12results about How to "Prevent misidentification" patented technology

The invention relates to a method and device for controlling a vehicle through voice outside the vehicle and a medium. The method comprises the steps of determining that pedestrians appear near a vehicle; intercepting a face image; judging whether the face image is in an input information base or not; if yes, detecting whether the vehicle is in a dormant state currently; if in the dormant state, waking up the intelligent voice module, the microphone and the loudspeaker to enter a working state; obtaining human voice; judging whether the voice is in an input information base or not; if yes, acquiring a voice instruction; converting the voice instruction into a text through a voice recognition technology; an intelligent voice module is used for understanding the text and generating a control instruction; determining an out-of-vehicle sound area where the sound source position of the voice instruction is located by using a microphone; judging whether the external sound area supports the control instruction or not; the invention provides a method for controlling a vehicle through a voice instruction outside the vehicle. When the vehicle is dormant, the specific part can be opened through the voice instruction, and the vehicle does not need to be started.

The invention relates to the technical field of data processing, and discloses a radio frequency fingerprint identification method based on a deep optimization adaptive attention mechanism, and the method comprises the steps: receiving a target equipment radio frequency signal, and carrying out the down-conversion, filtering and digital sampling of the target equipment radio frequency signal, and obtaining an original IQ data sample; preprocessing the original IQ sample to obtain a training data set; inputting the training data set into a convolutional neural network for pre-training to obtain a pre-training model; performing fine tuning on the pre-training model to obtain a downstream task model; and using the downstream task model to classify and identify unknown equipment data. Through time window segmentation and feature enhancement, the definition and consistency of radio frequency signals are improved, and meanwhile, the influence of noise and interference is reduced; and the effect of improving the capturing capability of the model on multi-scale and key features in a complex environment is achieved through the differentiated attention mechanism module.

The invention discloses reference management and garbage collection of application data. The application data reference management method comprises the steps that when it is detected that target application data is referenced by a target variable, and the target variable belongs to a target variable type, a logic address of the target application data is written into a physical address of the target variable; and taking the logic address corresponding to the physical address of the target variable as a reference address, and writing the reference address into a reference address table. According to the technical scheme, the reference address table is used for managing application data reference which is not recorded in the reference array, the application data are prevented from being mistakenly regarded as junk data and deleted, null exception caused by reference of the inexistent application data is avoided, the reference address table records the reference address instead of the logic address of the application data, and the application data can be stored in the application data. Content in an application data reference counter or a reference address table is prevented from being synchronously modified when the reference relation is cleared, and application execution efficiency is guaranteed.

The invention discloses a method for identifying a launching station number of a down-cast sonde, belongs to the technical field of high-altitude meteorological data detection, and aims to solve the technical problem that in the prior art, after the down-cast sonde is loaded into a launching device, the corresponding launching device cannot acquire the ID number of the loaded down-cast sonde through a conventional communication interface. The method comprises the following steps: S1, after airborne control equipment is powered on, performing in-situ detection on each channel of a radio frequency switch one by one; s2, the airborne control equipment records the in-place condition of the down-cast sonde in each channel of the radio frequency switch and the ID number of the corresponding down-cast sonde; s3, connecting the plurality of launchers with the corresponding down-cast sondes; s4, enabling the station number of the transmitting device to correspond to the channel number of the radio frequency switch; s5, enabling the ID number of the down-cast sonde to correspond to the station number of the transmitting device, and completing the identification of the station number. According to the method for identifying the put station number of the downward-put sonde, the identification efficiency of the station number is improved, and identification errors possibly caused by manual identification of the station number are avoided.

The invention provides an endoscope system capable of preventing misrecognition of a result of recognition processing and a result of non-recognition processing. An endoscope system (10) is provided with: a discrimination unit (73) that discriminates the type of an image obtained by capturing an image of a subject; a recognition unit (72) that performs recognition processing for recognizing the subject using the image; and a notification unit (65) that notifies whether or not the identification processing is functioning with respect to the specific type of image identified by the identification unit (73). The discrimination unit (73) discriminates the type of image by discriminating, for example, the operation mode, the type of endoscope used for imaging the subject, or the presence or absence of administration of a drug to the subject.

The utility model discloses a water quality treater which comprises a shell, an electrolytic bath, a control mainboard, a power supply module and a pipeline system, the electrolytic bath, the control mainboard, the power supply module and the pipeline system are arranged in the shell, the electrolytic bath is provided with a cathode chamber and an anode chamber which are provided with electrode plates, and the power supply module provides electric energy. The control mainboard controls the electrode plate to electrolyze oxidized water in the cathode chamber and electrolyze reduced water in the anode chamber respectively; the pipeline system is provided with a water inlet pipe for supplying water to the cathode chamber and the anode chamber, an oxidation water outlet pipe for discharging oxidation water to the cathode chamber and a reduction water outlet pipe for discharging reduction water to the anode chamber; an oxidation water outlet valve is arranged on the oxidation water outlet pipe, a reduction water outlet valve is arranged on the reduction water outlet pipe, and under the control of the control main board, the oxidation water outlet valve controls the oxidation water outlet pipe to discharge oxidation water, and the reduction water outlet valve controls the reduction water outlet pipe to discharge reduction water. Therefore, the purpose of separately discharging the oxidized water and the reduced water is achieved, and the service life of the electrolytic bath and the water outlet pipeline is prolonged.

This invention proposes a touch display device for use with an active pen. During touch sensing, the touch position is detected. During active pen sensing, an uplink signal is transmitted and a downlink signal is detected. When a downlink signal is detected, it is determined whether the distance between the active pen position and the touch position is less than or equal to a first preset distance; if so, the device does not enter active pen mode.

The present disclosure relates to a wireless communication system using a quadrature amplitude modulation scheme, and an object thereof is to effectively prevent erroneous recognition of data while utilizing a non-linear region of an amplifier. A transmission device includes: a transmission signal amplifier enabled to vary transmission power; a transmission power control unit that controls the transmission power; and an approximate pattern information notification unit that generates pattern information on an approximate constellation. In a case where 64 QAM is used as the modulation scheme, a 64 QAM distortion constellation is generated in the non-linear region of the transmission signal amplifier. In this case, a 64 APSK approximate constellation is generated, and pattern information on the 64 APSK approximate constellation is transmitted to a reception device. The reception device performs likelihood calculation for a reception point indicated by a data signal by using the approximate constellation corresponding to the pattern information.

This invention relates to the field of object recognition technology for nerves, and more particularly to a deep learning-based real-time intraoperative localization and monitoring system for recurrent laryngeal nerve surgery. The system includes a preoperative analysis module to acquire the expected nerve path sequence to determine a first variation risk region; an intraoperative acquisition module to acquire intraoperative video, electromyographic signals, and sequences corresponding to instrument positions; a priori constraint recognition module to determine the intraoperative nerve recognition path sequence and visual recognition confidence level; a path deviation correction module to correct the intraoperative nerve recognition path sequence and visual recognition confidence level; a distance state analysis module to determine whether the real-time distance between the instrument and the nerve is within a distance boundary state; and a warning monitoring module to determine the existence of hidden warning regions based on spontaneous electromyographic activity indicators combined with visual recognition correction confidence level, and to monitor the rate of change of spontaneous electromyographic activity indicators to indicate the nerve stress response state. This invention improves the accuracy of real-time intraoperative localization and monitoring of recurrent laryngeal nerve surgery.

The invention discloses a multi-source remote sensing image enhancement method, device and equipment and a storage medium, and the method comprises the steps: collecting a multi-source remote sensing image, carrying out the preprocessing of the multi-source remote sensing image, and constructing an initial image sample library; performing distortion elimination processing on the first-resolution image in the initial image sample library, and performing correction processing on the second-resolution image to obtain a paired image sample library; constructing a deep learning resolution enhancement network combining a stacked architecture and an attention mechanism; a progressive training strategy is adopted, and the target image sample library is utilized to train the deep learning resolution enhancement network; and inputting a to-be-enhanced remote sensing image into the well-trained deep learning resolution enhancement network to generate an enhanced multi-source remote sensing image. According to the method, through sample library construction, distortion elimination, network design and a training strategy, the generated image gives consideration to high spatial resolution and high temporal resolution at the same time, and the generalization ability and the enhancement effect of the model are enhanced.

The present invention relates to the technical field of commercial image management, and in particular to a periodically updated image management system and management method. The management system comprises an image library, an acquisition module, a feature extraction module, a feature matching module, a determination module and an update module. The feature extraction module is configured to obtain image features of display images of similar commodities; the feature matching module is configured to determine a candidate image set of a current period from the image library according to the number of clicks and image features of display images of similar commodities in the previous period; the determination module is configured to determine display images of commodities in the current period from the candidate image set based on the number of collections of commodities in the previous period; by accurately matching multiple features in the image library, the candidate image that best meets the current browsing trend can be selected from the image library, and the similarity is further selected for the misrecognition of the image, thereby effectively improving the image display effect under the premise of ensuring the discernibility of the displayed image.