111results about How to "The extraction result is accurate" patented technology

The invention discloses a packaging structure and a packaging method for a fingerprint recognition chip. The packaging structure comprises a substrate, a sensing chip, a plurality of wires and a plastic package layer. The substrate is provided with a first surface, and the first surface of the substrate is provided with a first welding pad layer. The sensing chip is located on the first surface of the substrate and provided with a first surface and a second surface, the second surface of the sensing chip is located on the first surface of the substrate, the first surface of the sensing chip is provided with a sensing area and a peripheral area, and the surface of the portion, in the peripheral area, of the sensing chip is provided with a second welding pad layer. The two ends of the wires are electrically connected with the first welding pad layer and the second welding pad layer respectively, the wires have top points with the maximum distance to the surface of the substrate, and a first distance is reserved between the top points and the first surface of the sensing chip. The plastic package layer is located on the surface of the substrate and the surface of the sensing chip and surrounds the wires and the sensing chip, and a second distance is reserved between the surface of the plastic package layer and the first surface of the sensing chip and is larger than the first distance. According to the packaging structure, the requirement for the sensitivity of the sensing chip is lowered, and application is wider.

Disclosed is a wafer-level fingerprint recognition chip packaging structure and method. The method includes: providing a substrate which comprises a plurality of induction chip areas and is provided with a first surface and a second surface opposite to the first surface, wherein a first surface of each induction chip area comprises an induction area; forming a covering layer on the first surface of the substrate; forming a plug structure in each induction chip area of the substrate, wherein one end of each plug structure is electrically connected with the corresponding induction area, and the other end of each plug structure is exposed out of the second surface of the substrate. By the forming method, the packaging process of a fingerprint recognition chip can be simplified, the requirement on sensitivity of the induction chip is lowered, and the packaging method is wider in application.

The invention relates to the field of image processing, and discloses a method of extracting a rectangular building from a remote sensing image. The method comprises the steps of obtaining a plurality of superpixel region blocks by superpixel segmentation of a remote sensing image; determining two seed points on a target building in the remote sensing image; merging the plurality of superpixel region blocks based on the determined seed points; carrying out corner detection for the remote sensing image; calculating to generate a corner distance saliency map related to each pixel point in the remote sensing image based on the corner detection result; carrying out binary segmentation for the corner distance saliency map; determining priori information based on the merging result of the plurality of superpixel region blocks and the distance saliency map after binary segmentation; obtaining a building segmentation result by segmenting the remote sensing image based on the priori information; carrying out morphological image processing for the building segmentation result; and obtaining a rectangular target building by rectangular fitting of the building segmentation result after the morphological image processing. The above the method can accurately extract the rectangular building from the remote sensing image.

Disclosed are a fingerprint identification chip packaging method and a fingerprint identification chip packaging structure. The fingerprint identification chip packaging method includes: providing a substrate; coupling a sensing chip on the surface of the substrate, wherein the sensing chip is provided with a first surface comprising a sensing area and a second surface opposite to the first surface and positioned on the surface of the substrate; forming a plastic packaging layer on the surface of the substrate, wherein the plastic packaging layer encloses the sensing chip, and the surface of the plastic packaging layer is flush to the first surface of the sensing chip; forming a covering layer with the thickness smaller than 100 micrometers on the plastic packaging layer and the first surface of the sensing chip. The fingerprint identification chip packaging method has the advantages that the fingerprint identification chip packaging structure formed by the fingerprint identification chip packaging method is simplified, and the requirement on sensitivity of the sensing chip can be lowered, so that the fingerprint identification chip packaging method and the fingerprint identification chip packaging structure are wider in application range.

The invention discloses a sea surface ship candidate region detection method based on visual saliency. The method comprises the following steps: step 1, acquiring a frequency domain feature map of anoptical remote sensing image containing a sea surface ship target by adopting an FT visual saliency model; step 2, acquiring an edge gradient feature map of the optical remote sensing image by adopting a Scharr edge detection operator; step 3, fusing the frequency domain feature map and the edge gradient feature map by adopting a Gaussian mixture function to obtain a fused feature map; and step 4,based on an otsu adaptive threshold segmentation algorithm, carrying out binarization on the fusion feature map, so that an area containing a ship target can be extracted from the image, and a sea surface ship candidate area can be obtained. The advantages of the FT visual saliency model and the Scharr edge detection operator in sea surface ship detection are fused, the influence of background noise is effectively suppressed, ship targets of all sizes are highlighted, and the candidate region extraction result is more accurate; the detection speed of the fusion model for extracting the sea surface ship candidate area is high.

The invention discloses an image color retrieval method and an image color retrieval system and relates to the field of image processing. The method comprises the steps of performing quantification processing to HSV (Hue, Saturation and Value) color spaces of pixels in an image to obtain quantified color values within a range of 0 to N, wherein non-uniform fuzzy quantification is performed to tone values; statistically collecting a histogram of the quantified color values of the pixel of the image and performing normalization processing by using the sum of the pixels of the image to obtain a color characteristic vector of the image; comparing the color characteristic vector of the image to the color characteristic vectors of other images to determine the similarity of the images. The method and the system disclosed by the invention have the advantages that since images related to the inquired image color are extracted from a database by taking the color characteristic vector after image quantification as a relevant judgment basis, the image retrieval is enabled to be more intelligent and the extraction result of color-related images becomes more accurate; since the images in the database are processed in advance by adopting an offline module, the execution speed is effectively improved.

The invention provides a method and a device for extracting an intracerebral brain tissue. The method comprises the following steps: (1) inputting an image, preprocessing the image for obtaining a smooth image and a skin mask, and performing binarization on the smooth image for obtaining a binary image; (2) according to a current corrosion radius, performing morphological erosion on the binary image for obtaining N current connected domains; (3) determining a human neck position through the skin mask, and determining a current final dividing result according to the N current connected domains and the human neck position; and (4) performing dividing leakage detection on the current final dividing result, if a leakage occurs in the current final dividing result, enlarging the current corrosion radius and returning to the step (2); and if no leakage occurs in the current final dividing result, determining the current final dividing result as the intracerebral brain tissue which requires extraction. The method and the device can be used for automatically and accurately extracting the intracerebral brain tissue and furthermore have advantages of low realizing complexity and high computing speed.

The invention discloses an epileptic electroencephalogram characteristic extraction method based on a wearable device. The extraction method includes the steps: outputting a classifier with characteristic migration through operational treatment of kernel principle component analysis, a multi-kernel support vector machine and migration characteristic calculation according to marked epileptic electroencephalogram data acquired by a hospital and unmarked epileptic electroencephalogram data acquired by the wearable device; performing characteristic extraction on the unmarked real-time epileptic electroencephalogram data acquired by the wearable device by the aid of kernel principle component analysis, and placing the unmarked real-time epileptic electroencephalogram data after characteristic extraction into the classifier to output predicting results. According to the method, image resources of image departments of various hospitals and image centers are integrated to form an epileptic electroencephalogram characteristic extraction method served for clinic and based on the wearable device, and clinical working efficiency is improved. The epileptic electroencephalogram characteristic extraction method can be rapidly and accurately used for characteristic extraction and result prediction of epileptic electroencephalogram of the wearable device.

The invention discloses a parameter extraction method for InP HBT small-signal models, which mainly solves the problems of the prior art, i.e. complex extraction process and inaccurate extraction results. The technical scheme of the invention is as follows: an open circuit soldering point structure and a short circuit soldering point structure are adopted to analyze an equivalent circuit to extract parasitic parameters; a collector open circuit state is adopted to analyze the equivalent circuit to extract external resistance parameters; a circuit network theory is utilized to analyze the scattering parameter S, impedance parameter Z and admittance parameter Y of the intrinsic part of an InP HBT device; and a layer-by-layer stripping method is adopted to determine a fixed expression for each intrinsic model parameter to extract intrinsic parameters. The parameter extraction method has the advantage of accurate, rapid and visual parameter extraction. A simulation result shows that a small-signal model result extracted by the method can perfectly fit the scattering parameter S of an actual device-testing result and can be used for directing circuit design and the determination of a large-signal model peripheral circuit.

The invention discloses a nucleic acid extraction device capable of preventing cross-contamination. The device comprises a motor, a motor control circuit, a stirring sleeve frame which is driven by the motor, stirring sleeves which are arranged on the stirring sleeve frame, and sampling boards, wherein guide grooves are formed on the stirring sleeve frame; each stirring sleeve is provided with a connecting part, and stirring sleeve feet which are formed by extending from the connecting part; the bottom of each stirring sleeve foot has a rotary double-leaf double-curved-surface structure; each stirring sleeve is in clamping fit with the guide grooves of the stirring sleeve frame through curve-shaped guide rails which are arranged on two sides of the connecting part; and the motor control circuit controls the motion of the motor according to input parameters of sample information and further controls the motion track of the stirring sleeves in deep holes of the sampling boards. The nucleic acid extraction device effectively prevents the cross-contamination caused by the mutual influence of aerosol and liquid droplets among samples in the automatic nucleic acid extraction process, and ensures that nucleic acid extraction results are accurate and reliable.

The invention discloses an SAR image change detection method based on an oriented difference chart. The problems that in the prior art, change information of SAR images of one area at different moments can not be accurately described, and the interference resisting capacity is weak are mainly solved. The SAR image change detection method includes the steps of 1, preprocessing two SAR images of one area at different moments; 2, respectively building a mean-value ratio difference chart and a logarithm ratio difference chart for the two preprocessed SAR images; 3, according to the thought of an oriented filter, enabling the mean-value ratio difference chart to serve as an oriented chart, enabling the logarithm ratio difference chart to serve as an input chart, and carrying out fusing to generate the oriented difference chart; 4, dividing the oriented difference chart through a mean-value cluster level set model based on local joint entropy, and obtaining the result of change direction. By means of the SAR image change detection method, interference of background area information can be effectively restrained, the change direction accuracy is improved, and the SAR image change detection method can be used for natural disaster evaluation, environment resource monitoring and urban construction planning.

The invention discloses a news key information extraction method and system. According to the method, a webpage is converted into a set of line numbers and a text through preprocessing; according to the feature that the probability of a sentence with the maximum number of words occurring in the news main body is very high, the start point and end point of the main body are searched starting from the middle part of the main body towards two ends, thereby extracting the news main body; a title is extracted according to a longest public substring algorithm; a regular expression is established; time is extracted by taking the line numbers as assisted judgment; a source is extracted by taking the line numbers as the assistance according to format features of the source; and the obtained contents are written into a local file by taking line breaks as separators in sequence according to the obtained news title, time, source and main body. The system has relatively high accuracy, is independent of a special webpage template and has relatively high practicability and robustness. Moreover, according to the method, the complexity is low; the extracted results are accurate; and the consumed resources are few.

The present invention discloses an SAR image change detection method based on significance and a deep convolutional network. The problem is mainly solved that the detection precision is low and the capacity of resisting disturbance is poor in the prior art. The implementation scheme of the method comprises the steps of: performing preprocessing of two SAR images of the same area at different moment to perform logarithm ratio operation and normalization to obtain a normalized logarithm ratio difference graph; performing significance feature extraction of the difference graph to obtain a saliency map; performing thresholding of the saliency map, multiplying the saliency map by a thresholding result to obtain a saliency feature map; performing clustering of the difference graph to obtain pre-classification results; selecting training samples from the pre-classification results; constructing a deep convolutional network, and performing training of the network by employing the training samples; and performing change detection of the SAR image by employing the trained network. The SAR image change detection method can effectively inhibit the disturbance of background area information, can improve the change detection accuracy, and can be used for natural disaster assessment, environmental resource detection and urban modeling planning.

The invention discloses a cluster-based automatic WMH extraction system. The system comprises a tissue type segmentation module, a DARTEL standard segmentation module, a non-brain tissue removal module, a WMH segmentation module and a WMH refined division module, wherein the tissue type segmentation module is used for preprocessing a plurality of RLAIR images and a plurality of T1 images of a plurality of participants; the DARTEL standard segmentation module is sued for mapping the images obtained by the tissue type segmentation module into a DARTEL space; the non-brain tissue removal module is used for removing non-brain tissues of the FLARI images and the T1 images in the DARTEL space, and respectively marking the obtained images as D-FLAIR images and D-T1 images; the WMH segmentation module is sued for segmenting the D-FLAIR images and the D-T1 images on the basis of the DARTEL space so as to obtain a WMH map; and the WMH refined division module is used for carrying out refined division on the WMH map so as to obtain an intracerebroventricular white matter hyperintensities area PVWMH and a deep white matter hyperintensities area DWMH. The system is an automatic WMH extraction system which is capable of segmenting white matter focuses on the basis of longitudinal data sets and carrying out refined division on the WMH and has favorable generalization performance.

The invention discloses a method for extracting the lowest point of the spatial wheel rim of a wheel track wheel. The method comprises the following steps of S1, laser displacement sensor arrangement; S2, laser displacement sensor data acquisition and three-dimensional coordinate transformation; S3, determination of a laser displacement sensor emitted laser plane; S4, spatial wheel rim data point projection and end surface projection line determination; S5, the fitting of the spatial wheel rim projection curve and the initial extraction of the lowest point of the spatial wheel rim; and S6, extraction of the final spatial wheel rim. The method for extracting the lowest point of the spatial wheel rim of the wheel track wheel of the invention has the advantages of being more accurate and practical, so as to provide a reliable guarantee for the non-contact online accurate measurement of the wheel set profile dimension parameter. The method is very suitable for rail transit monitoring engineering technology.

The invention belongs to the technical field of geoscience, and particularly relates to a deep alteration mineral three-dimensional modeling method. The method comprises the following steps: step 1, carrying out hyperspectral scanning on a drilled core image; 2, preprocessing the borehole core imaging hyperspectral data; Step 3, performing rock core alteration mineral mapping; 4, calculating the relative content of the rock core alteration minerals of the unit section; 5, displaying the relative content of the borehole alteration minerals in a three-dimensional mode; 6, carrying out parallel section projection; Step 7, carrying out three-dimensional space vectorization; And step 8, carrying out three-dimensional space interpolation on the alteration minerals. According to the method, the deep alteration mineral three-dimensional solid model is established, the deep drilling rock core alteration mineral extraction result is fine, objective and accurate, the modeling precision is high, and the method has important practical application value for deep geological research and prospecting exploration.

The invention provides a span-based fine-grained sentiment analysis method, and relates to the technical field of sentiment analysis. According to the invention, a sentiment analysis model based on aspects is established by selecting a sentiment analysis data set, the sentiment analysis model based on aspects is trained through a loss function and a training data set, and sentiment analysis of thetext is achieved by making the test data set to be subjected to sentiment analysis pass through the trained sentiment analysis model based on aspects. The network model provided by the invention is used for extracting aspects and corresponding sentiment polarities, and the model is a simple and effective joint model for extracting sentence aspects and corresponding sentiment polarities for sentiment analysis tasks. The model uses BERT as word embedding, then uses a loop control unit to extract the representation of each subtask, uses an interaction layer to consider the relationship between them, and finally performs aspect item extraction and sentiment analysis.

The embodiment of the invention discloses a method for quickly extracting equivalent three-dimensional pore parameters of shale. The method comprises the following steps: acquiring a scanning electronmicroscope imaging image of a shale sample; selecting a segmentation threshold, and dividing the electron microscope image into four components; dividing the pores according to the size of a connected region; splicing the organic holes and the inorganic holes extracted from the large area and the small area; extracting parameters according to an image recognition result; converting the acquired two-dimensional plane parameters into equivalent three-dimensional parameters according to dimension raising transformation; and obtaining the sample area of the RES scale. According to the method, theprocessing speed is higher, the extraction result is more accurate, and more effective information data is obtained; the method is in favor of accurately evaluating the reservoir gas storage capacityand the transport capacity of shale, and can rapidly distinguish organic holes and inorganic holes and extractequivalent three-dimensional parameters. In addition, the method has the advantages of simple operation, automatic completion of calculation, high calculation efficiency, and strong applicability, and can analyze and research various shale samples.

The invention relates to a pathology information extraction method based on dynamic pulse wave feature parameters and belongs to the technical field of fusion of information science and medical science. The pathology information extraction method comprises the following steps: first searching for starting points and finishing points of pulse waves; solving confidence intervals of the starting points and the finishing points of the pulse waves; judging whether the pulse waves are normal according to the confidence intervals of the starting points and the finishing points so as to select a normal pulse wave; working out the number of gaussian functions for fitting the pulse waves in the I normal period section; solving parameters ak and ck of gaussian functions for fitting the pulse waves in the i normal period section; then solving a parameter bk and a gaussian function expression of the pulse waves in the i normal period section; solving six feature points of the pulse waves in the i normal period section; sequentially solving six feature points of the pulse waves in each normal period section, and solving an average value of the feature points of the pulse waves in all the normal period sections, wherein the average value serves as case information to be analyzed. By means of the pathology information extraction method based on dynamic pulse wave feature parameters, fitting precision is higher, and the feature points of the pulse waves are extracted more accurately.

The invention relates to a task type multi-round dialogue method and system for an office scene. The method comprises the steps of performing scene definition on different office application tasks toobtain intentions and imperfection values of the different office application tasks; performing data acquisition on the intentions and the imperfection values of the different office application tasksto obtain acquired data; determining an intention recognition model and a imperfection extraction model by using a deep learning model training method and a customized rule processing method according to the acquired data; performing dialogue management until necessary information is collected; determining a current user intention and a current imperfection value according to the intention recognition model, the imperfection extraction model and the necessary information; and executing the service according to the current user intention and the current imperfection value. According to the invention, different office task scenes can be covered, the requirements of users are accurately met, and the office efficiency is improved.

The invention provides a method of extracting permanent scatterers by using a small data set SAR (Synthetic Aperture Radar) image sequence. According to the technical scheme, 8-15 SAR image sequences are utilized, the SAR image sequences are preprocessed, airspace signal to clutter ratio estimation and time domain signal to clutter ratio estimation are performed, and PS judgment is performed to obtain a PS extract result. The method has the beneficial effects that the implementation is simple, the accuracy is high, and the applicable range is wide.