235results about How to "Reduce the number of comparisons" patented technology

The invention provides a face tracking method with classification and identification functions, which comprises the following steps: 1. a sequence to be matched of a current frame is F1, F2; a sequence to be matched of a front frame is F1,F2to Fj; 2. Fm of the current frame is taken out in sequence, and an LBP code pattern is calculated; 3. the obtained LBP code pattern is compared with an LBP code pattern of the front frame so that Fn which has the maximal matching degree with the Fm is found out; 4. whether a person is the same person or not is judged according to the matching degree and the central position distance relation; if so, face images of the same of the person are classified into the same type, if not, Fj+1 equal to Fm is added in the FL sequence; 5. the step 2 to the step 4 are repeated until the matching sequence of the current frame is traversed; and 6. the step 1 is back, and the next frame is processed. The invention adopts LBP characteristics with a rotary uniform mode to classify and identify the identity of a person, classifies the same person into one kind, solves the problem that information is lost after the person is sheltered in the traditional face tracking method, expands the function of a face tracking system and improves the application value thereof.

An embodiment of the invention provides a method and a system for building indexes and a method and a system for retrieving the indexes. The method for building the indexes includes enabling an acquisition server to store acquired real data and generating index data according to storage positions of the real data; storing the index data into current buffer blocks according to a key value sequence; continuously arraying the index data in the buffer blocks into index blocks according to the key value sequence; sequentially arraying the index blocks to form index files. The index data comprise key values. The method for retrieving the indexes includes acquiring key values of target data; comparing the key values of the target data to key value ranges of the various index blocks in the index files and determining certain index blocks with the key value ranges covering the key values of the target data as target index blocks; searching the target index blocks to obtain index data with key values equal to the key values of the target data and using the index data as target index data; extracting storage positions of the target data from the target index data and reading the target data at the storage positions.

The invention provides a method for clustering optimization based on a Canopy algorithm. The method comprises the following steps of step (101) carrying out packet processing on all original data based on the Canopy algorithm in order to obtain N Canopy sets and a center of each Canopy set; and step (102) taking a number N of the Canopy sets as a divided number k constructed by a K-means clustering algorithm; taking the center of each Canopy set as a cluster center of a divided kth cluster; and adopting the K-means clustering algorithm to carry out clustering optimization processing on all original data based on the determined cluster number k and cluster center, and outputting a clustering optimization result. According to the method and the system, the Canopy clustering algorithm is used for preprocessing of the K-means clustering algorithm and is used for finding the proper k value and cluster center. The running time of the whole cluster is greatly reduced, the computational efficiency of the algorithm is improved, and the fault tolerance of the algorithm is increased.

The invention provides a hash-based method for searching a CAM (central address memory). The CAM has a data storage function and a high-speed data content searching function and also supports multiple instructions such as reading, writing and searching. However, the most important function of the CAM is the high-speed data content searching function, and the searching function is based on other functions of the CAM. Therefore, the core of the design of a CAM functional model is the design of a high-speed searching engine. The searching engine can maintain a sheet and receive external input data. The input data and each item in the sheet are completely compared (accurate) or partially compared (Don'tcare function) through a certain search algorithm in the searching engine, and then comparison results are output. The rapid CAM content search can be achieved via a hash function. When a request is searched through the CAM, hash calculation is carried out for data to be searched to obtain a hashkey, a first address of a CAM table item data linked list is obtained via access to a hash entry table according to the hashkey (as a hash entry table address), the access to the CAM data table item linked list is completed, the data to be searched and items in the CAM data linked list are compared one by one, and finally CAM search results are returned.

The invention discloses a face recognition method, and the method comprises the following steps: obtaining a face image of a to-be-recognized person, inputting the face image into a target network model, and obtaining a target group category label and feature data of a face; reading face image feature data with a target group category label and marked personnel identity information from a database; and comparing the feature data with the face image feature data, and determining identity information of the to-be-identified person according to a comparison result. According to the method, the comparison frequency can be reduced, the comparison time is shortened, and the face recognition efficiency is further improved. The invention further discloses a face recognition device and equipment and a readable storage medium which have corresponding technical effects.

The invention relates to a device for detecting gene fusion of circulating tumor DNA (deoxyribonucleic acid) samples. The device comprises a sequencing data acquisition module, a comparison module, a re-comparison module, a real fusion breakpoint judgment module and an output module. The device for detecting the gene fusion of the circulating tumor DNA samples has the advantages of high detection speed and stability and low resource requirement.

The invention discloses a method and a device for rapidly positioning water color abnormity based a remote sensing image, and belongs to the field of water pollution monitoring. The method comprises the following steps: obtaining the remote sensing image, and extracting a water area part; calculating the normalized water index of the water area, and performing mean calculation by using a mean operator; comparing the mean-calculated normalized water index distribution image with a water color abnormality judgment threshold to obtain a water color abnormity window distribution image; performingmask operation on the normalized water index distribution image by using the water color abnormity window distribution image to obtain a preliminarily-determined water color abnormity area distribution image; and comparing each pixel of the preliminarily-determined water color abnormity area distribution image with the water color abnormity judgment threshold to obtain a water color abnormity area. Scale conversion is performed by using the mean operator, so the comparison frequency is reduced, the processing time is greatly shortened, the speed and the accuracy of water color abnormity detection are improved, and a technical support is provided for the environmental emergency management department.

The invention discloses a successive-approximation analog-to-digital converter and a quantitative method based on an adaptive prediction interval thereof, and belongs to the technical field of analog-to-digital conversion. In the invention, a DAC module comprises two redundant capacitances; the quantitative method comprises the steps of switching a first redundant capacitance according to a firstcomparison result before the redundant capacitance is switched and obtaining a switched second comparison result; judging whether to continue quantification or to switch the second redundant capacitance according to a comparison of the first comparison result and the second comparison result and obtaining a switched third comparison result; and then judging whether to adjust a prediction intervalor to directly start the quantification from a highest digit according to a comparison of the second comparison result and the third comparison result. According to the successive-approximation analog-to-digital converter and the quantitative method based on the adaptive prediction interval thereof, a sampling point in a window size range of the prediction interval can be guaranteed to be predicted correctly; and through the correct prediction, the comparison times and the switching times of a capacitance switch in the DAC module can be greatly reduced.

The invention relates to an index read-write method based on compression merging and asynchronization updating. The method includes the steps that indexes are used for generating index cache files according to storage region serial numbers; incomplete index cache files or writing index cache files are filtered and removed; when the total number of the index cache files meets the asynchronization updating initiation condition, data of storage regions corresponding to the same storage region serial numbers in all the index cache files are read respectively, and the data are merged and sorted; data of storage regions corresponding to the current storage region serial numbers in an index database are read and decompressed; the data are merged, sorted and compressed, the executed data are covered and updated to the storage regions corresponding to the current storage region serial numbers in the index database, and metadata are updated; whether the ends of files of the index cache files are read is judged, if yes, the process is completed, if not, the storage region serial numbers are added with one, and the process is executed continuously. Compared with the prior art, the index read-write method based on compression merging and asynchronization updating has the advantages of being efficient, asynchronous, capable of reducing the data size for reading a magnetic disk every time and reducing the index comparison frequency and the like.

The system thereof comprises: a web server connected to the client side; a database connected to the server, a broadcast device connected to the database for use in saving the number information provided by the task request of group-sending message into the zone mapping table of memory. The method includes the following steps: the broadcast device gets the numbers provided by the task request of group-sending message from database and saves them into memory; according to the parent zone information, extracting the number section information of each zone corresponding to parent zone from database; after comparing the number information to the number section information of each zone corresponding to the parent zone, the number information is saved into the zone mapping table and then is sent to the short message center.

The invention discloses a multi-dimensional distance clustering anomaly detection method and system based on a time sequence, and belongs to the technical field of aviation safety. The multi-dimensional distance clustering anomaly detection method based on the time sequence comprises the following steps: step 1, preprocessing a trajectory data set, the preprocessing comprising cleaning and re-integration; 2, calculating the multi-dimensional similarity between the tracks; 3, for the multi-dimensional Hausdorff distance, constructing an inter-track similarity matrix; 4, carrying out a hierarchical clustering algorithm of the multi-dimensional hausdorff distance; selecting a hierarchical clustering algorithm in machine learning to perform hierarchical clustering based on the similarity matrix; and step 5, detecting the anomaly detection effect of the algorithm, constructing a track with anomalies in speed, direction, longitude and latitude, clustering the abnormal track with a normal track through the hierarchical clustering algorithm, and evaluating the clustering algorithm by selecting a correct rate, a precision rate, a recall rate and an F1 value.

The present invention discloses a method and a system for link list encoding. The link list coding method includes steps below: S102. A reference link list is compared with a current link list; S104. It is necessary to add/delete bit mask code, select an encoding method according to the comparison above and utilizes the selected encoding method to encode added/deleted bit mask code in the current or reference link list. The present invention greatly reduces quantity of element comparison between the current link list and the reference link list, improving encoding efficiency and enhances module reusability in a proposal of deletion at first and then addition by making functions of calculating added/deleted bit mask code, etc. independent.

The invention relates to a flow table rapid searching method and system under a high-concurrency network environment. The method comprises the steps of: 1) counting the flow entering a network interface, setting a buffering window of a buffering area according to the counted current flow state; 2) according to the size of the set buffering window, utilizing quintuple information to group arrived data packets; 3) according to a preset scheduling strategy, scheduling each cached group, and successively sending each group to a connection management module; and 4) extracting the quintuple information of each group by the connection management module, carrying out flow table searching, finding corresponding flow table items, and using the data packets in the groups to update the information of the flow table items. The flow table rapid searching method and system are mainly suitable for a high speed network flow processing system of a backbone link, the access cost of the connection management module under the high speed network environment is optimized, and the access efficiency of the flow table is improved.

The invention relates to the field of computer data transmission and discloses a method for increasing the read-write speed of spatial big data files. The method for increasing the read-write speed of spatial big data files comprises a zero-copy step, a file buffering step and a digest checking step and particularly comprises the steps of when a read-write access is performed on a spatial big data file, obtaining a message digest of the spatial big data file through the digest checking step and comparing the message digest of the spatial big data file with a message digest of a file in a cache system; if the spatial big data file has the same copy in the cache system, updating cache hit information of the copy, updating the weight W of the copy, adjusting the link position of the copy in a sequential linked list according to the weight W of the updated copy, and transmitting the copy through the zero-copy step; if the spatial big data file does not have the same copy in the cache system, creating the copy of the spatial big data file in the cache system, updating the weight W of the copy, and transmitting the copy through the zero-copy step. The method for increasing the read-write speed of spatial big data files has the advantages of overcoming data read-write bottlenecks and being beneficial to increasing data read-write speed on the whole.

The embodiment of the invention discloses a consensus establishment method based on a lot drawing mechanism, devices and a readable storage medium, and relates to the field of information processing.The method, the devices and the medium are used to reduce time consuming of consensus algorithms in the lot drawing mechanism. The method of the embodiment of the invention includes: acquiring first data, which are broadcast by an accounting node of an end-position block in a block chain, by a current node; using a local private key to carry out signature on a hash value of the first data to obtain second data; carrying out hash operation on the second data to obtain a local candidate hash value; then determining a broadcast moment, which corresponds to the local candidate hash value, according to the first algorithm; then broadcasting the local candidate hash value to the block chain at the broadcast moment corresponding to the local candidate hash value, determining alternate hash valuesaccording to an order of time when foreign candidate hash values are received; and selecting a target hash value from the local candidate hash value and the alternate hash values, wherein a node corresponding to the target hash value is an accounting node of a to-access block.

The invention relates to an image processing method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the feature extraction of a to-be-processedimage, and obtaining a first feature of the to-be-processed image; according to the first feature and class center features of a plurality of reference image classes in a feature library, determiningan image class of the to-be-processed image; and under the condition that the image category of the to-be-processed image is a first category in the plurality of reference image categories, updating acategory center feature of the first category according to the first feature and a plurality of pieces of feature information of the first category in the feature library. According to the embodimentof the invention, the image retrieval speed and accuracy can be improved.