30results about How to "Less input data" patented technology

The invention discloses a method for constructing an atmospheric pollutant diffusion model. The method includes the following steps of establishing three-dimensional space rectangular coordinates, determining factors which influence atmospheric pollution diffusion, simplifying a fundamental form of a Gaussian model, and obtaining the atmospheric pollutant diffusion model. The method can be widely applied to survey regions with various dimensions, and is particularly suitable for medium and small dimensions. Besides, the method is high in computational efficiency, low in requirement to meteorological data, and capable of reflecting a diffusing process of atmospheric pollution to the greatest extent in a simple mode. Under the conditions that a rang is small, a weather field is simple, the underlying surface and topographic conditions are relatively not complex, and secondary pollutant is not considered or processed simply, a concentration field calculated through the method and a concentration field calculated through a complicated numerical method are comparatively fit, errors are mainly derived from input parameters of the model (such as precision degree of pollution source data), the model is simple, input data are less, and programming and developing are convenient.

The invention belongs to the technical field of image information processing, and discloses a fire image recognition method based on depth learning. The method comprises the following steps: collecting smoke pictures and normal pictures in the early fire stage as a training set and a test set of a convolution neural network; Obtaining dark channel images of each image to compose a final training set and a test set; constructing Convolution neural model to determine whether smoke could be detected or not. Obtaining The smoke detection model by training the neural network, and testing the reddest pair model and evulating its performance. Compared with the prior art, the method of the invention improves the accuracy of smoke detection in a single image by utilizing dark channel images and depth learning, simultaneously improves the detection speed, and can be practically applied to the fire detection work of cities or forests.

A method for detecting small objects in a structured road based on deep learning comprises the steps that image data, containing the small objects, on the real structured road are collected, and the positions and the category information of the small objects in the structured road are marked through a manual method; Constructing a deep convolutional neural network suitable for small object detection in the structured road and a corresponding loss function; Inputting the acquired image and the labeled data into the convolutional neural network constructed in the previous step, updating the parameter value in the neural network according to the loss value between the output value and the target value, and finally obtaining an ideal network parameter. The invention provides a brand new network structure for the problem that the current neural network is poor in small object detection. On the premise that the calculated amount is not increased basically, the performance of small object detection is greatly improved, and the method can be conveniently deployed in an existing intelligent driving system, so that an intelligent driving automobile can detect dangerous objects on a road in along distance and respond in time, and the safety in the driving process is improved.

The invention discloses a test-case selection method based on user sessions and a hierarchical clustering algorithm. The method includes the following steps: acquiring server access logs, and carryingout sorting according to time; carrying out preprocessing and clustering to form a user session sequence set; calculating similarity distances among all user session sequences through using an improved user-session-sequence comparison algorithm; employing the improved condensing hierarchical clustering algorithm to cluster the user session sequences, and outputting final clustering results of test cases; and optimizing selection of the test cases through deleting redundant test cases. According to the method of the invention, representative user operation sequences can be quickly mined from the large number of server access logs to use the same as test cases, automation of test-case generation and optimization of test-case selection are realized, and subsequent work of automated functiontests of a server, performance tests, user behavior analysis and the like is facilitated.

The invention relates to the technical field of cooling towers and discloses a three-dimensional design method for a water collection apparatus of a high-level water collection cooling tower. The method comprises the following steps of S1, inputting design parameters of the water collection apparatus; S2, selecting a quadrant arrangement axis net; S3, determining an inner and outer zone boundary on the axis net; S4, arranging structural members on the axis net to form a simplified model with member occupation; S5, arranging the water collection apparatus based on the simplified model; S6, building a water collection apparatus model according to a symmetric relationship of quadrants; S7, generating a drawing and a material statistics report; and S8, generating a video animation. Through the design method, the three-dimensional video animation of a water collection apparatus arrangement scheme can be automatically generated only by inputting necessary basic parameters, so that a large amount of cooling tower design time is shortened in the whole process, the manual input error probability is reduced, and the working efficiency is improved.

The special yarn manufacturing apparatus of the present invention performs the speed change of the front roller based on a plurality of pattern data indicating changes in the thickness of the spun yarn, and manufactures a special yarn corresponding to the above pattern data including at least information on the distance between thick pitches. Wire. The CPU controls the first servomotor in accordance with an execution program for executing a repeating pattern of one cycle consisting of pattern data larger in number than the pattern data based on the pattern data stored in the working memory. , thereby spinning a spun yarn that performs a repeated pattern structure of one cycle.

The embodiment of the invention belongs to the technical field of information, and relates to a cloud native resource dynamic prediction method, which comprises the steps of performing relevancy sorting on acquired to-be-predicted resource data and performance index data based on a Pearson correlation coefficient to obtain a correlation relationship between the to-be-predicted resource data and the performance index data; defining a correlation threshold based on the correlation relationship; taking the performance index data greater than or equal to the relevancy threshold value as performance index time series data; performing transverse data expansion on the performance index time series data to obtain training data and test data; inputting the training data into a constructed time sequence neural network model for training to obtain a trained time sequence neural network prediction model; and inputting the test data into the time sequence neural network prediction model for prediction operation to obtain a resource prediction result. The invention further provides a cloud native resource dynamic prediction device, computer equipment and a storage medium. The prediction complexity can be reduced, and the prediction accuracy can be improved.

The invention discloses an intelligent unattended sensor system, which comprises a power supply conversion unit, and a seismic detection unit, a signal processing unit, a data processing unit and a communication unit which are connected in sequence, the seismic detection unit is used for converting a vibration signal into an electric signal, the signal processing unit is used for converting the electric signal into a digital signal, and the data processing unit is used for analyzing and processing the digital signal to obtain a target identification result. The invention also discloses a target identification method based on the system. The method comprises the following steps: S1, the seismic detection unit is used for converting a vibration signal into an electric signal; s2, the signal processing unit is used for converting the electric signal into a digital signal; s3, the data processing unit is used for processing the digital signal to obtain an identification result; and S4, the communication unit is used for outputting a target identification result. The system has the advantages of flexible layout, good concealment, high recognition rate, multiple recognition types, strong interference resistance, good integration, all-weather work and the like.

The invention relates to a signalized intersection lane queuing length estimation method based on single-section electronic police data. The method comprises the following steps that: (1) the vehicledeparture time headway information of each lane is acquired through electronic police data, wherein the electronic police data is the time stamp information of vehicles passing a stop line, includingdeparture time, direction number, lane number and vehicle model parameters; and (2) according to the timestamp information and a vehicle departure time headway sequence, for a single-lane layout condition and a multi-lane layout condition, a feature function is defined to describe the double characteristics of the close distance and the time headway fluctuation, the change points of the departuretime headway sequence are identified, and the vehicle queuing length and the omission ratio on each lane in the period are estimated. Compared with the prior art, the method considers the error and missing detection phenomena of electronic police equipment, and has the advantages of complete scene coverage, less input data, wide applicability and the like.

The invention relates to the technical field of information processing and particularly to a method for establishing a stock volatility prediction model. The method comprises the following steps of: 1, reading original data from a database, the original data including quotation data and financial statement data relating to stock; 2, constructing a risk factor according to the original data; 3, storing the factor constructed in the second step in a factor database; and 4, establishing a volatility prediction model by using the risk factor determined in step three. The method can perform decomposition according to different factors, can not only predict and quantify the risk and improve the stability of the risk prediction, but also decompose an investment risk into different sources so that investors can choose undertaken risks and avoided risks in a more pertinent way.

The invention discloses a real-time weather prediction algorithm based on an LSTM-DNN (Long Short Term Memory Network-Deep Neural Network) network model, which is mainly applied to speculation of future weather conditions. The algorithm is divided into two networks, namely a deep neural network and a long-short-term memory network. The deep neural network performs manual classification on input data according to various meteorological elements and current weather conditions, and builds a deep full-connection network for speculating weather conditions according to weather elements to classify the input data. The LSTM network carries out normalization and vectorization processing on input data so as to predict change values of various meteorological elements in a future time period. The output of the LSTM network is used as the input of the trained DNN network, and after the two networks are connected to form the LSTM-DNN network, the future weather condition can be speculated based on the current meteorological elements. According to the invention, the LSTM-DNN network model is applied to the meteorological field, and a real-time weather condition prediction method capable of being realized is provided.

The invention discloses a multi-category target recognition method and system for an unattended sensor system. The method comprises the following steps: S1, data preprocessing: framing original data collected by unattended sensor equipment; s2, signal feature extraction: extracting multiple segments of independent time domain features of the framed data at the same time; s3, convolutional neural network model training: taking the multiple segments of independent time domain features obtained in the step S2 as training samples, inputting the training samples into a convolutional neural network for training, and establishing a convolutional neural network model; and S4, signal recognition: recognizing the signal through the convolutional neural network model established in the step S3 to obtain an recognition result. The invention has the advantages of being easy and convenient to operate, small in calculation amount, high in recognition rate, capable of being applied to an embedded system and the like.

The invention discloses an object tracking method, a multi-target tracking model training method and related equipment. The method comprises the following steps: sequentially inputting each image frame of a target video into a multi-target tracking model to obtain target information of a target object corresponding to each image frame; tracking the target object according to the target information between the continuous image frames; the multi-target tracking model is used for performing feature extraction processing on the Nth image frame to obtain a first feature map, performing sampling processing on the first feature map to obtain a second feature map, fusing the second feature map and the target feature map to obtain a third feature map, and performing regression processing on the third feature map to obtain target information; under the condition that N is 1, the target feature map is empty; under the condition that N is greater than 1, the target feature map is a third feature map obtained by processing the (N-1) th image frame by the multi-target tracking model.

A method for estimating the range of a vehicle (1) when driven by means of a power supply is described. The method comprises the steps of: determining a current longitudinal acceleration (23) of the vehicle by triggering a request (21) to update the range estimate; determining a current driving resistance value (25) of the vehicle if the determined magnitude of the current longitudinal acceleration of the vehicle is less than a predetermined limit value; transmitting (13, 31) the calculated current driving resistance value and the current position of the vehicle to the cloud (2); determining acurrent driving resistance value of the plurality of vehicles (1) by means of the cloud, and comparing the same with a standard driving resistance value; determining a correction index (11) based onthe comparison; and calculating a range of the vehicle based on the correction index (29).

The invention discloses a method for calculating the rotor forced excitation hot spot temperature of an air-cooled non-salient pole synchronous phase modifier, and the method comprises the steps of inputting a calculation model and forced excitation parameters, and dividing a model geometry into a discretized heat flow node topological connection diagram; and gradually determining the area, depth,thermal resistance, resistance, heating power and other parameters of each heat flow node, listing the energy conservation equation of each node by taking the node temperature as an unknown variable,forming an equation set, and solving the equation set to obtain the rotor winding forced excitation hot spot temperature of the air-cooled phase modifier. According to the invention, only 24 parameters need to be input, the rotor forced excitation hot spot temperature can be automatically calculated, the heat flow node topological graph can be automatically drawn, the requirements for air coolingnon-salient pole synchronous motor rotor forced excitation hot spot temperature calculation and curve drawing can be well met, and the purpose of conveniently, rapidly and accurately obtaining the air cooling rotor forced excitation hot spot temperature is achieved. The invention has the characteristics of less input data, simplicity and convenience in operation, quickness in calculation and drawing, concise and intuitive heat flow node topological connection diagram and the like.