246results about How to "Solve redundancy" patented technology

The invention relates to the field of knowledge extraction and particularly relates to a knowledge extraction method and system based on a memory neural network and a device. The invention aims to solve the problem of information redundancy existing in the prior art. The method comprises a step of obtaining a possible relationship type in an input text and a semantic coding vector by using a convolutional neural network under the premise of giving a predefined relationship type, a step of carrying out semantic encoding by using a two-way long and short time memory neural network and obtaininga semantic vector, a step of taking the relationship type as an initial value of the two-way long and short time memory network and a first label in a decoding module and thus fusing the relationshiptype information into encoding information and label information of the decoding module, and a step of obtaining a label sequence by using a decoding module of a single-way long and short time memorynetwork and then obtaining structured information by parsing the label sequence. According to the knowledge extraction method and system and the device, the efficiency of structured information extraction is greatly improved, and the problem of information redundancy existing in the prior art is solved.

The invention discloses a method for improving the bandwidth utilization rate of a BFT consensus algorithm based on a block piece. In one blockchain network, when a node receives new transaction, andthe new transaction passing through the verification will be cached till a certain number of transactions are obtained or there is no new transaction in a specified time window, so the transactions will be taken as one block piece and broadcast is transmitted to other nodes, and other nodes will put the block piece into respective transaction block piece pools; For a main node, after a certain number of transaction block pieces are obtained or there is no new block piece in the specified time window, the block pieces are packaged into one block for consensus, and a prepared message sent by themain node just needs to transmit the Hash of the block pieces, and other nodes organizes the required blocks for consensus in the respective block pieces according to the block piece Hash table of the main node. The method guarantees the effectiveness of the BFT algorithm, also solves a problem of redundancy of the broadcast mechanism of the BFT algorithm, and improves the bandwidth utilization rate of the blockchain node.

The invention discloses an electronic nose signal processing method based on a mixing characteristic matrix. The method comprises the steps that first, characteristic extraction is conducted on an original data matrix to establish the mixing characteristic matrix, then a binary quantum particle swarm optimization algorithm is used for conducting characteristic selection on the mixing characteristic matrix, at last, the mixing characteristic matrix is fed into a classifier, a decimal quantum particle swarm optimization algorithm is used for conducting synchronous optimization on mixing characteristic sub matrixes and the parameters of the classifier, mode recognition is conducted to determine the mixing characteristic sub matrix with the highest recognition rate, and the sensor characteristic corresponding to the mixing characteristic sub matrix is selected to be used as the optimized characteristic of an electronic nose signal for conducting mode recognition. The electronic nose signal processing method based on the mixing characteristic matrix has the advantages that the defect that the complete information of an entire response curve can not be reflected only by extracting a single characteristic is overcome; the operation complexity is lowered, the problem of redundancy between sensors is effectively solved, the recognition rate of wound infection detection of an electronic nose is improved, and beneficial guidance can be provided for a doctor to choose a proper treatment method.

The invention relates to a data backup technology, in particular to a disk incremental backup technology in a computer system. An incremental backup method comprises the following steps of: adding a valid bitmap into an incremental backup program to mark a valid data memory unit; only backing up the increment of valid data during incremental backup to reduce data redundancy and improve backup efficiency; adding a monitoring bitmap into the incremental backup program to mark a data change memory unit; backing up the increment of the valid data when monitoring is valid; and backing up all validdata when the monitoring is invalid. When the program is totally normal, the incremental backup is efficiently finished, and when the monitoring program is abnormal, no loss of increment of the validdata can be also guaranteed.

The invention provides a Shiro authentication method based on Redis storage. Through placing a session into an Redis for caching, the memory consumption can be reduced to solve the problem of memory overflow; and as an expiration time algorithm is introduced to automatically eliminate a cache, the efficiency and the hit rate of the cache are improved, thereby guaranteeing data to be stored efficiently and reliably. The method comprises the steps of carrying out encapsulation on identification information of a user according to login parameters of the user to generate a unique identifier; according to the unique identifier, finding whether the corresponding cache exists in an Redis server; if the corresponding cache exists, obtaining login information of the unique identifier and encapsulating login successful information; and if the corresponding cache does not exist, calling a login authentication method and an authority authentication method for authentication, if the authentication is passed, encapsulating the login successful information, and if the authentication is failed, prompting the login authentication is failed.

The invention discloses a rail transit emergency power supply system and a control method, belonging to the technical field of rail transit power supply. The invention mainly comprises a matching design of a D-series high-speed train emergency power supply system and a hybrid power control method, wherein the emergency power supply system mainly comprises a fuel cell system, an energy storage device, an energy controller and the like. The matching design is mainly based on the running line of D-series high-speed trains under the emergency mode and the power and energy under vehicle parametersas constraint conditions, and the volume cost is used as the objective function for performing matching design to obtain a scheme of D-series high-speed train emergency power system. According to theinvention, the emergency power supply system of the D-series high-speed train can be effectively designed and controlled, the electric energy demand under the emergency situation is satisfied, and theenergy utilization efficiency is improved.

The invention provides a light field camera internal parameter calibration device and method, and the method comprises the steps: fixing a light field camera on an electric control translation platform, controlling the electric control translation platform through a motion controller, and driving the light field camera to translate to different positions; the calibration plate is arranged in frontof a main lens of the light field camera and is fixed on the angle rotating table, and the motion controller controls the angle rotating table to drive the calibration plate to rotate at different angles; the calibration plates under different postures are shot by rotating the calibration plates or moving the light field camera, the points on the calibration plates and the image points corresponding to the points are determined, the internal parameters and the external parameters under the corresponding postures of the light field camera are calculated through linear initialization, finally acost function is established, and an optimal solution is obtained through iteration. According to the invention, the internal parameters of the light field camera can be calibrated more flexibly andaccurately.

The invention discloses a gesture image key frame extraction method based on image similarity. The method comprises the steps of firstly, converting an input gesture video into a video frame image; classifying the video frame images into the static gesture images and the dynamic gesture images by adopting a Mobilenet-SSD target detection model; for the dynamic gesture image, adopting the cosine similarity to judge an image key frame; for the static gesture image, firstly adopting the cosine similarity for pre-extracting the key frame, then using a VGG16 network model for extracting the abstract features, carrying out the space gradient calculation, and according to the gradient difference of two adjacent frames of images, setting a self-adaptive threshold value to complete the judgment of the final key frame of the static image. According to the method, the problems of missing detection and redundancy of the video key frames are effectively solved, and by adopting the VGG-16 network model, the expression capability of the image is greatly enhanced, and the pre-extracted static gestures can be extracted again, so that the number of the static gesture key frames is reduced.

The invention provides a millimeter wave radar and video monitoring camera combined calibration method. According to the invention, the video detection technology is utilized to accurately identify the types of traffic participants, and the millimeter wave radar detection technology is combined to monitor the target in the positioning range in real time, so that the two technologies complement each other. A multi-sensor fusion model is utilized to establish an accurate coordinate transformation matrix among a radar coordinate system, a video sensor coordinate system, an image coordinate system and a pixel coordinate system, and spatial fusion of multi-sensor data is realized.

The invention discloses a domain adaptive unsupervised target detection method based on feature separation and alignment, and the method comprises the following steps: inputting paired RGB samples ofa source domain and a target domain into a two-stage target detection framework, calculating the detection loss through the label of the source domain, training the target detection framework, and obtaining a learned target detection model, wherein in the training process, the multi-level high-dimensional features are aligned; aligning the regional instance features of each group; graying the RGBsamples of the source domain and the target domain which are paired, carrying out feature separation, and separating out high-dimensional features related to detection and disturbance features irrelevant to detection. The invention provides an unsupervised domain self-adaptive target detection method based on feature separation and alignment, which can effectively solve the problems of backgroundinformation noise and candidate region redundancy in domain self-adaptive target detection.