51results about How to "Reduce the number of collections" patented technology

The invention discloses to a method and a method for generating a decision tree in a data mining system. The main technical scheme comprises steps as follows: A, a data set is traversed and set, so as to determine the unclassified data set corresponding to every candidate unit on the current layer of the decision tree; B, the attribute value of the data attribute corresponding to every candidate unit is determined according to the determined unclassified data attribute; C, the decision tree branch of every unit is generated according to the determined attributed value; and D, whether the data in the unclassified data set corresponding to the lower unit of every decision tree branch is identical in the attribute value with the pre-determined set data attribute or not is judged respectively, the units with a negative judgment are determined to be candidate units of the next layer of the current layer, the next layer serves as the current layer, and the step A is carried out again, and units with a positive judgment are determined as the last unit of the branch respectively. By adopting the technical scheme, the times for traversing the data set is reduced, as a result, the computational efficiency is improved, and the occupation of system resource is lessened.

The invention discloses an indoor positioning method for reducing fingerprint data acquisition workload. For a large number of point pairs with larger AP (Access Point) RSSI (Received Signal Strength Indicator) difference values, the physical distances are larger. The method comprises the following steps: calculating the RSSI 'distance value' between points by taking the RSSI values of multiple APs obtained at a certain position as high-dimension characteristic data of the position according to the characteristic for serving as the intimate relationship between two points; acquiring the relative position relationship of one low-dimension space between each object by using a multi-dimensional scaling analysis algorithm; evaluating the absolute positions of other points by taking a small number of sample points as anchor nodes. By adopting the method, the aim of reducing the acquisition number of sample points is fulfilled. A hierarchical region reducing method is put forward on the basis in order to further fulfill the aim of reducing the positioning error. Compared with a KNN (K Nearest Neighbor) algorithm, the method has the advantages that the sampling points can be reduced, and meanwhile the positioning error is reduced.

The invention relates to artificial intelligence, and provides a damage detection model training method and device, a vehicle damage detection method and device, equipment and a medium. The method comprises the steps: inputting an obtained damage sample image into a damage detection model containing a first parameter, extracting damage features, and generating an intermediate convolution feature map; inputting the intermediate convolution feature map into a mask prediction branch model containing a second parameter; outputting a training result through the damage detection model according to the damage features, and obtaining a mask result through the mask prediction branch model; obtaining a first loss value through the first loss model, and obtaining a second loss value through the second loss model; obtaining a total loss value; and when the total loss value does not reach the convergence condition, iteratively updating the first parameter and the second parameter until the total loss value reaches the convergence condition to obtain a trained damage detection model. According to the invention, the damage type and the damage area can be quickly identified. The invention also relates to a blockchain technology. The damaged sample image can be stored in the blockchain.

The invention provides an injector defect detection method based on semantic segmentation. The injector defect detection method comprises the following steps: S1: performing syringe image acquisition,S2, converting a json file into a png mask picture by using a code; S3, performing image processing, thus completing the manufacturing of an injector data set; S4, constructing a full convolutional neural network model; inputting the training data set into a semantic segmentation network; and iterating the model parameters to obtain a convergent segmentation model; S5, testing a model, inputtingthe test data set into a semantic segmentation network, and obtaining injector segmentation maps; S6, exporting a model file; and S6, carrying out defect detection through a semantic segmentation network. The injector defect detection method performs defect detection through the semantic segmentation network, automatically judges whether the injector has defects or not is faster and more accurate,is quicker and more accurate, compared with traditional manual detection, can improve the automation degree of the manufacturing process, can greatly reduce the enterprise cost, and meanwhile, can improve the productivity.

The invention belongs to the field of food inspection technology, and discloses a food data analysis method based on cloud computing. The food data analysis method based on the cloud computing includes the steps of photographing information of a food production process by a camera unit; processing, by a processing unit, information transmitted by the camera unit; the processing unit identifying the information through a built-in image identification and determination module; transmitting the processing unit information to a monitoring unit through a wireless routing unit; the monitoring unit monitoring the information transmitted by the routing unit through a network cloud server; sending out a warning prompt on illegal information; and carrying on multi-aspect control through data sharing between the network cloud server and the monitoring unit by using a mobile terminal. According to the invention, the supervision is greatly enhanced, a more accurate understanding of the processing process of a factory is gained, the processing process is more transparent, opportunistic and irregular production methods are eliminated, and black-hearted manufacturers are also restrained from producing harmful foods; and at the same time, the examination work is better facilitated.

The invention discloses a high-speed evanescent field frequency shift super-resolution microscopic imaging system and imaging method, and the method comprises the steps: employing an on-chip light-emitting device to generate evanescent fields with three wavelengths transmitted in the same direction at the same time, and employing a color camera to collect a frequency shift image generated by the three-wavelength evanescent fields which illuminate in the same direction at the same time, wherein the three wavelengths correspond to frequency shift signals of three wave bands of red, green and blue, and the illumination of each wavelength has different frequency shift amounts. In the image reconstruction process, multi-wavelength information shot at the same time is separated out to be subjected to interpolation processing, and therefore sample information with different frequency shift amounts and different illumination directions is obtained. And finally, image reconstruction is carriedout, high-frequency and low-frequency spectrums of the sample in different directions are spliced by using a splicing algorithm in a spectrum space, and finally, a high-resolution image breaking through the optical diffraction limit of the traditional microscope is recovered. According to the invention, under the condition of sacrificing a certain image sampling rate, the image acquisition quantity of the evanescent field frequency shift super-resolution method can be effectively reduced, and the imaging speed is improved.

The invention belongs to the technical field of image recognition, and discloses a large-scale commodity recognition method based on deep learning, and the method comprises the following steps: S1, building a commodity detection model; S2, inputting a to-be-detected picture into the commodity detection model, and obtaining all commodity positioning data in the to-be-detected picture; S3, establishing a commodity classification model; and S4, inputting the to-be-detected picture into the commodity classification model, and obtaining corresponding commodity category data according to all the commodity positioning data. The method solves the problems that in the prior art, the precision is difficult to meet the requirements of large-scale projects, the requirements of training samples are extremely high, rapid iterative updating cannot be realized, the reusability of the samples is low, the equipment requirements are high, the feature expression capability is limited, the workload is hugeand the operation speed is low.

The invention belongs to the technical field of image processing, and discloses an image processing system in orthopedic anesthesia based on optimized sedation management and regional blocking. The image processing system comprises an ultrasonic image obtaining module, an operation state detection module, a main control module, an image processing module, a three-dimensional navigation module, a storage module and a display module. Through the image processing module, a doctor is prevented from manually identifying lesion tissue, the image identifying efficiency is improved, the doctor does not need to have a high lesion identification ability, and the labor cost is reduced. Through the three-dimensional navigation module, a virtual bone image of a surgical spot of a user and a three-dimensional image of the specific location, where a surgical instrument enters, in the user body are overlaid on a real vision of the doctor, in the surgical process, real-time navigation is conducted, thesurgical operation efficiency of the doctor is greatly improved, and a surgery is more safely, accurately and efficiently implemented.

A flash memory address translation layer system connected with a flash memory and an EMS read/write controller comprises: an instruction buffer for connecting with the EMS memory read/wrote controller; a logical address buffer for connecting the EMS memory read/wrote controller; a data buffer for connecting with the EMS memory and the EMS memory read/wrote controller; a pair of assistant controller for providing an instruction assistant function for controlling data's read/write; a microprocessor for connecting the instruction buffer and two assistant controllers; an address translation unit for connecting the microprocessor; a flash memory address buffer for connecting with the assistant controller, address translation unit and the flash; a adjustable address translation layer unit for connecting with an assistant controller; a quick fetch instruction and data buffer for connecting with the adjustable address translation unit. The invention can greatly reduce the number of invalid data collection and occupy capacity of the flash memory, and has a quick mechanism with space efficiency to improve and quicken the efficiency which the logical address corresponds to the real flash address.

The invention relates to a microblog information priority collection method based on multiple strategies, which can timely and effectively obtain the information of a blogger by constructing a multi-strategy priority collection method under the condition of limited collection capability. The method comprises the steps of firstly, constructing a classification model to screen bloggers, removing junk bloggers, and dividing the remaining bloggers into three categories according to the number of microblogs and the number of fans; secondly, aiming at different types, constructing different acquisition strategies; clustering the blog sending time of the large V bloggers, and extracting the optimal collection time of the large V bloggers; and through the microblog statistics of the bloggers, training a regression model, predicting the activity values of the bloggers, and sorting the bloggers according to the activity values; and finally, designing a multi-strategy microblog priority acquisition method by integrating the three types of acquisition strategies, and maintaining the timeliness of the acquisition strategies by periodically updating the acquisition queue. Experiments show that the hotspot microblog information can be effectively obtained in time, and the collection number is greatly increased.

The invention belongs to the technical field of application of artificial intelligence to grain storage, and particularly relates to a granary grain storage quantity parallel detection method, which comprises the following steps of: firstly, generating virtual data samples according to real data samples by adopting a mixed data generation method, and then inputting the virtual data samples into parallel detection models, wherein a primary layer adopts decomposition condition fuzzy reasoning to perform feature extraction, realizing virtual and real data feature interaction through a feature sharing layer, realizing multi-layer feature extraction according to target function design, and then outputting the grain quantity through a regression layer; in order to further improve the generalization ability, adopting an interval type-2 fuzzy integrated virtual-real interaction parallel detection model, and finally outputting an interval type-2 fuzzy reasoning system driven by real data. The granary grain storage quantity parallel detection method is high in detection result precision and high in model generalization capability, and has important technical application and popularization values.

The invention discloses a waste oil recovery and flow direction supervision method. The waste oil recovery and flow direction supervision method comprises the following steps of S1, scanning a code and registering by a user, uploading equipment operation data, and archiving in a supervision background; S2, collecting waste oil; S3, measuring the content of the waste oil; S4, uploading metering data to the supervision background; S5, dispatching staff to recover the waste oil; S6, confirming the recovery cost; and S7, performing statistical management on the quantity and the amount of the recovered waste oil; According to the waste oil recovery and flow direction supervision method provided by the invention, paid recovery can be carried out on the waste oil after oil-water separation, the willingness of merchants to treat waste by themselves is improved, the recovered waste oil can be used for producing biodiesel, various pressures of municipal environmental sanitation are reduced, theenvironment is greatly protected, the problem of difficult recovery of kitchen waste in small and medium canteens and restaurants is solved, the waste collection times and disposal cost are greatly reduced, supervision background data can also provide remote supervision data for government functional departments, and whether the restaurants perform source processing on waste generated by the restaurants or not is judged.

The invention relates to a method for real tree interactive modeling based on sparse images. The method comprises the steps that an interactive branch extracting and reconstructing module starts from the sequence of n images forming an angle of 90 degrees with an input overlay tree, data of two-dimensional center positions and the diameters of branch nodes of all levels are interactively compiled and stored in the image 1, a branch hierarchical structure is established, the image 1 is searched for the corresponding position of a main branch on the image n with the middle images 2 to n-1 serving as references, the corresponding position of the main branch is marked, interactive compilation is conducted on the image n, and depth information corresponding to the branch node of each level in the image 1 is determined; according to the extracted information of the two-dimensional center positions, the depths and the thicknesses of the branch nodes and based on a branch space bounding box, perspective correction is conducted; a final three-dimensional model is generated by a leaf generating module according to the phyllotaxis rule in the phytology. According to the method, the reconstructed tree model is similar to a real tree form, the topological structure of the branches can be well maintained, and three-dimensional reconstruction of the tree on a common accuracy level is achieved.

The invention discloses a method and a device for carrier aggregation and a terminal device, relating to the communication technology field. The method for carrier aggregation comprises steps of receiving pseudo-distribution carriers issued by a base station, detecting whether each pseudo-distribution carrier is occupied, reporting the detection result of each pseudo-distribution carrier to a base station in order to make the base station determine whether to perform aggregation on the pseudo-carrier as a member carrier according to the detection result. The embodiment of the invention can avoid the interference on invoking on the same carrier.

The invention relates to the technical field of data processing, in particular to a load balancing method and device. The method is applied to a load balancing system, and comprises: when it is detected that the number of cluster nodes changes, dividing the number of current task sets by the number of current nodes to obtain a quotient d and a remainder n; setting the number of bearable task sets of n first-class nodes with the maximum number of currently borne task sets as d+1, and setting the number of bearable task sets of second-class nodes except the first-class nodes as d; and adjusting the task set currently borne by each node, so that the number of the task sets currently borne by each node is equal to the number of the corresponding bearable task sets.