35results about How to "Solve the technical problem of low detection accuracy" patented technology

The invention discloses a detection method and device for similar video files. The method comprises the following steps of obtaining an amount of key frames with the same fingerprint in the first key frames of a first video file and the second key frames of a second video file, wherein a ratio of the amount to the amount of the second key frames is recorded as a first ratio; obtaining the same video fingerprint sequence in the first audio fingerprint sequences of the first video file and the second audio fingerprint sequences of the second video file, wherein a ratio of the range of an audio fingerprint serial number in the audio fingerprint sequence to the amount of audio fingerprints in the second audio fingerprint sequence is recorded as a second ratio, and a ratio of the amount of audio fingerprints in the audio fingerprint sequence to the range of the audio fingerprint serial number in the audio fingerprint sequence is recorded as a third ratio; and according to at least one of the first ratio, the second ratio and the third ratio, judging whether the first video file and the second video file are similar or not. By use of the method, the technical problem in the relevant art that accuracy for detecting the similar video files is low is solved.

The invention provides a target detection method based on multi-source heterogeneous data cognitive fusion. The method is used for solving the technical problem of low target detection precision caused by a shielded target contained in an image in the prior art, and comprises the following steps of: preprocessing data; obtaining a training data set and a test data set; constructing a target detection model H based on multi-source heterogeneous data cognitive fusion; performing iterative training on the target detection model H based on multi-source heterogeneous data cognitive fusion; obtaining target detection results. On the basis of a target detection network, an input optical remote sensing image and an SAR image are fused, an optical remote sensing feature map and an SAR feature map after feature extraction are fused, and the detection results of the optical remote sensing image and the SAR image are fused, so that a target detection model learns the features of the SAR image andthe features of the optical remote sensing image, and the technical problem that the detection precision of the image containing a shielded target is low is solved.

The invention discloses a blast furnace sinter particle size detection method and system based on RGB and laser feature fusion, and the method comprises the steps: constructing a deep learning network of an encoder-decoder structure through obtaining an RGB data set and a laser data set, obtaining an RGB initial feature tensor and a laser initial feature tensor, constructing a multi-source feature weighted fusion sub-network at the last layer of the fusion branch, carrying out adaptive weighted fusion on the RGB initial feature tensor and the laser initial feature tensor based on the multi-source feature weighted fusion sub-network to obtain a fusion feature tensor, and inputting the fusion feature tensor into a decoding process to obtain a segmented image. The technical problem that the existing blast furnace sinter particle size detection precision is low is solved, the multi-source feature weighted fusion sub-network is added in the deep learning network of the encoder-decoder structure, the adaptive weighted fusion of the multi-source features can be realized, and the detection precision of the blast furnace sinter particle size is improved. The laser supplement effect is fully exerted, and the detection precision of the blast furnace sintered ore particle size is improved.

The invention discloses a face key point detection model training method and a face key point detection method. The training method comprises the following steps: acquiring a face sample image carrying preset face labeling information; inputting the face sample image into a Pnet network of an MTCNN network to obtain a face candidate box corresponding to the face sample image; inputting the face candidate frame and the face sample image into an Rnet network of the MTCNN network to obtain a face target frame corresponding to the face sample image; and inputting the face target frame, the preset face labeling information and the face sample image into an Onet network of the MTCNN network to obtain a trained MTCNN network model, and a key point regression function in the Onet network is a preset WingLoss function, when an existing face key point detection method is used for face detection, the problems of key point drifting is easy to occur, so that the face key point detection accuracy is relatively low are solved.

The invention provides a reagent management device and extraction equipment, and relates to the field of solid-phase extraction. The reagent management device comprises a box body, a reagent container, a weighing assembly and a controller. The weighing assembly is arranged in the box body. The reagent container is placed on the weighing assembly. The weighing assembly is connected with the controller. The weighing assembly of the reagent management device weighs the residual amount of a reagent without being influenced by fluctuations of the liquid level of the reagent, so that the detection precision of the residual amount of the reagent is high. Even if a capacitive liquid level sensor cannot reliably detect or even cannot detect the residual amount of the reagent with a weak dielectricconstant, the residual amount can also be accurately detected.

The invention discloses a message identification system, method and device. The system comprises at least one collector which is used for extracting message features of collected messages, carrying out first feature matching on the message features and pre-stored virus features, and sending the feature matching successful messages to an aggregator; the aggregator which is in communication connection with the collector and is used for receiving the messages with virus features, carries out secondary feature matching on the messages through a preset feature set and sending the feature matching successful messages to a data processor; and the data processor which is in communication connection with the aggregator and is used for classifying the messages on which the secondary feature matching is carried out by the aggregator, and determining classes of terminals which send the messages. According to the system, the method and the device, the technical problem that the detection precision is low is solved.

The invention discloses a method, a device and equipment for detecting a high-voltage line, wherein the method comprises the steps of acquiring magnetic field strengths of a plurality of coils; basedon the magnetic field strength of the plurality of coils, determining the position of a high-voltage line; and displaying the position of the high-voltage line. The method, the device and the equipment settle technical problems of only capability of generating an alarm when a detector approaches the high-voltage line and low detecting accuracy in prior art.

The invention discloses a high-temperature fluid mass flow online detection method, device and system, and the method comprises the steps: synchronously collecting a front image of a high-temperature fluid in a positive direction and a side image of the high-temperature fluid in a side direction, carrying out the image processing of the front image and the side image, extracting a high-temperature fluid flow line according to the side image, according to the high-temperature fluid flow line, establishing prior flow velocity distribution along the high-temperature fluid flow line, establishing a displacement field of a first region of interest by using a cross-correlation method according to the prior flow velocity distribution, and obtaining the high-temperature fluid mass flow according to the high-temperature fluid sectional area and the displacement field of the first region of interest, so that the technical problem that the existing high-temperature fluid mass flow detection precision is low is solved. The sectional area of each point of the high-temperature fluid is calculated in real time by synchronously collecting the images of the high-temperature fluid in the positive direction and the side direction, meanwhile, the flow velocity and the sectional area of an interested area on the high-temperature fluid are cooperatively obtained through the front image and the side image collected in the orthogonal direction, and therefore real-time accurate measurement of the mass flow of the high-temperature fluid is achieved.

The invention discloses a device for fault arc detection and positioning in a mains supply environment. Comprising a high-speed AD acquisition module, a low-speed AD acquisition module, an isolation module, a signal conditioning module, a current transformer module, a switch module, a DA output module, a screen display module, a fault arc positioning analysis module, a fault arc detection analysis module, a fusion analysis module and a fault arc generation module, the DA output module is used for receiving a detection signal and sending the detection signal to the fault arc generation module, the fault arc generation module is used for generating a fault arc and generating a reflection signal when the detection signal encounters the fault arc, and the high-speed AD acquisition module is used for acquiring the detection signal generated by the DA output module and the reflection signal generated by the fault arc generation module. The technical problem that an existing arc fault detection device can only perform fault arc detection and cannot realize fault arc detection and positioning and fault arc type detection at the same time can be solved.

The invention discloses a detection method and device for controlling a driving mechanism and a cooking utensil. The method comprises the steps that in the heating process of the cooking utensil, a driving mechanism is controlled to work so as to drive a pressure limiting valve to act; obtaining change values of cooking parameters of the cooking utensil within a preset time period, the cooking parameters which comprises at least one of temperature and pressure; and determining whether the control driving mechanism works normally or not based on the change value of the cooking parameter. According to the invention, the technical problem of low detection accuracy of the control driving mechanism in the prior art is solved.

The invention discloses a fluid flow velocity measurement method based on a terahertz high-speed echo effect. The method comprises the steps: generating a source signal and a reference signal, detecting the terahertz echo, reflected by a fluid, of the source signal, obtaining a time-domain spectrum brightness deviation value according to the terahertz echo and the reference signal, establishing a time-domain spectrum brightness deviation-flow velocity relation model according to different time-domain spectrum brightness deviation values corresponding to different flow velocities, and obtaining the flow velocity of the fluid according to the time-domain spectrum brightness deviation-flow velocity relation model and the time-domain spectrum brightness deviation value corresponding to the to-be-detected fluid. According to the invention, the technical problem that the existing non-contact fluid flow velocity device has low detection precision on the flow velocity of the fluid in a complex and severe environment is solved; the characteristic of high penetrability of terahertz waves is ingeniously utilized, and terahertz signals can be received in a complex environment, so that the flow velocity of the measured fluid can be accurately measured; and the non-contact characteristic is achieved, so that great significance is achieved for flow velocity measurement of the fluid in the complex environment.

The invention discloses an image processing method and device, a storage medium and a computer terminal. The method comprises the steps that a target image is acquired, and the target image comprises a target object; and detecting the target image by using a target detection model to obtain a detection result of the target object, the target detection model being obtained by training a pre-training model through a first training sample, the pre-training model being obtained by training the first training sample, and the first training sample being obtained by performing data enhancement processing on an initial training sample. The technical problem of low detection precision of a detection model obtained by training due to few training samples and low sample precision in related technologies is solved.

The invention provides a detection device for a lathe workpiece hole. The detection device comprises a supporting rod, a dial indicator, and a seat body, and further comprises an ejector rod mechanism, the supporting rod is provided with a side opening cavity and a first through hole, wherein the side opening cavity comprises a first cavity body and a second cavity body communicating with the first through hole, the cavity diameter of the first cavity body is larger than that of the second cavity body, a detection rod stretches into the first through hole in a sliding mode, one part of an ejector rod is internally arranged in the second cavity body, the side wall of the ejector rod is in contact fit with the head of the detection rod, the other part of the ejector rod sequentially penetrates through a sleeve embedded in the first cavity body and a locking nut arranged on the side wall of the supporting rod to be in contact fit with the workpiece hole to be detected, two spherical screws are symmetrically arranged on the side wall of the sleeve, the spherical portions of the spherical screws abut against the inner wall of the sleeve, and the spring sleeves the side wall of the portion, between the sleeve end close to the dial indicator and the wall of the second cavity body, of the ejector rod. Compared with the prior art, the detection device has the advantages that the detection precision is improved, meanwhile, the operation is convenient, and the detection efficiency is improved.

The invention discloses an industrial control system traffic anomaly detection method based on an OCSVM and a K-means algorithm, and the method comprises the steps: obtaining a normal traffic data set from an industrial control system through a traffic obtaining tool, carrying out the protocol analysis of the traffic data for each piece of traffic data in the traffic data set, so as to obtain the attribute of the traffic data and the attribute value, wherein all attribute values corresponding to the traffic data form attribute value vectors, the attribute value vectors corresponding to all the traffic data form a to-be-detected data matrix with the size of m * n, and dimension reduction processing is performed on the to-be-detected data matrix by using a principal component analysis (PCA) method to obtain a matrix after dimension reduction processing; and inputting each row in the matrix after dimension reduction processing into the trained single-class support vector machine OCSVM detection model to obtain a label value corresponding to the row, wherein the label values corresponding to all rows in the matrix form a label value column vector. According to the invention, the technical problem that the abnormal traffic is less or difficult to obtain in the existing traffic anomaly detection method of the industrial control system can be solved.

The invention discloses an Internet of Things equipment access detection method based on an SMOTE and a parallel random forest. The method comprises the following steps: obtaining a plurality of pieces of equipment fingerprint information of Internet of Things equipment, analyzing each piece of equipment fingerprint information to obtain a feature attribute value corresponding to the equipment fingerprint information, and constructing a feature matrix according to the feature attribute values corresponding to all the obtained equipment fingerprint information; inputting the feature matrix into a trained equipment identification classifier to obtain an output result, and querying a corresponding vulnerability and a type thereof in a local vulnerability library according to the output result; and judging whether the queried vulnerability is a medium-high risk vulnerability or not according to the type, and if so, forbidding access of the Internet of Things equipment, otherwise, allowing the access of the Internet of Things equipment. According to the invention, the technical problem that the detection accuracy of uncommon Internet of Things equipment is low due to the fact that Internet of Things equipment is diversified in variety and large in difference in an existing Internet of Things equipment access detection method based on machine learning or deep learning can be solved.