49results about How to "Narrow down the range of recognition" patented technology

Embodiments of the present application provide a new word recognition method, apparatus, computer device, and computer-readable storage medium. The method comprises the following steps of: acquiring atext corpus; according to a preset sentence endpoint, dividing the text corpus into pieces with a length of 2-N is a natural number, N >= 2; judging whether a candidate word satisfies a preset condition; determining a candidate word as a candidate new word if the candidate word satisfies a preset condition; judging whether the candidate new words are included in the preset vocabulary; and determining the candidate new word as a new word if the candidate new word is not included in the preset vocabulary. Embodiments of the present application are based on natural language processing, and textcorpus is accurately segmented through preset sentence endpoints to obtain candidate words, thereby improving segmentation accuracy, identifying new words through identifying candidate words and candidate new words, and effectively improving accuracy and efficiency of new word discovery.

The invention relates to the field of teaching, and discloses an online classroom video knowledge point identification method and device, which can accurately identify all knowledge points contained in a classroom video. The method comprises the following steps: determining a knowledge point domain of a classroom video; acquiring a classroom video, which comprises a teaching courseware part and ablackboard writing part; for the teaching courseware part, dividing the teaching courseware part into multiple sections of videos according to courseware page switching time points, and identifying afirst set of knowledge points contained in each section of video based on the knowledge point domain; for the blackboard writing part, dividing the blackboard writing part into multiple sections of videos according to blackboard writing updating time points, and identifying a second set of knowledge points contained in each section of video based on the knowledge point domain; dividing the audio corresponding to each section of video into a plurality of audio clips according to audio pause points, and identifying a third set of knowledge points contained in each audio clip based on the knowledge point domain; and calculating a set of the knowledge points contained in the classroom video according to the first set, the second set and the third set.

The invention relates to a character information identification method for a medical image and a device thereof. The method comprises the steps that input of selecting a character area of an operator on the medical image is received, and the character area under character information identification is confirmed according to the input; image preprocessing operation is performed on the character area so that non-character information is removed and a clean character image is obtained; the character image is cut so that characters are split out and single character images are obtained, and the obtained single character images are identified so that preliminarily identified characters are obtained; the preliminarily identified characters are corrected; and the corrected identified characters are verified according to the preset verification conditions so that unreasonable and illegal identified characters are removed, a verification result is outputted, and the verification result includes the verified identified characters. The identification range is reduced according to the received input of the operator, identification efficiency is enhanced, and the preprocessed and preliminarily identified characters are corrected and verified so that identification accuracy is further enhanced.

The invention discloses a peripheral visual field danger identification method based on human vision and machine vision coupling. The method is characterized by firstly, according to an eye tracker experiment, acquiring driver eye moving data, and extracting a coordinate point area reflecting that a driver visual line type is stare; then, carrying out three standard area division based on human eye imaging principle on a camera image, comparing a driver stare point area with standard division areas, and removing the standard division area where the stare point area belongs; and finally, processing the two remaining standard division areas according to a priority, and identifying whether an object is moving and determining whether the object is dangerous to a vehicle. A multi-target dangeridentification problem in the front visual field of a driver is solved, an image identification range is reduced; and priority processing is performed on the peripheral visual field area, the monitoring of a key area is reinforced, and real time image processing efficiency is increased.

The invention relates to the field of encrypted flow identification, in particular to a VPN flow identification method based on SSL encryption. A Bit-level DPI fingerprint generation technology is adopted to identify SSL flow aiming at the problem of higher missing identification rate of the traditional SSL encrypted flow fingerprint identification method, and a bit signature is generated througha specific rule after preprocessing data, the running length of the bit signature is coded, an SSL state conversion machine is generated by using the coded bit signature to identify SSL flow, SSL VPNflow is identified by using a bidirectional GRU network flow identification model based on an attention mechanism on the basis of the identified SSL flow, and the SSL VPN flow is sent to the SSL stateconversion machine. The method at least comprises the steps of data set acquisition, data preprocessing, Bit-level DPI fingerprint generation, SSL flow identification, SSL flow set data preprocessing, SSL VPN encrypted flow identification and index result analysis.

The invention discloses an information processing method and electronic equipment. Firstly, M images are determined on the basis of parameter information, the M images at least comprise one first image, then non-facial feature information in images to be recognized is extracted, the non-facial feature information and the feature information of a first object are compared, and when comparison results show that the non-facial feature information and the feature information of the first object are matched, first identification is used for calibrating areas to be recognized. According to the technical scheme, the problem that multiple factors influence recognition accuracy is solved, the characteristics that the non-facial features of an object to be recognized are fewer in change and high in consistency under the conditions of being in similar time range or at an identical shooting spot are utilized, the purpose of recognizing the object is achieved through comparison of the non-facial feature information and the feature information of the first object, the technical problem that the multiple factors influence the recognition accuracy in the prior art is solved, and the information processing method and the electronic equipment have the technical advantages of being high in recognition rate, stable in recognition accuracy and high in recognition accuracy.

The invention discloses a method and system for recognizing texts in an image. The method comprises the steps: firstly extracting N text positioning regions in a to-be-recognized image through a textdetection model, and correspondingly obtaining N to-be-merged text positioning boxes; wherein each to-be-merged text positioning box is a region contour of a corresponding text positioning region in the to-be-identified image; according to the vertex coordinates of each to-be-merged text positioning box, merging all the to-be-merged text positioning boxes to obtain K to-be-detected text positioning boxes; wherein N > = K > = 1; based on the text recognition model, recognizing texts in all to-be-detected text positioning boxes to obtain to-be-detected texts; comparing the to-be-detected text with keywords in a preset violation keyword dictionary to obtain an recognition result corresponding to the to-be-identified image; according to the technical scheme, the accuracy and the recognition efficiency of text recognition in a complex image can be improved.

The embodiment of the invention discloses a human body recognition method, device and equipment and a computer readable storage medium. The method comprises the following steps: acquiring a temperature image of a target human body and a left-eye view image and a right-eye view image corresponding to the target human body; obtaining a depth image of the target human body according to the left-eye view image and the right-eye view image corresponding to the target human body; processing the temperature image of the target human body and the depth image of the target human body to obtain a targetimage; and recognizing the target image according to a pre-trained human body recognition model. According to the invention, the non-limb area is directly excluded through the temperature image of the target human body; the method is advantaged in that non-human body identification is avoided, the range beyond the preset distance is eliminated through the depth image, the recognition range is reduced, problems of low efficiency and low error-tolerant rate of color-based recognition can be effectively solved, recognition accuracy is improved, extra equipment support is not needed, and cost isreduced.

The embodiment of the invention provides a bus lane occupation vehicle-mounted snapshot processing system based on 5G communication, and the system is characterized in that an image collection unit collects an original image; an image analysis unit analyzes a first target image from the original image; a license plate recognition unit recognizes complete license plate information from the first target image; the image analysis unit analyzes a second target image from the original image; and the license plate recognition unit recognizes complete license plate information from the second target image. According to the embodiment of the invention, the license plate information of the social vehicles occupying the bus lane is identified as little as possible, the data volume of the snapshot processing of the social vehicles is reduced, and the consumption of system resources is reduced. In addition, under the condition that license plate information of social vehicles occupying bus lanes is recognized as little as possible, the license plate recognition range can be expanded to non-bus lanes, and therefore the integrity of license plate information recognition is improved.

The invention discloses a deep learning remote sensing image vessel target identification method based on threshold constraint, and the method comprises the following steps: (1) carrying out the threshold segmentation of a remote sensing image through an OTSU threshold segmentation method, and achieving the sea-land separation; (2) extracting shape features of the remote sensing image; (3) on the basis of a deep learning YOLOv5 algorithm, carrying out multi-scale connection fusion on the pyramid network structure at the bottom layer; (4) aiming at the characteristics of the remote sensing image ship target, designing an anchor frame according to the shape characteristics of the remote sensing image ship target; (5) introducing focusing classification loss as a loss function of a YOLOv5 algorithm to carry out regression convergence; (6) on the basis of an improved YOLOv5 algorithm, inputting a vessel sample for training to obtain a model; and (7) carrying out vessel target identification on the remote sensing image according to the trained model. The method optimizes the anchor frame and the loss function according to the shape features of the vessel, improves the generalization performance of the model, and improves the recognition precision of the vessel.

The invention discloses a gesture recognition method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining a to-be-recognized human body image, and determining a face region in the to-be-recognized human body image; and based on the face area, determining a gesture position prediction area in the to-be-recognized human body image so as to perform gesture recognition in the gesture position prediction area. The face image area of the current to-be-recognized human body image is recognized instead of human body detection, so that the situation that the detection range is unstable due to the fact that the human body is in a non-standing state at the moment or is shielded by other objects is avoided; the possible gesture position prediction area ofthe gesture is speculated according to the face image area, so that the recognition range during single detection and recognition of the image is reduced, the calculation amount of gesture recognitionoperation is greatly reduced, system jamming easily caused when multiple recognition objects are recognized at the same time is avoided, and meanwhile, the recognition accuracy is also improved.

The invention discloses a high-permeability strip judgement method based on HALL curves. The method comprises the following steps that S1, a double logarithmic curve of a water-oil ratio WOR, a water-oil ratio differential coefficient WOR' and production time of an oil well is drawn; S2, according to the double logarithmic curve, a water channeling type is identified, and a well group area where ahigh-permeability strip possibly exists is found out; S3, the HALL curves are drawn for all water injection wells in the well group area which is found out; S4, a typical HALL curve is determined, and the HALL curves of the actual water injection wells are compared with the typical HALL curve; when the HALL curves of the water injection wells deviate from the typical HALL curve and are lower thanaverage HALL curves of other wells, the water outlet type of the oil well controlled by the water injection well is judged as the high-permeability strip water outlet type. The invention aims at providing the high-permeability strip judgement method based on the HALL curves. The problem that whether or not abnormal water outlet is caused by the high-permeability strip is difficultly determined inthe prior art is solved. The aim of improving the identification accuracy of the high-permeability strip is achieved.

The invention discloses a target hand analysis method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring a to-be-recognized image in a to-be-detected area; determining a first suspected target according to the to-be-recognized image and a preset hand contour; if the first suspected target belongs to the human body area, determining the first suspected target as a second suspected target; and if the gesture of the second suspected target accords with the preset gesture, determining that the second suspected target is a target hand. According to the embodiment of the invention, the target hand is determined by judging whether the target conforms to the preset gesture or not, and misrecognition caused by the fact that some objects are similar to the hand contour is avoided.

The invention discloses a method for recognizing animal categories through voice, which comprises a voice input module, a voice analysis module, a microcontroller module and a voice database module, and is characterized in that the voice analysis module is respectively connected with the voice input module and the microcontroller module, and the microcontroller module is connected with the voice database module. According to the animal sound recognition method provided by the invention, after most sound of the animal boundary is stored, the content of the sound can basically cover all animal types, and according to the classification mode of the border line, all types to which the animals belong can be directly output and recognized after the animals are accurately recognized, so that the recognition accuracy is improved.

The embodiment of the invention discloses a content data processing method, a report data processing method, computer equipment and a storage medium. A picture is extracted from the video, the target area of the content data in the picture is firstly positioned, and the content data is further recognized in the target area, so that on one hand, the recognition range can be narrowed, the recognition speed can be improved, on the other hand, other redundant information does not need to be recognized, and recognition accuracy can be improved. The content data and the audio data are further associated, and the associated content data and audio data are provided, so that the content data and the audio data can be conveniently combined for common use, part of the content data and the corresponding audio data can be selected, and rapid use of the key content data is facilitated.

The invention provides an automatic detection system and method for an engineering machinery fault display instrument. The automatic detection system comprises a fault simulation sending module, a camera module, a computer software module and a communication module, the computer software module is in data communication with the fault simulation sending module and the camera module through the communication module, and the fault simulation sending module is connected with a detected instrument through the communication module. The fault display instrument can be automatically detected, and a detection report can be generated; the camera module and an improved character recognition algorithm improve the recognition rate and recognition speed of characters, recognition data and detection results are stored locally, and a detection result document is automatically generated, so that the data are traceable, and the results are easy to export.

The invention discloses an image processing method of ultrasound images, ultrasound equipment and a storage medium. The image processing method includes: obtaining a target ultrasound image of a to-be-detected target, wherein the target ultrasound image at least includes a region-of-interest; performing image recognition on the region-of-interest in the target ultrasound image to determine a healthy area and a lesion area; and performing image mark processing on the lesion area, so as to track the lesion area. Through the method, the recognition efficiency of the lesion area can be enhanced, and the positioning accuracy of the to-be-detected target can be increased as well.