154results about How to "Realize precise identification" patented technology

The invention discloses an improved weighting region matching high-altitude video pedestrian recognizing method (KS-WRM (KS-weighting matching region)). The improvement is reflected as follows: 1) confirming a candidate region by utilizing Kalman filtering; 2) fixing a camera or not by adopting different pedestrian detecting strategies; 3) proposing a pedestrian detecting method of an HLS (hue, lightness and saturation) model with multi-characteristic fusion; and 4) dividing pedestrians based on CA obvious region detection of context sensing. According to the invention, pedestrians can be recognized accurately under complex occasions of blur details, complex backgrounds, noise influences and the like. And the improved weighting region matching high-altitude video pedestrian recognizing method (KS-WRM) provided by the invention can be used for pedestrian detection and recognition at places with high visitor flow rate, such as remote sensing satellite images, malls, subway stations, railway stations, airports, as well as intelligent traffic control, intelligent vehicle auxiliary driving, pedestrian flow rate statistics and analysis places.

The objective of the invention lies in providing a robot charging method and device. The method comprises the steps: enabling robot to move to a position nearby a potential or registered charging pile according to an environment map and the self-positioning information of the robot, i.e., the posture of the robot in the environment map; carrying out the linear extraction of structural data of an identification through a laser collection device after the robot moves to a short-range pile-searching butt joint region, recognizing the charging pile through combining with the preset identification template structural data, and carrying out the butt joint of the charging pile with the robot for charging when the collected structural data of the identification and the identification template structural data meet a preset matching degree; carrying out the physical connection confirmation according to the voltage condition after the connection of the robot and the charging pile, and guaranteeing the charging correct behavior. The method can achieve the precise recognition of the charging pile, is high in efficiency, is reliable, and effectively irons out defects of a series of pile butt joint jittering, interference and wrong recognition caused by the sensitive characteristics of hardware and the environment interference in the prior art.

The invention discloses a precise recognition method of overlapped shielded fruits, which concretely comprises an image acquisition step, a target object extraction step, a target object depth calculation step, a shielded edge area determining step, a shielded edge information obtaining step, an overlapped shielded fruit separating step, an incomplete target fruit repairing step and a fruit centroid and depth coordinate calculation step, wherein in the target object extraction step, an acquired image is processed to extract fruits, branches and leaves in the image, in the shielded edge information obtaining step, complete information of a shielded edge is obtained through methods of edge detection, edge thinning and edge connecting, and in the overlapped shielded fruit separating step, an overlapped shielded complete edge image is overlapped with a fruit image after being perverted to realize separation of overlapped shielded fruits. For picking robots of fruits, such as apples and citrus, the precise recognition method of the overlapped shielded fruits disclosed by the invention can realize precise recognition for the overlapped shielded fruits, and thus the practical progress of the picking robots is expected to be pushed.

The invention discloses a part defect prediction method and device and electronic equipment, and relates to the fields of artificial intelligence, deep learning, cloud computing and computer vision, in particular to the aspect of industrial quality inspection. According to the specific implementation scheme, the method includes: obtaining plane images of multiple planes of the part; cutting the plane image according to a reference size to obtain a plurality of sub-images; inputting the plurality of sub-graphs into a pre-trained defect prediction model to obtain a defect prediction result of each sub-graph; and inputting the defect prediction result into a pre-trained defect identification model to obtain a defect identification result of each sub-graph, the defect identification result comprising a defect type and a defect grade. Through the above scheme, on one hand, accurate identification of real defects of the part can be realized, so that the accuracy of a defect detection resultis improved, and the probability of misjudgment on whether the part has defects or not is reduced; and on the other hand, the input data of the defect prediction model is convenient to process, and the labor cost can be reduced.

The invention discloses a message type identification method, which core idea is to match the preserved UDP port number of RTP based on the application with the port number carried in the UDP herder of the input text, if they match, then the input text is identified to be RTP packet based on the application of UDP, if not match, then it is identified to be the pure UDP packet or other agreement packets based on UDP application. The invention also discloses packet type identification device corresponding to the method. The invention discloses a method and system to compress the header of packet based on the above packet type identification method.

The invention discloses a digital picture book rapid identification method and a system, a computer readable storage medium; the method comprises the following steps: pre-extracting cover feature points of digital picture books stored in a digital picture book stock, and carrying out quantification clustering; extracting cover feature points of a to-be-identified digital picture book, and carryingout quantification clustering; using a cluster-based cover feature point coupling algorithm to identify the digital picture book in the digital picture book stock; if the highest first coupling rateis between the first and second thresholds, selecting a plurality of digital picture books from the digital picture book stock so as to form a candidate digital picture book set; using a non-linear transformation-based cover feature point coupling algorithm to identify digital picture book in the candidate digital picture book set. The method combines the cluster-based cover feature point couplingalgorithm with the non-linear transformation-based cover feature point coupling algorithm so as to identify the digital picture book in the digital picture book stock, thus fast and accurately searching digital picture books.

The invention provides an industrial process fault diagnosis method based on similarity local spline regression, and relates to the technical field of fault monitoring and diagnosis. The method includes the steps of collecting industrial process data, performing partial labelling and standardization processing, obtaining forecast labels by using an LSR method, processing the forecast labels by using a similarity analysis method, performing label correction on fuzzy fault identification points, then constructing an online diagnosis model based on a spline function, obtaining a coefficient matrix by using a ridge regression method, collecting new data during the industrial production process, obtaining the corresponding labels through the online diagnosis model and the coefficient matrix, and performing fault diagnosis. The industrial process fault diagnosis method in the invention solves the problem in fault diagnosis of multi-source heterogeneous data including a large number of physical and chemical variables, images, audio, video and the like in an industrial production process, without the need for fault identification of all industrial production data, thereby saving a lot of time and manpower, and can greatly reduce false alarms and improve the accuracy and sensitivity of fault detection.

The embodiment of the invention discloses a processing method, a processing device and a processing system for an engineering event. The method comprises the following steps of: extracting event attribute information from an obtained event; acquiring event source information for the production of the event from the event attribute information, wherein the event source information comprises the sub-object information of network equipment; and searching for rule item records matched with the event source information from configured engineering event rules by taking the event source information as an index condition, and if the matched rule item records are found, identifying the event as the engineering event. Therefore, the event status of the network equipment is subdivided to a sub-object of the network equipment, and the event produced by the sub-object of the network equipment is identified to realize the accurate identification of the engineering event.

The invention relates to a point cloud and multi-view fused vehicle-mounted laser point cloud multi-target identification method, which comprises the steps of constructing a depth model PGVNet based on an independent point cloud object to perform surface feature category prediction by performing point cloud local feature extraction on the independent point cloud object by utilizing a point cloud feature extraction module; generating a multi-view image of the independent object, and extracting an optimal view feature by using a view feature extraction module through view grouping and group feature fusion; fusing the optimal view feature and the point cloud feature by using a point cloud view feature fusion module based on an attention mechanism to obtain a point cloud global feature fused with attention; and finally, using the classifier MLP is to predict the category of the independent ground object target on the vehicle-mounted laser point cloud surface. According to the method, on one hand, the problem of information redundancy between similar views is reduced, on the other hand, the optimal view features can be used for guiding the model to learn point cloud local features, themodel classification precision is improved, and a new research method is provided for vehicle-mounted laser point cloud roadside multi-target fine classification.

The invention discloses an asset management method, device and equipment, and a readable storage medium. The method comprises the steps of obtaining traffic information of traffic accessing a target server; analyzing the traffic information to obtain access asset information corresponding to each traffic; aggregating the access asset information based on a destination IP address, a destination port and a transport layer protocol type, and obtaining statistical information of assets corresponding to the target server, wherein the statistical information comprises access times corresponding to each asset. According to the method, the access asset information is aggregated by analyzing the traffic accessing the target server and taking the IP address, the port and the transport layer protocol type of the target server as keys, so that the access condition of the assets accessed by the target server can be aggregated, and the statistics of the asset use condition of the target server is realized; and the assets of the target server are identified by using the IP address, the port and the transport layer protocol type, so that accurate identification of the assets in the data center is realized.

Embodiments of the invention provide an identity recognition method and apparatus, a storage medium, and electronic equipment. The method includes: outlining a face image from a target image; performing super-resolution on the face image to obtain a face image reaching a preset pixel value; and performing identity recognition on the face image reaching the preset pixel value. Through super-resolution of the face image and identity recognition of the face image after super-resolution, the probability of identity recognition of the face image is increased, and accurate recognition of the face image in the target image is realized.

The invention relates to the field of intelligent video analysis, in particular to a distributed intelligent video analysis system, method, apparatus, equipment and storage media, The system comprises: an intelligent video analysis platform and a plurality of local video processing devices, a local video proces device comprises a front-end and an IP video source, the IP video source is used for collecting or storing the IP network video stream which needs intelligent video analysis. The front-end computer is used for acquiring the IP network video stream which is output from the IP video source, identifying the IP network video stream, and sending the initial identifying information to the intelligent video analysis platform. The intelligent video analysis platform is used to receive the preliminary identification information, process the preliminary identification information based on the depth learning algorithm, obtain the identification result, and send the identification result back to the front-end computer. With the distributed deployment of low-cost front-end, the video is accurately identified, and at the same time, the cost is greatly saved, the performance of intelligentvideo analysis is fully utilized, and the performance resources of video processing are effectively saved.

The invention provides a method for detecting the number of people in a target place, a recommendation method, a detection system and a medium. The method for detecting the number of people in the target place comprises the steps of obtaining an environment image of a target place; determining a target detection area in the environment image; determining a face human head target frame and a non-face human head target frame in the target detection area; and according to the face head target frame and the non-face head target frame of each target detection area in any time period, determining the number of people going in and out of the target place in the time period. According to the invention, accurate identification of the head target frame in the target detection area is realized, and the number of people going in and out of the target place in any time period can be accurately determined.

The invention provides a complex scene-oriented license plate recognition method, which is used for performing license plate recognition on a to-be-detected image in a complex scene so as to recognizea character recognition result of a license plate, and is characterized by comprising the steps of S1, preprocessing the to-be-detected image to obtain a preprocessed image set; S2, building a YOLOv3license plate detection model; S3, training a YOLOv3 license plate detection model of the license plate detection training set; S4, building an end-to-end license plate recognition convolutional neural model; S5, training an end-to-end license plate recognition convolutional neural model; S6, inputting the preprocessed image set into the trained YOLOv3 target detection model to perform license plate target detection so as to obtain a license plate position; S7, processing the corresponding to-be-detected images according to the license plate positions in sequence to obtain a license plate image set; and S8, inputting the license plate image set into the trained end-to-end license plate recognition convolutional neural model to obtain a character recognition result of the license plate.

The invention discloses a construction method and device and a recognition method and a recognition device for a deep learning recognition model of an inclined license plate. The construction method comprises the steps: determining a license plate coordinate from a collected license plate image, and calculating an affine parameter; constructing a deep learning network framework for identifying theinclined license plate; and training a positioning network by using the collected data set, and training a license plate character recognition network through the trained parameter model and the trained license plate data set. The invention provides a recognition network framework based on a deep learning method for inclined license plate recognition, and the technical effect of greatly improvingthe recognition precision of the inclined license plate can be achieved.

The invention relates to a device for accurately identifying the semi-active laser target azimuth, which belongs to the technical field of laser. The device comprises an optical system and an array detection system, wherein the optical system comprises a concavo-convex lens, a biconvex lens, a first convex-concave lens and a second convex-concave lens; the four lenses are coaxial; the array detector is positioned on a focal plane of combined focal length of the four lenses; the center of a detection surface of the array detector is positioned on optical axes of the four lenses; optical energyis focused on the array detector through the optical system; the array detector is used for converting the received optical signal into an electrical signal; and the optical system is used for ensuring that light with different angles is incident to corresponding pixel positions according to difference of performance parameters of the array detector. According to the device disclosed by the invention, the aim that reflection lasers with different angles have signal response at corresponding positions of the array detector is achieved by using the design of the optical system and the selectionof the array detector, thereby the accurate identification of the target azimuth is achieved; and the device is applied to military affairs so as to improve accurate strike capability of a weapon system.

The invention provides a three-dimensional composite imaging method of a hyperspectral laser, a system and an unmanned self-propulsion device. The method includes steps of scanning an underwater suspicious zone through hyperspectrum so as to determine a detecting target in a suspicious zone; scanning the detecting target through a laser, and acquiring a backward scattering signal; confirming the distance information with the detecting target through the backward scattering signal; preforming interference shielding treatment on intensity information, spectrum information, and polarization information of the backward scattering signal, and acquiring the position information of the detecting target after analyzing the distance information with the detecting target; according to the intensityinformation, spectrum information, and polarization information of the backward scattering signal after removing interference, generating a three-dimensional image of the detecting object. Through combination of active laser scanning and hyperspectral passive imaging, the polarization information and spectral information of the detecting target are extracted; through information fusion of the both, speciality inversion of the small underwater target is completed, thereby realizing the accurate identification and high-precision location of the detecting target under the water.

The invention discloses a shelf commodity display information identification method, device and equipment and a storage medium, and the identification method comprises the steps: extracting the angle information of shelf commodities from a target image, correcting the target image based on the angle information of the shelf commodities, and obtaining a corrected image; and carrying out shelf commodity information identification according to the corrected image to obtain commodity display information in the target image. According to the method, the target image is corrected to obtain the corrected image at the standard angle, so that the influence of different shelf postures on the commodity display recognition accuracy can be avoided; and meanwhile, accurate identification of image commodity display information is realized according to the corrected image, and an accurate data basis can be provided for subsequent intelligent distribution display recommendation.