31results about How to "Reduce the impact of recognition accuracy" patented technology

The invention belongs to the field of close-range photogrammetry and relates to a detecting and identifying method for an annular coding mark point. The method includes that first, canny edge detection is conducted on a collected image, an outline centroid is closed through a series of limiting conditions and calculation, and noise and non-coding mark points are filtered; then least square ellipse fitting is adopted, coding mark point location is conducted, an ellipse fitting error is combined to judge a partition coding mark point outline, and the outline is filled; finally, ALPC transformation for transforming a local concentric ellipse into parallel straight lines is provided, the ALPC transformation is conducted on the partitioned coding mark point, and transformed image characteristics are used for decoding. By means of the detecting and identifying method, location of the coding mark point can reach a sub pixel level, local shape characteristics of a concentric ellipse are transformed into shape characteristics of parallel straight lines which are easy to detect and calculate, coding mark point identifying speed is improved, and effects of an included angle of a camera optical axis and a coding mark point normal on the coding mark point identifying accuracy can be reduced.

The invention is suitable for the medical field, and provides a plasmodium infected erythrocyte identification method and a device thereof. The method comprises the following steps: acquiring the forward diffusion light signal and the lateral diffusion light signal of cells in a sample, and an optional fluorescence signal; obtaining a first two-dimensional point diagram according to the forward diffusion light signal and the lateral diffusion light signal, or obtaining a three-dimensional point diagram based on the forward diffusion light signal, the lateral diffusion light signal and the fluorescence signal; and identifying cells appearing in the preset area of the first two-dimensional point diagram or the three-dimensional point diagram as plasmodium infected erythrocytes. The method has a high identification precision.

The invention relates to an OCR recognition method based on a deep learning model and a terminal, and belongs to the field of data processing. The method comprises the steps of segmenting a preset character segment image into a plurality of single-character images to obtain a single-character image set; sequentially identifying elements in the single-character image set by a preset first OCR deeplearning model to obtain a first feature vector set; wherein one single-character image corresponds to a first feature vector; according to a preset feature database, each first feature vector in thefirst feature vector set is converted into a corresponding single character, and a single character set is obtained; wherein a record in the feature database stores a single character and a feature vector corresponding to the single character; and arranging elements in the single character set to obtain a character string corresponding to the character segment image. The anti-interference capability of OCR character recognition is improved.

The present invention discloses an action sequence detection-based user identity identification method applied to an indoor WiFi environment. The invention aims to solve the technical problem of low accuracy of an existing user identity identification method based on WiFi signals. According to the technical schemes of the method, sensing is realize through using devices such as commercial WiFi and notebooks; sensed data are preprocessed, so that the quality of the data can be improved; features are extracted from the data so as to depict user identities; a classification model is constructed; and the classification model is invoked to calculate the probability distribution of user identities of single actions; the probability distribution results of all identifiable actions in action sequences are put into statistics, so that identity identification is realized; and in the process of the identity identification, the data are used to accurately depict waveforms, and the user identities are judged many times through the action sequences, and therefore, the influence of complex environments on identification accuracy can be decreased, and high-accuracy identity identification is realized based on the results of a plurality of times of judgment.

The invention relates to an audio identification method and device, a storage medium and electronic equipment. The method comprises the following steps: performing human voice extraction on an audio to be identified to obtain a target audio only containing a human voice, wherein the audio to be identified corresponds to a target song; performing voiceprint feature extraction on the target audio toobtain target voiceprint feature information; and according to the target voiceprint feature information, determining whether the audio to be identified belongs to an original singer version of the target song or not, wherein the original singer version of the target song is from an original singer of the target song. In this way, whether the audio to be identified belongs to the original singerversion or a cover version of the song can be identified more accurately, so that the accuracy of audio identification is improved, the stability is high, and the influence of the environment is relatively small.

The embodiment of the invention provides a three-dimensional face recognition method and device, terminal equipment and a computer readable medium. The method comprises the steps of acquiring a near-infrared image and a depth image of an actual scene; normalizing the near-infrared image and the depth image, and extracting a first multi-dimensional feature vector of a human face in an actual scene;calculating a deflection angle of the face according to the key points of the face in the near-infrared image, rotating a pre-input face three-dimensional model according to the deflection angle, andperforming planar projection on the rotated face three-dimensional model to obtain an input near-infrared image and a depth image of the face; carrying out normalization processing, and extracting asecond multi-dimensional feature vector; and calculating the similarity between the first multi-dimensional feature vector and the second multi-dimensional feature vector, and performing face recognition according to the similarity, thereby effectively recognizing faces with different deflection angles, improving the three-dimensional recognition efficiency, and reducing the influence of the background on the face recognition precision.

The invention relates to a Raman spectrum data analysis method and device based on deep learning. The method comprises the following steps: analyzing Raman spectrum data characteristics of a substance; manually labeling substance Raman spectrum data categories, and establishing a Raman spectrum training set, a verification set and a test set; in order to solve the problems that Raman spectrum data is preprocessed, spectrums are easily interfered by ambient light and Raman spectrum data of the same type of tissues are different, constructing a deep residual neural network model based on multi-scale feature fusion, employing ResNet50 as a model backbone network, fusing Raman spectrum feature information, and improving the spectrum space semantic information characterization capability; training the Raman spectrum analysis model by using the training set, and evaluating the performance of the model on the verification set and the test set; and finally, deploying the trained model to edge computing equipment, and constructing a Raman spectrum data analysis device. The Raman spectrum data can be efficiently and accurately analyzed and identified, and the method can be applied to different types of Raman spectrum equipment.

The invention provides a vehicle positioning method and device. The method comprises the following steps: acquiring reference models of lanes, wherein the reference models are generated according to vertical accelerations changing along with time in directions vertical to ground of reference vehicles travelling the lanes; when a target vehicle travels, measuring the process that the vertical acceleration of the target vehicle changes along with time to obtain a test sample; and determining the target lane of the target vehicle from the lanes according to level of similarity between the test sample and the reference models of the lanes. The method can reduce the influence of environmental factors on lane recognizing accuracy and improve the lane recognizing and positioning precision and reliability, so that the problem that the recognizing accuracy is easily interfered by environmental change in the prior art is solved.

The invention discloses an English secondary braille alphabet conversion system based on image acquisition and correction. The system comprises an image acquisition and processing device, the image acquisition and processing device comprises a scanning unit and an identification module. The identification module is electrically connected with a CCD/CMOS sensing module and an image conversion module. The CCD/CMOS sensing module is electrically connected with the scanning unit; the image conversion module is electrically connected with an image uploading module; the image uploading module is electrically connected with a cloud database, when the system operates, the scanning unit is required to perform original video acquisition firstly, the scanning unit comprises a monitoring camera, and after the scanning unit acquires video data and image data of a first hand, the scanning unit is used for acquiring the video data and image data of the first hand. The system has the advantages that the recognition unit is electrically connected with the noise reduction module; and the image synthesis filtering unit synthesizes the processed high-frequency sub-band coefficient and low-frequency sub-band coefficient to obtain an image without noisy points, thereby ensuring that subsequent text recognition is accurate and errorless.