51results about How to "Improve recognition quality" patented technology

An arrangement for the optical evaluation of harvested crop in a harvesting machine includes a camera located for taking images of threshed out crop elements (grain and material other than grain) being conveyed within a channel of the machine. An electronic image processing system receives the images and identifies individual objects in the images, these objects being classified into predetermined object classes by way of comparing at least one of color, contour or texture features of the individual objects and corresponding characteristics of reference objects filed in a data bank. The processing system then determines the absolute or relative numbers or areas of the individual objects assigned to the respective object classes.

In a method and a device for traffic sign recognition, at least one significant feature for a traffic sign is determined which is standardized for a region. The region that corresponds to the determined feature is determined. At least one classification feature and / or at least one classification method is defined depending on the determined region for the recognition of the traffic sign and / or at least one further traffic sign. The recognition of the traffic sign and / or the at least one further traffic sign is performed by using the defined classification feature and / or the defined classification method.

The present invention provides a method and system to improve speech recognition using an existing audio realization of a spoken text and a true textual representation of the spoken text. The audio realization and the true textual representation can be aligned to reveal time stamps. A speech recognition can be performed on the audio realization to provide a hypothesis textual representation for the audio realization. The aligned true textual representation can be compared with the hypothesis textual representation. Single word pairs from the true and the hypothesis textual representations can be selected where the representations are different. Similarly, single word pairs can be selected from each representation where the representations are identical. A word or pronunciation database can be updated using the selected single word pairs together with the corresponding aligned audio realization.

The invention provides a multi-label anti-collision arithmetic in a high frequency / super high frequency RFID identification system, the arithmetic comprises the steps that: resetting; a read-write machine sends a label inquiring order; a responder judges the received inquiring order; the read-write machine receives and judges the label inquiring response; the read-write machine processes the data of the responder; pd value is set as 1, Q value is set as Q; the read-write machine sends an inquiry variation order and the responder returns, the read-write machine receives and judges the label inquiring response; the read-write machine processes the data of the responder; the arithmetic can circulate in a manner of rapid elimination and has high practical identification efficiency to the labels taking place the collision, which can rapidly reduce the number of responders within the work range of the read-write machine; in one hand, the identification efficiency is enhanced by rapidly reducing the number of the responder, in other hand, the identification quality can be further promoted by diminishing the disturbance of electromagnetic environment within the work range of the read-write machine.

An arrangement for the optical evaluation of harvested crop in a harvesting machine includes a camera located for taking images of threshed out crop elements (grain and material other than grain) being conveyed within a channel of the machine. An electronic image processing system receives the images and identifies individual objects in the images, these objects being classified into predetermined object classes by way of comparing at least one of color, contour or texture features of the individual objects and corresponding characteristics of reference objects filed in a data bank. The processing system then determines the absolute or relative numbers or areas of the individual objects assigned to the respective object classes.

The invention provides an iris algorithm based electric power security information identification and intelligent management system. The system comprises an identification terminal located on an operating condition site and a remote management module connected with the identification terminal, the identification terminal comprises a personnel identification unit, a tool and appliance identification unit and an alarm unit, the remote management unit comprises a controller and a storer connected with the controller, and the storer is connected with the alarm unit. The personnel identification unit comprises an iris image processing based video monitoring module, and the tool and appliance identification unit comprises a scanner, an identification tag and a reader connected with the scanner. The iris biometric identification technique and the RFID (radio frequency identification) barcode technique are combined with wireless network applications, users are provided with accurate and quick personnel and capital equipment data collection, statistics and management, and personnel entry-exit management in important places and real-time track management including tracking, positioning, maintenance and the like on capital equipment in the whole life cycle of the capital equipment are realized.

The invention provides a method for identifying a character in an image. The method is characterized by comprising the following steps: S1, longitudinally projecting a binary image acquired by the image; S2, segmenting the image for the first time according to the longitudinally projected projection value and a first preset threshold value to acquire character width; and S3, segmenting the image for the second time according to the character width. By the method for identifying the character in the image, the character can be identified from the image well and accurately, the identification quality is high, the identification process is easy to implement, and system resources are saved.

The invention discloses an image search method based on deep learning. The image search method greatly enhances the search efficiency through information input, searching strategies, search, result discrimination and output to a GIS database. A traditional image search method is low in query efficiency and reduces work efficiency because collected sample data is huge and the bandwidth and the scaling capacity are limited. The disclosed image search method has the advantages that the accuracy in the image search process and recognition quality are improved through deep learning; the searching strategies include distance and road preceding-following relations, and images can be searched according to specific situations; under the situations like dense roads, the road preceding-following relation search strategy is adopted, the distance search strategy is used in open regions, the quantity of collected data samples are greatly reduced, and the search speed is increased; distributed-type search and centralized-type search can be utilized for the image search, the appropriate search approach is selected according to actual conditions, accordingly the bandwidth and the scaling capacity are saved, and the query efficiency is greatly improved.

The invention provides a data information management system for power equipment, and the system comprises a data collection end disposed on the power equipment, a management system and a processing system, wherein the management system and the processing system are respectively connected with the data collection end. The data collection end comprises an equipment parameter collection terminal anda video information collection terminal, which are disposed on the power equipment. The management system comprises a control host, a data concentrator and a data exchanger, wherein the data concentrator and the data exchanger are connected with the control host. The processing system is of a multilayer processing structure, and comprises an equipment layer, a network layer, a scheduling layer anda transmission layer, wherein all layers are respectively provided with an independent information preprocessing unit. The system can meet the development demands of information integration of powerequipment, can enable terminal equipment which cannot achieve the mutual data direct connection and data unified communication previously to achieve the autonomous call and communication function, achieves the communication of power system equipment and expandable cascading of the data collection design, and is worth of being widely popularized and used.

In a method and a device for traffic sign recognition, at least one significant feature for a traffic sign is determined which is standardized for a region. The region that corresponds to the determined feature is determined. At least one classification feature and / or at least one classification method is defined depending on the determined region for the recognition of the traffic sign and / or at least one further traffic sign. The recognition of the traffic sign and / or the at least one further traffic sign is performed by using the defined classification feature and / or the defined classification method.

A method of identifying stricken-out characters in handwriting, comprising parsing a scanned image into regions and objects, defining objects containing handwritten characters, applying structural or feature classifiers for primary character recognition, applying one or more supplemental feature classifiers preliminarily trained by strike-out characters, and identifying a stricken-out character if any. The stricken-out character may be further examined by special procedures, either automated or manual.

The invention provides a method for automatically recognizing taxus cuspidata in high-resolution remote sensing images. The method has the advantages that only few ground samples are needed for high-accuracy recognition of the taxus cuspidata in the high-resolution remote sensing images, the advantages of low cost and wide coverage of the remote sensing images are fully utilized, the taxus cuspidata can be recognized from the remote sensing images on the basis of the few samples according to its unique features, high recognition quality is acquired, and accordingly, high social and economic value is achieved.

The invention relates to an automatic detection method for an airport of a large-size space-borne SAR image, which comprises the following steps of: A, performing SAR image segmentation based on a maximum posterior probability criterion; B, extracting interesting targets according to the pixel intensity and the area of the airport; C, extracting dimensions and rotary invariable parameters of the interesting targets; D, training a support vector machine classifier according to the dimensions and the rotary invariable parameters of the interesting targets, and classifying the interesting targets to obtain an initial detection result; and E, performing parallel runway detection on the initial detection result, and retaining the targets with parallel runways as final detection results. Compared with the prior art, the method has the advantages of high identification speed, good identification quality and high identification accuracy rate.

The invention relates to the technical field of bar code recognition, in particular to a bar code-based automatic recognition method, which comprises the following steps of: acquiring bar code data of a bar code acquisition image; based on the barcode data, acquiring an identification value of the barcode acquisition image, and determining the quality of the barcode acquisition image; if the identification value of the bar code acquisition image is greater than or equal to the identification threshold value of the bar code acquisition image, generating a bar code identification qualified signal; determining an optimal bar code based on the data qualified bar code acquisition image and the change qualified bar code acquisition image, and completing automatic identification work through a bar code identifier; according to the invention, the bar code acquisition quality can be effectively evaluated, the qualified bar code image can be determined, the bar code can be better identified, and the bar code identification quality can be greatly improved.

The invention provides a power transformation equipment data management system based on image information recognition, and the system comprises an image information recognition module disposed on power equipment, and a terminal machine module connected with the image information recognition module. The image information recognition module comprises an image information unit, a management information unit and a fault alarm unit, wherein the fault alarm unit is connected with the image information unit and the management information unit. The image information unit comprises a visible light camera and a multispectral sensor group based on spectral image processing, and the management information unit includes a scanner and a read / write module connected with the scanner. The system can employa multispectral biological recognition technology system, which can be connected with a central control system host of a transformer substation and can achieve the bidirectional communication, the equipment information recognition technology and the access control technology for the identity recognition of a persona and the equipment information recognition through a spectrum algorithm, avoids the illegal access, and achieves the management and control of equipment.

A keyword detection method includes: obtaining an enhanced speech signal of a to-be-detected speech signal, the enhanced speech signal corresponding to a target speech speed; performing speed adjustment on the enhanced speech signal to obtain a first speed-adjusted speech signal having a first speech speed, the first speech speed being different from the target speech speed; obtaining a first speech feature signal according to the first speed-adjusted speech signal; obtaining a detection result according to a first keyword detection result corresponding to the first speech feature signal, the detection result indicating whether a target keyword exists in the to-be-detected speech signal; and performing an operation corresponding to the target keyword in response to determining that the target keyword exists according to the detection result.

The invention discloses a turnout switch point closeness detection method based on an image detection technology. The method comprises the following steps: identifying a switch point: tracking the switch point; identifying a fork tip edge; and comparing the fork tip motion trail with the normal motion trail, and carrying out fork tip closeness state detection. The invention provides the turnout junction point closeness detection method based on the image detection technology, non-contact real-time nondestructive detection of turnout junction closeness is realized, the detection efficiency andaccuracy are high, and the turnout junction point closeness detection method is fast and efficient and is very worthy of popularization.

The invention provides an intelligent recognition method and system for an ultra-large format document based on block parallel. The method comprises the following steps: acquiring image data of a to-be-recognized document; a self-adaptive blocking strategy is adopted, dynamic blocking is conducted according to an available memory of a system or image features, a certain overlapping area is kept between every two adjacent blocks, and then blocking tasks are converted into an asynchronous task queue; using a thread pool to process each block in parallel, or adopting a distributed computing architecture to distribute tasks to a plurality of computing nodes for execution; detecting a text region and a table structure, and performing text recognition on the detected text region through a text recognition model; processing results of all the blocks are collected and combined according to the recorded position information; and performing confidence scoring according to an identification result output by the text identification model to optimize a text region, and performing visualization processing and layout analysis on the optimized text to generate an output result in a standard format.

In a method for operating a fuel injection device of a motor vehicle, fuel is conveyed under high pressure into a high-pressure reservoir, the pressure of which is controlled with the aid of a pressure regulation and from which fuel is taken for injection via at least one fuel injection valve under the control of a control device. A control pressure deviation of the pressure in the high-pressure reservoir is detected and the existence of a balance disturbance between quantity of fuel supplied to the high-pressure reservoir and quantity of fuel taken from the high-pressure reservoir is deduced from the control pressure deviation. If a balance disturbance exists, operational influencing of the fuel injection device counteracting the balance disturbance is produced and / or a corresponding report is generated.

A data property recognition apparatus, includes a storage unit; a model data acquisition processor acquiring a plurality of model sets of data entries, each data entry individually representing an identified property common to the model set and being of a data type common to the model set; a feature vector generation processor receiving an input set of data entries, recognizing a data type common to the input set of data entries from among a plurality of supported data types, selecting a set of statistical characteristics representing the input set of data entries in dependence upon the recognised data type, generating a value of each of the selected set of statistical characteristics from the input set of data entries, and outputting a feature vector composed of the generated values of statistical characteristics.

The invention provides a fingerprint sensing device. A control circuit controls a part of point light sources to irradiate a fingerprint of a user. Reflected light generated by using the plurality of point light sources to irradiate the fingerprint of the user forms a light-emitted pattern including a plurality of reflected light patterns on a sensing layer, wherein each of the reflected light patterns is provided by a corresponding point light source, and each of the sensing units senses the reflected light patterns corresponding to at least two point light sources.

The invention discloses a machine and depth vision fusion-based weld ultrasonic image defect identification method. The method comprises the following steps of performing ultrasonic imaging on a to-be-detected object by utilizing ultrasonic detection equipment; a defect labeling model is used for automatically positioning a welding seam area of the ultrasonic detection image; obtaining texture feature data of a weld seam image in the defect labeling image by using an image recognition technology, and performing category recognition on the texture feature data through a collaborative algorithm; and outputting an identification result to a front-end interface, and automatically describing the shape of the weld joint by utilizing cartographic software to realize weld joint defect identification of the to-be-detected object. According to the method, the defect area can be automatically positioned, various weld defects can be identified, the weld defect identification quality can be effectively improved, the identification speed can be increased, the dependence on artificial experience can be effectively reduced, the labor cost can be reduced, the identification process can be visualized step by step, the identification quality can be improved, the identification process can be obviously accelerated, and the identification efficiency can be improved. And the robot welding process speed is met.

The invention relates to a fatigue analysis platform for an audience of an electronic reader. The platform comprises a face image capture device embedded in a front frame of the electronic reader andused for performing image data capture on the face of the audience of the electronic reader to obtain and output a high-definition face image, an electronic ink touch screen fixedly mounted in the front frame of the electronic reader and used for displaying related contents of e-books stored in the electronic reader and displaying text contents corresponding to an eye fatigue level of the audiencein real time when receiving the eye fatigue level of the audience, and a stereo speaker embedded in the front frame of the electronic reader and used for playing a corresponding fatigue alarm voice file when the received eye fatigue level of the audience exceeds the limit. Through the platform, the harm on the eyes of the audience due to excessive reading can be effectively avoided.

The invention belongs to the field of fish food deep processing, and discloses an automatic counting detection system, a fish body quality prediction method and a terminal in the fish deep processing process, workbenches are located on the two sides of a first conveying belt, and water faucets for flushing waste generated on the workbenches at any time are installed on the workbenches; a button for controlling the speed of the first conveying belt is installed on the workbench and used for regulating and controlling the speed of the first conveying belt, and processed objects enter the funnel-shaped counting system. And the objects counted by the funnel-shaped counting system are orderly arranged on the second conveying belt and enter a subsequent process. The counting device is accurate in counting and is not influenced by external temperature, light and humidity. The detection method disclosed by the invention is simple and high in precision, and meets the requirement of on-site rapid detection in a fish deep processing process; the future quality of deep-processed food can be predicted according to fish body detection information, and the method is more practical and can be widely applied to the field of fish deep processing management.

The invention provides a grape stem accurate identification method based on deep learning post optimization. The grape stem accurate identification method comprises the steps: 1, collecting and marking a grape stem picture; 2, identifying the marked grape fruit stem picture through a deep learning model Mask R-CNN to obtain a pre-identified grape fruit stem area; 3, according to the pre-recognizedgrape stem area, obtaining grape stem contour coordinates and grape stem area coordinates; 4, obtaining H, S and V channel values of the grape stem contour and H, S and V channel values of the grapestem area; 5, respectively calculating an H channel average value, an S channel average value and a V channel average value of the grape stem area; 6, selecting the grape stem contour as a seed point,and carrying out regional growth according to a regional growth criterion based on the seed point. The method can achieve the recognition of the grape stems in a natural environment, and solves a problem that the grape stems are difficult to recognize under a natural orchard scene because of the change and shielding of light.

The invention discloses a communication enterprise group customer member identification method, which comprises the steps of performing feature engineering according to basic training data to obtain key features; selecting training data from the basic training data according to the key features, and training an identification model by using the training data; evaluating an identification effect ofthe identification model according to a preset index, and adjusting model parameters according to an evaluation result until the evaluation result reaches a preset value so as to obtain an optimal identification model; and according to the key features, performing communication enterprise group customer member identification on newly added service data by utilizing the optimal identification model. According to the invention, the key features which have important influence on identification are obtained through feature engineering, the blindness of feature selection is avoided, and the identification quality, performance, effect and identification efficiency are further improved. The invention further discloses a storage medium and computer equipment.

The invention relates to a color difference analytical device based on an image recognition technology. The color difference analytical device comprises a rack and is characterized in that a hollow photo studio is arranged on the rack, a digital camera for shooting workpieces is arranged in the photo studio and connected with an image analysis device for image color difference analysis, and a feeding device for feeding the workpieces into the photo studio and a discharging device for discharging the workpieces are disposed on a table. According to the color difference analytical device, defects in the background art are overcome, color difference analysis of digital images is performed on the workpieces by using the digital image recognition technology, and automatic feeding and discharging devices are designed for automatic feeding and classified discharging work after recognition. Compared with the existing detection means, the color difference analytical device has the advantages of high efficiency and recognition quality and the like.