the correlation measurement of the inquiry and the subtopics, sequencing the subtopics according to the importance of the subtopics, and selecting the front T important subtopics to obtain an ordered sequence of the subtopics correlative with the inquiry; and circularly obtaining representative sentences from the subtopic sequence in turn, and connecting the representative sentences together to generate an abstract. The method uses the topic segmentation technique so that the abstract is in a limited length range and comprises the important information in a document set as much as possible, provides more targeted services, can adjust the content of the abstract according to a user inquiry topic, and can achieve the interactions with users.

The invention discloses a face recognition method of a deep convolutional neural network, which reduces the time complexity, and enables a weight in the network to still have a high classification capacity under the condition of reducing the number of training samples. The face recognition method comprises a training stage and a classification stage. The training stage comprises the steps of (1) randomly generating a weight wj between an input unit and a hidden unit and an offset bj of the hidden unit, wherein j equals to 1,...,L and represents the number of the weight and the offset, and the total number is L; (2) inputting a training image Y and a label thereof, by using a forward conduction formula hw, b(x)=f(W<T>x), wherein hw, b(x) is an output value, x is input, and an output value hw, b(x<(i)>) of each layer is calculated; (3) calculating the offset of the last layer according to a label value and an output value of the last layer; (4) calculating the offset of each layer according to the offset of the last layer, and acquiring the gradient direction; and (5) updating the weight. The classification stage comprises the steps of (a) keeping all parameters in the network to be unchanged, and recording a category vector outputted by the network of each training sample; (b) calculating a residual error delta, wherein delta=||hw, b(x<(i)>)-y<(i)>||<2>; and (c) classifying a tested image according to the minimum residual error.

The invention provides an incremental data acquisition method and an incremental data acquisition system; wherein, the method comprises that: a relational database snapshot list corresponding to a target list is created according to the structure information of the target list; extraction of snapshots in the target list with time points appointed is completed with the created snapshot list; and snapshot intercomparison is performed according to the snapshots with different time points so as to capture incremental data. The invention has the advantages that the snapshot list is adopted to store snapshots, and maintenance of metadata information of the incremental data is not needed, thus versatility is enhanced; furthermore, intercomparison time of snapshot record is shortened, computer system computation is decreased, and computer system resources are saved through creating the cluster index of the snapshot list.

The invention relates to a kinect-based action training method, which can realize human body action extraction without marking points by acquiring human body action date through kinect, so that the action training is simpler and more practical. The kinect-based action training method is divided into two modes: on-line training and off-line training, wherein during the on-line training, the trained actions can be automatically divided into plurality of stages, a user can learn the actions by tracing the prompts of each stage, whether the user completes the actions of each stage or not is judged by kinetic energy and potential energy of the actions of the user, and important key points of the actions of the stage are automatically computed and are prompted to the user; and during the off-line training, the user firstly completes all the action by self, and the actions of the user can be automatically matched with standard actions by being caught by the kinect, comparison analysis can be carried out according to the matched actions so that the marking can be carried out according to the difference of the actions of the user and the standard actions on skeleton direction, and thus, the user can more intuitively find the difference.

The invention discloses a method and a system for segmenting building roof patches from large-scale image dense matching point clouds, and aims to solve the problem of how to accurately acquire building roof three-dimensional patches from color three-dimensional point clouds generated from inclined images. The method for segmenting the building roof patch from the large-scale image dense matchingpoint cloud comprises the following steps: ground filtering and vegetation filtering; removing the ground and vegetation in the image point cloud; carrying out building monomerization; removing non-building points, and realizing building clustering; removing the vertical surface of the building; segmenting, repairing and refining the roof surface patches.

The invention discloses an image super-resolution reconstruction method based on sparse representation and deep learning, and solves the problems that the image super-resolution process is complex incalculation and the quality of a reconstructed image is poor. The method comprises the following implementation steps: collecting and extracting training data blocks and a chromaticity and brightnessdictionary for combined optimization training; independently reconstructing a high-resolution image block; Carrying out high-resolution image reconstruction of sparse representation; Training a residual error network based on deep learning to optimize high-frequency details; Image super-resolution reconstruction. In order to prevent an edge effect and a fuzzy effect, chroma and brightness data aredistinguished and independently reconstructed; in order to optimize high-frequency detail information of a sparse representation output high-resolution image, the high-resolution image based on sparse representation reconstruction is input into a residual network, and a high-frequency residual image is output through four times of convolution feature extraction and feature fusion and input bitwise addition to reconstruct a super-resolution image. The method is low in calculation complexity, high in image reconstruction quality and widely applied to the fields of remote sensing monitoring, criminal investigation, traffic management and the like.

The invention discloses a method for remapping resource demand dynamic change based on a virtual network, relating to the internet technical field. The method comprises the following steps of: establishing a physical network model, a virtual network model and a physical network available resource model; according to the physical network model and the virtual network model, establishing the original virtual network mapping model; according to the physical network model and the original virtual network mapping model, determining available physical network resources; for the resource demand dynamic change of the virtual network, clearly determining a virtual node and a virtual link to be remapped; determining an order for joining the remapping process of the virtual node to be remapped; according to the order, completing the node mapping and the relative link mapping to the virtual node to be remapped; and updating the available physical network resources. The method disclosed by the invention aims to achieve an optimization purpose of minimum resource consumption, and therefore, the resource cost of each virtual network in the remapping process can be minimized so as to accord with the actual situation.

The invention discloses a polarimetric SAR classification method based on a semi-supervised convolutional neural network. According to the method, first, image data is input, and super pixel segmentation is performed; training samples and test samples are extracted; a neighborhood preserving and sparse filtering deep learning network (NDSFN) is adopted to perform unsupervised pre-training on parameters of the convolutional neural network, and a needed filter set is obtained; a feature map is obtained through convolution operation; fuzzy processing is performed on the feature map through downsampling; parameter settings of the NDSFN are adjusted according to the filter size in a second convolution layer to obtain a new feature map; a Softmax classifier is utilized to perform image classification on the training samples; and a small quantity of marked samples are adopted for slight adjustment to obtain a polarimetric SAR terrain classification. Through the method, by constructing a novelsemi-supervised convolutional neural network (SNCNN) model, the problem that a traditional supervised convolutional neural network has a high requirement for marked samples is effectively solved.

The invention discloses a low-complexity polarization code decryption SCL algorithm based on segmented verification assistance, and the algorithm selects a parity check code, repeatedly uses the parity check code in a decryption process, and achieves the performances of SCL-CRC24. Moreover, compared with a conventional scheme, the algorithm is better in low signal to noise ratio anti-noise performance and error rate. In addition, the spatial complexity of the algorithm is lower than that of the SCL-CRC24, the time complexity is greatly reduced, and the decoding speed is greatly improved. Compared with a CRC-24 verification algorithm sacrificing multiple information bits, the algorithm employs a parity check method, enables verification elements to be distributed in the information bits, is repeatedly used in the decoding process, and is lower in time complexity than the prior art.

The invention discloses an unmanned aerial vehicle routing inspection image retrieval system and method based on an electric transmission line and a GIS. The unmanned aerial vehicle routing inspection image retrieval system comprises an image processing server, an external signal input device, a result display device, an unmanned aerial vehicle and a ground monitor station. The external signal input device inputs external signals or instructions to the image processing server, the result display device outputs a processing result of the image processing server, the ground monitor station reads the flight state information and routing inspection result data of the unmanned aerial vehicle and sends the read data to the image processing server, and the ground monitor station is in wireless communication with the image processing server. By means of the unmanned aerial vehicle routing inspection image retrieval system and method based on the electric transmission line and the GIS, the data retrieving and processing efficiency of the routing inspection result can be improved, the image processing cost in routing inspection is lowered, the image retrieving and processing speed of the server is increased, and the processing time cost of the routing inspection result is lowered.

The invention discloses a knowledge map construction method for medical images, belonging to the field of knowledge maps. The construction process includes a step of knowledge representation with a framework theory representation method, a step of knowledge acquisition, wherein the knowledge sources of entity, attribute and attribute value extraction is unstructured data, a step of knowledge fusion with the integration of acquired new knowledge and the elimination of ambiguity, a step of knowledge processing with the knowledge reasoning and quality assessment of the data which is subjected tothe knowledge fusion and the adding of qualified data to a knowledge map, and a step of knowledge update with the update of the knowledge map according to the update and development of medical image knowledge. According to the characteristics of the medical image knowledge, and the knowledge acquisition rate is greatly improved with the unstructured data such as textbooks and academic journals asa source of knowledge.

The invention relates to a weighted association rule mining method based on a data source partition matrix, and belongs to the technical field of manual intelligence, machine learning and data mining. By using a manual intelligence method, implied valuable expert knowledge can be mined from massive and complex industrial text data and expressed in an association rule manner, and a support can be provided for establishment of a knowledge base in an expert system. According to the method, a classic association rule mining algorithm is analyzed; a method for rapidly computing a support degree based on the data source partition matrix is provided aiming at the problem that an Apriori algorithm is low in efficiency as a data set is scanned repeatedly; and a weighted calibration method based on probability is used for weighting data item sets during an association rule mining process aiming at the problem that the Apriori algorithm treats data items in the transaction data set equally. Experiments verify that the provided method can reduce space complexity and time complexity of traditional association rule mining.

The invention discloses a task unloading method based on power control and resource allocation, and relates to the field of industrial Internet of Things. The method comprises the steps: establishinga cross-domain network model of an industrial field; constructing a calculation model of an equipment task; according to the model, constructing a mixed integer nonlinear programming model for communication power control, resource allocation and calculation unloading problems; decomposing a problem into three sub-problems; solving an optimal communication power and a resource allocation strategy by utilizing convex optimization knowledge, a Lagrange multiplier method and a KKT (Karush-Kuhn-Tucker) condition; after substituting an original target function, solving an optimal decision of a taskcalculation position by utilizing a deep reinforcement learning algorithm, and obtaining an optimal strategy of communication power, resource allocation and the calculation position of task unloading.The method can obtain the optimal strategy in industrial network task unloading, and has the technical effects of reducing the task delay, reducing the equipment energy consumption and ensuring the service quality.

The present invention relates to the design and implementation of a three stage computer-aided screening system that analyzes fundus images with varying illumination and fields of view, and generates a severity grade for diabetic retinopathy (DR) using machine learning. In the first stage, bright and red regions are extracted from the fundus image. An optic disc has similar structural appearance as bright lesions, and the blood vessel regions have similar pixel intensity properties as the red lesions. Hence, the region corresponding to the optic disc is removed from the bright regions and the regions corresponding to the blood vessels are removed from the red regions. This leads to an image containing bright candidate regions and another image containing red candidate regions. In the second stage, the bright and red candidate regions are subjected to two-step hierarchical classification. In the first step, bright and red lesion regions are separated from non-lesion regions. In the second step, the classified bright lesion regions are further classified as hard exudates or cotton-wool spots, while the classified red lesion regions are further classified as hemorrhages and micro-aneurysms. In the third stage, the numbers of bright and red lesions per image are combined to generate a DR severity grade. Such a system will help in reducing the number of patients requiring manual assessment, and will be critical in prioritizing eye-care delivery measures for patients with highest DR severity.