43results about How to "Effectively found" patented technology

The invention relates to the technical field of computer vision, in particular to an instance-level image search method based on multiple layers of feature representations. According to the method, feature representations, on different layers, of images are learnt through a deep convolution neural network model for instance-level image search so that different images of the same article can be effectively found. A coding learning process is introduced on the basis of a traditional network model, and features from multiple convolution layers are automatically encoded, so that the extracted features are more robust, and influences of background and noise data on the features are reduced. Meanwhile, the features integrate local information and class information. The invention further discloses a loss function based on multiple tasks. By optimizing the function, the learnt features have good generalization performance, and the learnt features can be well used for distinguishing inter-class images from intra-class images of different things.

The invention discloses a method for quickly optimizing an overall maximum power point of a photovoltaic array under local shadow. Under the local overshadowing condition, output characteristics of the photovoltaic array present multiple peak points. For searching an overall maximum value, the method comprises the steps of adopting a modified fruit fly algorithm to perform overall range searching, confirming a range where maximum power is located, then finding a maximum power point in a sub-region in a confirmed region by improving a golden section method and using the maximum power point as a working point of the whole photovoltaic array. By adopting the two steps, the overall maximum power point of the photovoltaic array can be quickly optimized.

According to an autonomous parking system and method based on cloud sharing and map fusion, autonomous parking based on tracking is adopted as a whole, only local path planning is carried out, and the problems that an existing global path planning method is high in computing power requirement and poor in real-time performance are solved. Track map sharing, track map selection and track map fusion are carried out at a cloud end, so that a map does not need to be established by a user, track map multiplexing is carried out only according to selection of GPS coordinates, scene semantic information, track label classification and the like, and a tracking route is optimized; and a cloud-vehicle end-user terminal cooperative communication mechanism is established, the cloud is used for storing, processing and multiplexing the map, the requirement of vehicle end hardware for storage performance is lowered, and meanwhile real-time performance and reliability of autonomous parking are guaranteed through real-time communication between the vehicle end and the user terminal. The parking method is efficient in parking and high in success rate.

The invention relates to an image retrieval method based on hierarchical clustering, proposes a system schema based on hierarchical clustering network image retrieval on the basis of the traditional image search engine through keyword semantic analysis and image visual clustering, and mainly includes that: (1) the image retrieval results are clustered in semantic level; (2) the image retrieval results are clustered in visual characteristic level; and (3) the hierarchical clustering navigation display is realized in a rapid and efficient way. Compared with the traditional image search engine, the invention adds the clustering idea to the image retrieval. The keywords are semantically analyzed to form different subjects; meanwhile, the images are also analyzed in visual characteristic level to cluster similar images in a class, and finally the retrieval results are displayed for the user in classification mode through a concise and clear interface, so as to help users quickly and efficiently find out the required target image from the retrieval results with unsorted subjects.

The invention discloses a live broadcast platform user interestingness mining system and mining method based on association rules and relates to the technical field of data mining. The system comprises a data acquisition module for acquiring live broadcast platform user behavior information from a server and generating sampling transaction event library and a testing transaction event library, an association rule model establishing module for adopting a spark calculation framework to conduct frequency mode mining on the testing transaction event library to obtain an association rule model, a user interestingness output module for using the testing transaction event library as an input variable to input the association rule model, obtaining an output variable of the association rule model and using the output variable as user interestingness contents. The live broadcast platform user interestingness mining system has the advantages of being short in computing cycle and high in testing result accuracy and practicability.

The invention discloses a valley delimitation searching method for tracking a multi-peak maximum power point of a photovoltaic power system. The valley delimitation searching method comprises the following steps of: for avoiding falling into a local maximum value during searching, partitioning a region after searching a valley by utilizing a valley delimitation method; for ensuring to obtain a global maximum value, performing sub-region search on a partitioned region and then determining region in which a maximum power point is located; finally, finding out a global maximum power point as a working point of a system by a variable-step-length searching method in the determined region. Through the algorithm design above, capability of responding complex environmental change of a practical photovoltaic power system can be strengthened, the shortage of a single-peak searching method is made up, and better maximum power tracking effect is reached.

When an icon associated with a folder one level lower is displayed within a window, files and folders included in the folder one level lower can be shown to a user in an easy-to-understand manner at a glance by previewing icons displayed in relation to files and folders two levels lower on the icon associated with the folder one level lower according to layout information belonging to the folder one level lower, and the files and folders two levels lower included in the folder one level lower can be previously confirmed by the preview even if the user actually performs no operation for specifying the folder one level lower.

The invention discloses a radiographic image weld seam extraction method based on a beetle antennae search algorithm and mathematical morphology. The method comprises the following steps: firstly carrying out preprocessing on a radiographic image, graying the radiographic image to obtain a gray level image, and carrying out two-dimensional adaptive wiener filtering processing at the same time; then carrying out threshold segmentation on the gray level image, searching an optimal threshold value for binarization processing on the radiographic image by utilizing the beetle antennae search algorithm and a clustering idea, and carrying out the threshold segmentation on a to-be-segmented radial gray level image according to the searched optimal threshold value, so as to obtain a binary image; and finally extracting a weld edge of the radiographic image, and carrying out denoising, object deletion and boundary extraction on the binary image, so that the weld seam edge of the radiographic image is obtained. The method disclosed by the invention combines the beetle antennae search algorithm with the clustering idea, the radiographic image with a binary weld seam can be rapidly and effectively obtained, and multiple basic operations of binary morphology are also organically combined together, so that the method is applicable to edge extraction of a weld seam image, and a smooth and continuous weld edge which is not interfered by defect areas can be extracted.

The invention provides a network communication anomaly alarming method and Ethernet switch equipment. A switching control logic chip is arranged between an Ethernet port physical (PHY) chip and a light-emitting diode (LED) indicator lamp. The method comprises that: a management central processing unit (CPU) system queries data messages received by an Ethernet port at the interval of a set period, controls the switching control logic chip to transmit a normal working indicator signal to the LED indicator lamp of the Ethernet port and indicate a corresponding state if judging the data messages received by the Ethernet port so far in a current period are all normal data messages, otherwise controls the switching control logic chip to switch the indicator signal into an anomaly alarming indicator signal and transmit the anomaly alarming indicator signal to the LED indicator lamp of the Ethernet port and indicates the corresponding state according to the anomaly alarming indicator signal by the LED indicator lamp of the Ethernet port.

The invention provides a random access memory (RAM), a network processing system and an RAM table look-up method. By dividing the RAM into at least two virtual memory banks, storing an identical business table in at least two virtual memory banks and determining the appropriate virtual memory banks by calculation of a network memory so as to have an access, access flow pressure of the RAM is effectively reduced while the table look-up speed of the RAM is improved, and network transmission performance is also improved. Simultaneously, the same business table can be stored in the virtual memory banks of the RAM, the table look-up speed of the RAM is improved to the largest extent, a network processor can calculate virtual memory bank address information in a Hash algorithm, not only is the counting process simple and convenient, but also the appropriate virtual memory banks can be found effectively, and table look-up efficiency and the network transmission performance are further improved.

The invention provides a key new word finding method based on multidimensional word and sentence characteristics and sentiment analysis. The key new word finding method comprises the following steps:step 1, grasping needed comments from an e-commerce platform; step 2, pre-processing the comments; step 3, carrying out word segmentation on the comments by adopting an NLPIR word segmentation tool; step 4, constructing and filtering repeated strings of a text subjected to the word segmentation by utilizing the multidimensional word and sentence characteristics; step 5, manually screening accuratenew words of a training sample; step 6, counting digital combinations and new class combinations of the new words; step 7, adjusting a filtering threshold value in step 4 according to a counting result and adding the new class combination and filtering to obtain a repeated string set of the tested sample; step 8, adding a user dictionary for optimizing the word segmentation; step 9, carrying outdependence grammar analysis on a word segmentation result by utilizing LTP (Long Term Potentiation); step 10, carrying out sentiment marking on governing words in a dependency relation by utilizing aCRF++ tool; step 11, taking the repeated strings which are sentiment words or the repeated strings with the governing words which are the sentiment words as the key new words. The key new word findingmethod provided by the invention is used for carrying out new word mining on a lot of comment texts of certain e-commerce products to obtain the new words which play a key role in the comments, so that the word segmentation accuracy is improved and a good foundation is provided for the mining and analysis work of the texts.

The invention discloses a system platform for health service management, which is characterized in that: health, service stationing, merchant stationing, commodity management and personal center are included; The health is a platform service system interface composed of a health service institution, The service entry is an examination and evaluation system interface established for practitioners,the merchant entry and commodity management is an authentication or management system interface established for merchants or institutions engaged in the health service industry, and the personal center is a management system interface for customers or health service practitioners to carry out self-management and regulation; By interacting with the four modules of commodity management and personalcenter, health service providers such as dieticians, health administrators and baby nurses are provided with a platform to realize the service value. The marketing channel for the health commodity operators which is suitable for their own commodity characteristics is established; users are allowed to find their own suitable dietitian or health management specialist.

The invention provides a damping controller coordination parameter optimization method based on a steepest descent method. The method comprises the steps of firstly, respectively constructing a wind power plant model, a photovoltaic power station model, a PSS model and a PODC model; secondly, analyzing the influence of the PSS and the PODC on the stability of the power system by establishing a dynamic model of the power system; then, respectively constructing a target function based on the characteristic value and a target function based on the damping ratio, and constructing a multi-target function; carrying out search optimization on the multi-objective function by utilizing SDA to obtain optimal PSS and PODC parameters; and finally, verifying the optimized power system by utilizing characteristic value analysis and a dynamic time domain simulation method to obtain optimal PSS and PODC parameters, so that the small interference stability of the power system can be improved. Accordingto the invention, SDA is adopted to optimize coordination parameters of PSS and PODC, so that extra enough damping is provided for low-frequency oscillation existing in a power system, and the small-interference stability of the system is improved.

The embodiment of the invention relates to the technical field of operation and maintenance control, and discloses a laser container finding method, a hoisting system, a controller and a storage medium. The laser container finding method provided by the invention is applied to the controller in the hoisting system. The hoisting system comprises a hoisting device and the controller; eight laser devices are arranged on the periphery of the hoisting device; four laser devices in the eight laser devices are distributed on four end points of a pair of opposite sides of the hoisting device; the other four laser devices are distributed on the other pair of opposite sides of the hoisting device in pairs; and the eight laser devices are respectively in communication connection with the controller.The method comprises the following steps of obtaining laser emission distance values measured by the eight laser devices; and controlling the lifting device to move according to the laser emission distance values reported by the eight laser devices until the reported laser emission distance values conform to the preset conditions. The high-precision container finding can be realized at low cost.

The invention discloses a method for extracting welding seams of radiographic images based on beetle whiskers algorithm and mathematical morphology. Firstly, the radiographic images are preprocessed, grayscaled to obtain grayscale images, and two-dimensional self-adaptive Wiener filter processing is performed at the same time. ; Then threshold the gray-scale image, use the beetle-beetle optimization algorithm and clustering idea to find the best threshold for binarization of the ray image, and perform threshold segmentation on the ray gray-scale image to be segmented according to the optimal threshold searched to obtain Binarize the image; finally extract the weld edge of the radiographic image, perform denoising, object deletion and boundary extraction on the binarized image, so as to obtain the weld edge of the radiographic image. The invention combines the beetle beetle optimization algorithm with the clustering idea, can quickly and effectively obtain the binary image of the ray weld, and at the same time organically combines a variety of basic operations of binary morphology, which is suitable for weld images The advanced edge extraction can extract smooth, continuous weld edges that are not disturbed by defect areas.

The invention provides a recommendation method and system for a recruitment system, and relates to the technical field of the Internet. The recommendation method comprises the steps: obtaining the personal information of a plurality of interviewees, comparing the plurality of pieces of personal information, and obtaining the same information or similar information of the plurality of pieces of personal information; carrying out the classification according to the same information or similar information, and classifying the interviewees with the number of pieces of same information or similar information being greater than a preset threshold value into an employment applying group; obtaining the browse records or delivery records of all persons in the employment applying group, and storingthe browse records or delivery records into a database; comparing the browse record or delivery record of each person with the database; and pushing the recruitment information which is not browsed orapplied by each interviewee in the database to the corresponding interviewee. The method and system provided by the invention can effectively help the interviewees find the unseen recruitment positions or information, enlarge the selection range for the interviewees, and enable the interviewees to quickly and accurately find the positions matched with the conditions of the interviewees.

The invention belongs to the technical field of digital communication and block chains, and relates to a block chain storage optimization method based on a remainder system and an RAFT algorithm in order to achieve rapid updating of block chain data and simply, efficiently and substantially reduce the storage data volume of a block chain, which comprises the following steps: selecting n remainderbases with the bit width of b; utilizing the characteristic of remainder system parallel computing, so that each node only needs to update the local remainder in parallel to finish data updating; performing hybrid error detection and correction based on an extended Chinese remainder theorem CRT II and a consensus algorithm Raft; in the data recovery process, using a hybrid error detection and correction scheme based on a CRT II and a Raft algorithm, splitting a whole remainder system into a plurality of two-base remainder systems, adding two remainders determined by the Raft algorithm into theremainder systems to serve as redundant bases, and judging whether malicious data exist or not according to the overflow judgment theorem. The invention is mainly applied to block chain occasions.

The invention relates to a novel website navigation realization method. The method comprises a football-shaped network navigation framework (1), a football-shaped navigation keyword button (2) and a pre-view window (3) and comprises the following steps: coding the football-shaped network navigation framework (1) with HTML codes; adding special effects to football-shaped navigation with java scriptsuch that the football-shaped network navigation framework can rotate with a mouse while adjusting rotary speed according to the distance from the mouse to navigation with appearance of the content pre-view window (3); extracting keywords for information on one website; inputting extracted keywords into the button (2) part left by football-shaped navigation; adding website information pre-view information into a navigation button; adding a navigation function into a front page of the website; enabling the football-shaped navigation to rotate by means of the mouse in order to find out what a user wants, putting a mouse cursor to see pre-view information and entering a relevant site by clicking on the navigation button. The novel site navigation realization method enables the overall sit tobe well-organized and clear. Therefore, the user can find out required information in the most-effective manner.

The invention discloses a shopping basket recommendation method and device based on course purchase and electronic equipment. The method comprises the following steps: collecting course purchase data entries; according to an FP-Growth algorithm, obtaining a frequent pattern through the data items; according to the frequent mode, recommending a second course based on the purchased first course, wherein the first course and the second course are both in the frequent mode; and repeating the steps by using the data of the newly purchased course. The method can help students to quickly and effectively find associated courses of a certain course, so that the students can understand knowledge points with a coherent thinking, and the understanding of the students on the knowledge points is improved.

An autonomous parking system and method based on cloud sharing and map fusion, which adopts tracking-based autonomous parking as a whole, and only performs local path planning, which solves the problem that the current global path planning method requires high computing power and low real-time performance. Poor problem; track map sharing, track map selection, and track map fusion are carried out in the cloud, so that there is no need to build maps by yourself, and only need to select and reuse track maps according to GPS coordinates, scene semantic information, track label classification, etc., optimized Tracking route; establish a cloud-vehicle-user terminal collaborative communication mechanism, use the cloud to save, process and reuse maps, reduce the storage performance requirements of the car-end hardware, and at the same time, the real-time communication between the car end and the user terminal, The real-time and reliability of autonomous parking are guaranteed. The parking method is efficient in parking and has a high success rate.

The invention relates to the technical field of combination instrument fault diagnosis, and discloses a combination instrument fault diagnosis method, device and equipment and a storage medium. The method comprises the steps: recording the abnormal information of a combination instrument during the operation of the combination instrument, and obtaining a preset design rule; compiling the abnormalinformation according to the preset design rule to generate an abnormal code; obtaining a preset diagnosis code, and comparing the abnormal code with the preset diagnosis code; and positioning the abnormal information according to a comparison result to obtain a fault range of the combination instrument. According to the invention, due to the fact that the abnormal information is compiled into theabnormal codes, the problems of the combination instrument can be directly converted into the corresponding abnormal codes, the abnormal codes are compared with the preset diagnosis codes, and the fault range of the combination instrument is positioned, so that compared with the prior art that a corresponding fault piece needs to be taken for analysis, the method of the invention has the following characteristics that the position of the fault can be found more quickly and effectively so as to improve the troubleshooting efficiency.

The present invention provides a RAM, a network processing system and a RAM table look-up method, by dividing the RAM into at least two virtual memory banks, and storing the same business table in at least two virtual memory banks, and by The calculation of the network memory determines the appropriate virtual memory bank to access, while increasing the table lookup rate of the RAM, it effectively reduces the access traffic pressure of the RAM and improves the network forwarding performance. At the same time, the same business table can also be stored in the virtual memory bank of the RAM to maximize the table lookup rate of the RAM, and the network processor can use the hash algorithm to calculate the address information of the virtual memory bank, which not only makes the calculation process more efficient It is simple and more effective to find a suitable virtual memory bank, further improving the efficiency of table lookup and network forwarding performance.

The invention discloses an answer extraction method and system based on deep learning. wherein an implementation of the method includes the steps of collecting knowledge fragments related to the answer of a question to be answered according to keywords in the question to be answered; based on the type of the question to be answered, analyzing the part-of-speech tagging of the knowledge fragment, and using the knowledge fragment containing the type of the question to be answered as a candidate answer; using the trained word2vec model to calculate the similarity between the keywords and words contained in the knowledge fragment in the candidate answer, and finding words that meet the similarity requirements as target candidate answers; substituting the words in the target candidate answers into the question to be answered to obtain a number of target statements, sorting each target statement by judging semantic information of each target statement, and using the target statement with thehighest score as the target answer. Through the extraction method and system, the answers of multiple words with higher similarity can be found, thereby improving the accuracy of the system.

The invention provides a reference crop evapotranspiration calculation method based on remote sensing data and belongs to the field of quantitative irrigation design and agricultural water resource management. The method comprises the steps that firstly, training weather data of a researching area is acquired, an output data set with corresponding training reference crop evapotranspiration ET0 asa training set is calculated, training remote sensing data of the researching area is acquired as an input data set of the training set, and a reference crop evapotranspiration calculation model of the researching area is established; then EY0 and remote sensing data are verified, and the precision of the model is verified; during use, the remote sensing data in a required time period is acquiredand input into the verified model, and the model outputs the reference crop evapotranspiration in the selected time period of the researching area. According to the reference crop evapotranspiration calculation method, by combining less weather station data with easily acquired remote sensing data, the method can be actually applied to area reference crop evapotranspiration calculation of an areawithout weather station data and an area with sparse weather stations and effectively assist in local irrigation planning and agricultural water management.

The invention relates to the field of information technology and discloses a code comparison method, device and system during continuous integration. The method comprises the steps that a new version code and an old version code are acquired; the new version code and the old version code are compiled; the new version code and the old version code obtained after compilation are compared; and a varying source code is acquired and displayed according to the comparison result of the new version code and the old version code obtained after compilation. Through the code comparison method, device and system, comparison of the new version code and the old version code can be realized in a true sense, changes of the code is precisely determined, a development team can find the difference between the new version code and the old version code conveniently, quickly and effectively, and more efficient development is realized.

The invention discloses an attention fine-grained classification method based on weak supervision position positioning. The method specifically comprises the following steps: 101) a preprocessing step, 102) an information processing step, 103) an information progress step, 104) a feature data step and 105) a classification confirmation step. The invention provides an attention fine-grained classification method based on weak supervision position positioning, which can be used for positioning areas with relatively small inter-class differences and carrying out positioning and amplification.

The invention discloses a nail taking device used in the department of orthopaedics. The nail taking device comprises a hollow periosteal detacher, a screwdriver head, an operating handle, an arthroscope, a button, a spring and an injection hole, wherein the screwdriver head is located inside the hollow periosteal detacher, one end of the screwdriver head is connected with the operating handle through the spring, the button is arranged outside the operating handle, and the arthroscope and the injection hole are further arranged outside the operating handle. The nail taking device has the advantages of being convenient to operate, capable of saving time and effort, safe and effectively and is suitable for popularization and application.