1470results about How to "Improve the extraction effect" patented technology

The invention provides a method for detecting an X-ray mammary gland lesion image based on feature pyramid network under transfer learning. The method comprises the steps: 1, establishing a source domain data set and a target domain data set; 2, establishing a deformable convolution residual network layer by a deformable convolution and extended residual network module; 3, establishing a multi-scale feature extraction sub-network based on a feature pyramid structure through a feature map up-sampling and feature fusion method in combination with a deformable convolution residual network layer;4, establishing a deformable pooling sub-network sensitive to the focus position; 5, establishing a post-processing network layer to optimize a prediction result and a loss function; and 6, migratingthe training model to a small sample molybdenum target X-ray mammary gland focus detection task so as to improve the detection precision of the network model on the focus in the small sample image. According to the method, a transfer learning strategy is combined to realize focus image processing in a small sample medical image.

The invention relates to a method for integrated comprehensive management of gas in a low-permeability single coal bed, which can effectively solve the problem of comprehensive management and utilization of the gas. A technical scheme of the method comprises the following steps of: firstly, drilling holes, wherein a high-torque drilling machine and a rifle or a trihedral drill stem are used for solving the phenomenon of clamping the drill stem caused by collapsing and jetting of the holes in the hole drilling process; and then, performing hydraulic fracturing permeability improvement on the coal body, drilling holes for pre-split blasting, performing gas drainage, and recycling the drained gas, thereby implementing the integrated management of the gas. With the method provided by the invention, the permeability of the coal body is improved, a pre-draining rate of the gas is improved, and the drained gas is recycled, so that the drilled holes are radically prevented from being blocked in the construction process, the problem of difficulties in draining the gas in the low-permeability coal bed with a drilling tool having insufficient capability in the drilling construction is solved, consequently, outburst risk of the coal and the gas is reduced, outburst accidents of the coal and the gas are reduced and even avoided, and mine production safety and lives of coal mine workers are ensured, so that the method brings great economic and social benefits.

The present application relates to an answer selection method and apparatus based on an improved attention mechanism and electronic device. The method comprises the following steps: performing word vector conversion on the obtained text data and the problem data respectively to obtain word vector representations of each word of the problem data and the text data respectively; processing the problem data and the text data through a loop neural network model to obtain a problem semantic vector representation and a text semantic vector representation, respectively; processing the problem semanticvector representation and the text semantic vector by stacking the attention mechanism layer to obtain a contextual representation of the fusion problem information; obtaining an answer correspondingto the question data from the text data based on the context representation of the fusion question information and the question semantic vector representation. In this way, the system for machine reading comprehension is optimized through a specific model architecture incorporating an improved attention mechanism to improve the effect of short text answer extraction.

The invention discloses an event atlas construction system and method based on multi-dimensional feature fusion and dependency syntax. The event graph construction method based on multi-dimensional feature fusion and dependency syntax is realized through joint learning of event extraction, event correction and alignment based on multi-dimensional feature fusion, relationship extraction based on enhanced structured events, causal relationship extraction based on dependency syntax and graph attention network and an event graph generation module. According to the event graph construction method and device, the event graph is constructed through the quintuple information of the enhanced structured events and the relations between the events in four dimensions, and the defects that in the priorart, event representation is simple and depends on an NLP tool, the event relation is single, and the influence of the relations between the events on event graph construction is not considered at the same time are overcome. According to the event atlas construction method provided by the invention, the relationships among the events in four dimensions can be randomly combined according to different downstream tasks, and the structural characteristics of the event atlas are learned to be associated with potential knowledge, so that downstream application is assisted.

The invention provides a remote sensing image building extraction method and system based on depth learning which comprises the steps of sample preparation, model training, precision evaluation, building prediction, merging and vectorization. The invention also relates to a remote sensing image building extraction system based on depth learning, a storage medium and an electronic device. Based onthe improved RCF boundary constraint model, the invention extracts the urban single building contour, the rural isolated building contour and the peripheral boundary of the rural building dense group,at the same time, a U-Net semantic segmentation model network structure is improved, and the improved U-Net is utilized to classify the images at pixel level. Finally, the two models are fused, and the depth learning model is trained by a large number of building sample label data, so that the network model by fusing the improved U-Net and the RCF is used to extract the buildings on the sub-meterGao Fen 2 remote sensing images, so that the automatic and effective building vector data extraction is realized, and the time cost and labor cost of manual rendering is greatly reduced.

A directional water pressure blasting method for improving coal bed gas permeability includes utilizing water pressure blasting splitting to change the structure of the coal and rock mass, firstly drilling holes inside a coal bed, performing high-pressure hydraulic cutting seam, and forming a deep flat slot around the drilling holes of the coal bed after cutting; and filling explosive into the drilling holes of the coal bed, then injecting water and sealing the hole, detonating the explosive to perform water pressure blasting, and enabling the periphery of the drilling holes of the coal bed to generate directional expanding fractures under the function of detonation pressures. Performing water pressure blasting in a high-gas low gas permeability coal bed can organically combine the waterpower and the explosive force of the explosive fully and reasonably, the water pressure blasting enables the cracks generated by hydraulic cutting seam to further expand and a directional expanding cross fracture net is formed on the periphery of the drilling holes of the coal bed, so that the gas permeability of the coal bed is greatly improved, the problem that the gas of the coal bed is hard to extract due to the small number of the cracks can be effectively solved, the concentrated stress of the coal bed is transferred to the deep of the coal mass, and the capacity of gas emission is improved. The directional water pressure blasting method for improving the coal bed gas permeability is simple in process, convenient to operate, safe, reliable, and good in controllability and effects.

The invention relates to a microblog-oriented emotion entity searching system. The emotion entity searching system comprises a user interface (1), a query expansion module (2), a query processing module (3), an emotive information mining module (4), an emotive information judging and index building module (5) and a reverse index building module (6). The user interface (1) is used for interaction between a user and the system, and the user can submit a query request through the user interface and obtain a feedback result; the query expansion module (2) is used for carrying out word relation mining on microblog corpus data and building a weighting word relation graph in combination with a WordNet ontology base; the query processing module (3) is used for converting the query request of the user into query key words or query statements and for carrying out query expansion on the basis of the word relation graph built by the query expansion module (2), wherein the query key words or the query statements can be accepted by an index base; the emotive information mining module (4) is used for performing emotion mining on the microblog corpus base and generating a judging rule for emotion entities and emotion polarities; the emotive information judging and index building module (5) is used for judging the emotion entities and emotion polarities, building an emotive information index and storing the emotive information index; the reverse index building module (6) is used for building a reverse index for microblog text information and storing the reverse index. The microblog-oriented emotion entity searching system solves the problems that difficulty exists in microblog emotion entity extraction, emotion polarity analysis, emotion entity search and the like, and a novel intelligent searching product is provided for analyzing and monitoring social networking public opinions.

The invention relates to an under-pressure grouting and hole sealing device and method for gas extraction drill air bag blocking. The device is characterized in that the device comprises an extraction pipe, a first tip air bag, a second tip air bag, a slurry outlet pipe and a grouting pipe, wherein the extraction pipe is composed of a first extraction pipe body, a second extraction pipe body and a third extraction pipe body, the first extraction pipe body is sleeved with the first tip air bag, a one-way needle valve is arranged on the first tip air bag, the slurry outlet pipe with a one-way check valve arranged inside penetrates through the first tip air bag, the outer end of the slurry outlet pipe is communicated with the grouting pipe, the third extraction pipe body is sleeved with the second tip air bag, a one-way needle valve is arranged on the second tip air bag, the two ends of the second extraction pipe body are connected with one end of the first extraction pipe body and one end of the third extraction pipe body respectively, the tip of the other end of the third extraction pipe body is provided with a plurality of sieve holes which allow gas to inflow, and the first tip air bag and the second tip air bag are filled with expansion agents in advance. The device and the method can be widely applied to permanent hole sealing for coal mine underground gas extraction hole drilling.

The invention discloses an image super-resolution reconstruction method based on residual learning, and the method comprises the steps: obtaining a public image data set for a training system, and processing the public image data set to obtain a training set and a test set; obtaining a to-be-processed low-resolution image; constructing and fusing all modules of a system core neural network to forma deep convolutional network based on residual learning; calculating network prediction and label loss, and adjusting network parameters according to the loss; training a deep residual network and inputting a to-be-processed low-resolution image to the deep residual network; outputting the reconstructed high-resolution image, evaluating the image reconstruction precision, and completing the imagesuper-resolution reconstruction. According to the invention, the extraction capability of image features is improved, the generalization and high efficiency of an image reconstruction system are improved, and the image reconstruction precision is more accurate.

The invention provides an attention model-based image identification method and system. The method comprises the steps of firstly obtaining an input feature graph with an image matrix in a shape [W,H,C], wherein W is width, H is height and C is a channel number; and secondly performing space mapping on the input feature graph by using a preset space mapping weight matrix, performing activation through an activation function to obtain a space weight matrix, and multiplying the space weight matrix by the image matrix of the input feature graph by bit to obtain an output feature graph, wherein the preset space mapping weight matrix is a space attention matrix [C,1] with attention depending on image width and height, at the moment, the shape of the space weight matrix is [W,H,1], or the presetspace mapping matrix is a channel attention matrix [C,C] with attention depending on the image channel number, at the moment, the shape of the space weight matrix is [1,1,C]. The pertinence of feature extraction can be effectively improved, so that the extraction capability of image local features is enhanced.

The invention relates to a construction method for enhancing the extraction effect of a ground coal bed gas, comprising the following steps of: (1) constructing a vertical cave well, forging caves on coal beds with coal bed gases needed to extract, and placing fiberglass casings into the caves; (2) constructing a vertical well section and a lower branched main horizontal well hole of a horizontal well, respectively forwards constructing the lower branched main horizontal well hole and a horizontal branched well hole positioned in front of a lower branched cave after butting with the vertical cave well, returning to a position nearby the rear end of the vertical cave well and then constructing a horizontal branched well hole positioned at the rear of the lower branched cave; (3) after a lower branched well hole is constructed, carrying out side drilling to construct an upper branched well hole; (4) after an uppermost branched well hole is constructed, plugging inclined well sections of main horizontal well holes inside the horizontal well from a side drilling windowing point to a coal seeing point by using cementing cement; (5) drilling and flushing the vertical cave well, installing discharge and extraction equipment, and extracting the coal bed gas by discharging water and reducing pressure on the ground. The invention can shorten the construction period, reduce the construction and production cost, decrease the lands used for construction and extraction and enhance the extraction effect of the ground coal bed gas.

The invention discloses a method for performing through-going seam cutting, pressure relief, permeability enhancement and quick tunneling on a soft and high-outburst coal seam, and in particular is applied to quick tunneling in a coal seam which is high in outburst danger and a laneway is difficult to support in a coal mine. The method comprises the following steps of: forming a network type fractured net by high-pressure jet flow seam cutting, pressure relief and permeability enhancement technologies, and enhancing an effect of extracting strip-type coal seam gas on a coal laneway subjected to through-going drilling and pre-extraction; and forming a network type filling region by pressing and injecting filling materials, so that the strength and stress distribution of a coal body are improved, the coal laneway is easy to maintain in the tunneling process, a phenomenon that the gas rushes out of a tunneled laneway is avoided, gas accidents are eliminated in the tunneling process, and an aim of safely and quickly tunneling the laneway is fulfilled.

A method for intelligently checking financial documents, a deive and a storage medium are provided in the invention which relates information extraction tecnology. The method comprises: combining natural language processing with information extraction, in the aspect of paragraph text, through effective word segmentation, part-of-speech analysis, named entity recognition, dependency syntax analysisand semantic role analysis, the effect of text information extraction and paragraph text semantic understanding being greatly improved, and the unstructured text data being transformed into structured financial tuples. For tables, use python, the best parser for text tables Docx transforms the table into a data structure that can be processed by computer, then divides the table into regions according to the header information, understands the table semantics, accurately realizes the table element location and element extraction, and transforms the semi-structured table data into structured financial tuples; Finally, the parsed financial structured tuples can be checked with the data of financial statements, and also can be checked for context consistency.

The invention relates to an event trigger word extraction method based on auto-encoder fusion document information, which comprises the following steps: generating a training set by using unlabeled free text corpora, and training a GRU model to construct an auto-encoder; pre-processing and labeling the training corpus, and extracting to-be-identified words; obtaining a document vector of the to-be-identified word in a document where the to-be-identified word is located by the auto-encoder, and taking the document vector as a global feature of the to-be-identified word; expressing word vectorsand entity types of the to-be-identified words in a distributed manner to serve as local features of the to-be-identified words; carrying out vector splicing on the global feature and the local feature to obtain context features of the to-be-identified word; inputting the context features to the Bi-GRU model to perform multi-classification to identify whether the to-be-identified word is an eventtrigger word and the corresponding event type of the to-be-identified word.

The invention relates to a method for preparing a health-care food (therapy) product (nutrient juice wine) by use of extracts from cordyceps militaris and cocoon, aiming at extracting multiple nutrient substances with natural activity from cordyceps militaris and cocoon wastes to produce different types of dietary supplements for fully improving the utilization rate of cordyceps militaris and cocoon. The method is characterized in that the nutrient components can be extracted from the cordyceps militaris and the cocoon through freezing and thawing, combined enzymolysis, and a water extraction and alcohol precipitation step-by-step extraction method, and the health-care food (therapy) product has three formulas (in part by weight) according to different functions (requirements): A. nutritive extract of cordyceps militaris and cocoon 85% and assistant 15%; B. nutritive extract of cordyceps militaris and cocoon 95% and assistant 5%; and C. nutritive extract of cordyceps militaris and cocoon 98% and assistant 2%, wherein the assistant can be starch, beta dextrin and the like, can be combined with one of A, B and C formulas to produce the health-care food (therapy) product with different efficacies through the processes such as sterilization, enzymolysis, precipitation, residue filtering, saccharification and fermentation.

The invention relates to the field of mine support and gas extraction, in particular to a gob-side entry retaining and sectional coal seam gas pre-extracting method for a single low-permeability outburst coal seam. A plurality of bedding drill holes are drilled in an air inlet crossheading and an air return crossheading of a working surface towards the working surface in the inclined direction of the coal seal in a parallel mode, and a pneumatic gas extraction device is used for carrying out gas extraction; gas of an adjacent working surface is extracted along the inclined bedding drill holes of the coal seam in the downstream direction of a transportation crossheading of the working surface, and an extraction pipeline is communicated for extracting gas of the coal seam of the adjacent working surface; after extraction, robbing is carried out on the coal seal of the working surface; extraction is not stopped until finish, and the steps are repeated until extraction of the coal seam is finished. According to the method, coal mining without pillars is achieved, the recovery rate of coal is improved, the total gas extraction time is saved, the gas extraction effect is improved, and coal mining and gas extracting are achieved at the same time.

The invention discloses a video classification method and device based on deep learning. The method comprises the following steps of acquiring a data set and a category label of the video; and equallydividing the obtained video into N segments, at the training stage, randomly selecting M frames for each video, stacking the video frames into a multi-channel image; and passing through a convolutional neural network, extracting video space weighted characteristics of the video divided into N segments, and passing through a convolutional neural network, extracting video time weighted characteristics of the video divided into N segments, and obtaining video multi-scale features of the video divided into the N segments, fusing the N segments of video features to calculate a prediction score, obtaining a final video classification, storing the trained video classification model, and classifying the test video by applying the trained video classification model. Through the mode, the effect and accuracy of video feature extraction can be improved.

The invention discloses a method for reducing the content of cadmium in middle rice. The method includes the following steps that after middle rice of the first year is harvested, middle rice straw is returned to the field directly, and tillage is conducted so that the middle rice straw can be evenly mixed with soil; hyperaccumulation plants are planted in a paddy; after the hyperaccumulation plants are removed from March to May in the second year, soil heavy metal passivator is scattered on the surface of soil of the paddy; tillage is conducted, and base fertilizer is applied 3-5 days later; tillage is conducted, and rice is transplanted; during planting of rice, a foliar resistance control agent is sprayed in the later tillering stage and the booting stage of rice respectively, wherein spraying time interval is 5-10 days; field drying is conducted in the later tillering stage, natural drying is conducted in the maturation stage, and flooding irrigation is conducted in other development stages. According to the method, crop production and straw resource utilization are not affected, the dual targets of reducing the paddy soil cadmium content and reducing the rice cadmium content are achieved under the combined action of plant extraction, soil passivation, foliar resistance control and water management, and effects are more remarkable compared with other combination technologies.

The invention provides a microwave-assisted extraction and hydrofracture cooperative coal seam anti-reflection method and belongs to the technical field related to coal mine down-hole hydraulic fracturing. The method includes the steps that a microwave antenna is connected to the innermost end of a coaxial waveguide tube and sent into a fracturing drill hole, the outer most end of the coaxial waveguide tube is connected with a waveguide tube converter, the waveguide tube converter is connected with a rectangular waveguide tube, and the rectangular waveguide tube is connected with a microwave generator; organic solvent and high pressure water are mixed, hydrofracture is conducted on the drill hole through a fracturing tube, meanwhile, the microwave generator is turned on, radiation drilling is conducted through the microwave antenna, fracturing is conducted, a seepage is provided for the solvent, the organic solvent dissolves small organic molecules in coal so that a chambering effect is generated, and methane desorption is promoted through the heat effect of microwaves. According to the method, hydrofracture, solvent extraction and microwave radiation are combined, the porosity and permeability of the coal seam are greatly improved, methane desorption is promoted, and thus the methane extraction effect is greatly improved.

The invention relates to the field of flow detection, as an advanced low power consumption two-wire system vortex-shedding meter, comprising a piezo-electric transducer, a differential charge amplifier, a voltage amplifier, a programmable amplifier, a low pass filter, a voltage follower, a band-pass filter set, a band-pass select switch circuit, a peak detection circuit, a wave shaping circuit, asinglechip, a human-computer interface circuit, a 4-20mA output and power management circuit, a constant flow source, a temperature sensor, a pressure sensor, a differential amplifier and a 16-bit analog-to-digital converter. Mini FFT spectral decomposition is used to be combined with the analog band-pass filter set to process signals of a vortex sensor and a multicycle frequency measurement method of precision is used, thereby having temperature and pressure compensating circuit and function, outputting 4-20mA direct current and having two-wire system work. The advanced low power consumptiontwo-wire system vortex-shedding meter has greatly improved measure ratio, measurement precision and capacity of resisting disturbance.

The invention provides an edge-enhanced multi-scale remote sensing image building semantic feature extraction method. The method comprises the steps of obtaining a large amount of remote sensing imagedata to construct a remote sensing image data set Images, and obtaining a building semantic binary image label Masks, namely a real label, corresponding to each remote sensing image in the remote sensing image data set Images; constructing a building semantic extraction network EEMS-Unet model; training the building semantic extraction network EEMS-Unet model by utilizing the remote sensing imagedata set Images and the corresponding building semantic binary image label Masks to obtain the trained building semantic extraction network EEMS-Unet model; and inputting a remote sensing image to besubjected to building semantic feature extraction into the trained building semantic extraction network EEMS-Unet model, and extracting building semantic features in the remote sensing image to obtain a pixel-by-pixel prediction result Maskpred corresponding to the remote sensing image. According to the method, the problems of incomplete extraction, incomplete details and poor edge effect of buildings with different scales in high-resolution remote sensing images such as satellites and unmanned aerial vehicles can be effectively improved.

The invention relates to a method for preparing nutrient (liquid) beverages and granules of healthcare food (therapy) products by using extractives of cordyceps militaris and cocoons and aims to meet market requirements by producing nutrient healthcare products with different types by using various nutrient substances extracted from waste cocoons of the cordyceps militaris. The method is characterized by comprising the following steps of: firstly extracting the nutrient substances from the cordyceps militaris and the cocoons by adopting freeze thawing, combined enzymolysis and a water-extraction alcohol-precipitation sequential extraction method, and producing nutrient liquors and granules with different functions according to different functions (requirements) and different proportions. According to the method for preparing the nutrient (liquid) beverages and the granules of the healthcare food (therapy) products by using the extractives of the cordyceps militaris and the cocoons, the synergistic effects among nutritional ingredients are adequately exerted, and the use ratios of the cordyceps militaris and the cocoons are improved.

The invention relates to a data extraction method, a device, computer equipment and a readable storage medium. The data extraction method includes: segmenting to-be-processed data to obtain data sets;inputting the data sets into a preset neural network model to obtain initial extracted data and feature labels corresponding to the initial extracted data; extracting corresponding target extracted-data from the initial extracted data according to a preset rule template; and carrying out association output on the target extracted-data and feature labels corresponding to the target extracted-data.According to the above-mentioned data extraction method, no limit of an extraction rule exists when the different-format data are extracted, data extraction is carried out for mapping relationships of the data sets and the feature labels through custom extraction rules, an error rate in extracting the different-format data can be decreased, and an extraction effect is better.

The invention discloses a method for preparing health food capsules by using cordyceps militaris and cocoon extracts, and aims to extract various nutrient substances with natural activities from the cordyceps militaris and cocoon to produce different types of nutrient health products, so that the utilization rate of the cordyceps militaris and cocoon is fully improved, and the market demand is met. The method is characterized by comprising the steps of extracting the nutrient substances from the cordyceps militaris and cocoon by performing freeze thawing, combined enzymolysis and aqueous alcohol precipitation sub-step extracting methods; adding a proper quantity of auxiliary materials into the nutrient substances to fully and evenly stir and sieve to prepare the particles; filling the particles into the capsules; wiping off residual powder adhered to the capsule shells; waxing; polishing; and sterilizing the capsules in a radiation mode after the capsules are packaged so as to obtain the finished product.

The invention relates to a porous linear control hydraulic fracturing method. The method comprises the following steps of: conducting hydraulic fracturing drilling on a pre-mined coal-rock mass in a coal mining working face, a transportation lane or a return air lane till the set positions; commonly connecting a plurality of hydraulic fractured holes to a high-pressure water injection pump by adopting an equal-quantity diverter; producing an oval-like water pressure crack band around each hole under the action of pore water pressure; and enabling the water pressure crack bands of the adjacent drilled holes to run through each other by controlling the spacing interval between the drilled holes to realize linear control fracturing. According to the method disclosed by the invention, oriented slitting of a coal-rock layer is realized, and the method has the advantages of being conductive to oriented processing of a hard top plate, fracturing of hard top coal in the vertical direction, oriented transfer of stress, control of rock burst and control of coal and gas outburst, and improving the effect of gas extraction. The method is simple, beneficial for construction convenience, safe, reliable and good in effects, and has extensive practicality.

The invention belongs to the technical field of coal mine gas disaster prevention and control as well as coal-bed gas extraction, and discloses a simulating device and method for a diameter change rule of a gas extraction borehole. The device comprises a true triaxial pressure test device, an optical borehole deformation observation device, a coal mass crack evolution observation device and a main machine. The magnitude of stresses in different directions is changed through the true triaxial pressure test device, and the diameter change rule of coal masses with different strength at different stress states can be researched by the aid of the optical borehole deformation observation device; real-time observation and image shooting can be performed on crack evolution of coal masses around the extraction borehole through a microscope and a camera of the coal mass crack evolution observation device. The device and the method have the important significance in determination of the reasonable gas extraction time and the effective extraction radius as well as optimization of the borehole arrangement manner.

The invention discloses an information system fault diagnosis method based on a convolutional neural network. The method comprises the steps that operation monitoring data is acquired; the operation monitoring data is utilized to create an operation monitoring data matrix; the operation monitoring data matrix is input into a fault diagnosis model to obtain a diagnosis result, wherein the fault diagnosis model is obtained by utilizing a historical operation monitoring data matrix and an operating state label to train the convolutional neural network. Obviously, the convolutional neural network is utilized to create the fault diagnosis model, fault diagnosis is performed on the operation monitoring data, the diagnosis result is obtained, and therefore a user can maintain an information system through the diagnosis result; by use of the convolutional neural network, feature extraction capability and classification capability are improved, diagnosis accuracy is improved, and support is provided for secure operation of the information system. Besides, the invention accordingly discloses an information system fault diagnosis device based on the convolutional neural network.

The invention relates to a horizontal well forced gas drainage method with coal bed gas subjected to staged fracturing. In order to solve the problems that hole forming performance of horizontal drilling in a coal bed is poor and buried drill accidents can happen easily and to achieve the forced gas drainage effect on tectonic soft coal, the well track of a horizontal well is controlled in mudstone or sand stone with a certain range above a coal bed top boundary, and the horizontal well is docked with a drainage vertical well at a far end. An ultra deep penetration perforating technology is used, according to the requirement for certain hole density and orientation, orientated perforating is conducted downwards on the horizontal well section in a segmented mode, a steel sleeve, a well cementation cement ring and a mudstone top plate are penetrated, and the horizontal well is communicated with a target coal layer below. Then, a clear water sand carrying fracturing technology is used, staged fracturing is conducted on the target coal bed according to the requirements for large discharge capacity, large liquid amount, high prepad fluid ratio, large sand amount and middle sand ratio, and the purpose of increasing production is achieved.

A Group III nitride semiconductor light-emitting device includes an electrically conductive support; a p-electrode provided on the support; a p-type layer, an active layer, and an n-type layer, which are formed of a Group III nitride semiconductor and are sequentially provided on the p-electrode; an n-electrode which is connected to the n-type layer; a first trench extending from the surface of the p-type layer on the p-electrode's side to reach the n-type layer; an auxiliary electrode which is in contact with the surface of the n-type layer serving as the bottom of the first trench, but is not in contact with the side walls of the first trench; and an insulating film which exhibits light permeability and covers the auxiliary electrode and the bottom and side walls of the first trench.

The invention discloses a parallel multichannel convolutional neural network and a construction method thereof, and an image characteristic extraction method and relates to the machine learning technology field. As depth sensors such as Kinect are widely applied, acquisition of a multi-mode image is made to be more convenient, so characteristic extraction research based on the multi-mode image information is of great importance. According to the method, a sub mode convolutional neural network model is established for each channel of the multi-mode image, and each mode depth characteristic vector is extracted; for acquiring multi-mode image characteristics having unified property, weight connection is established at a special full-connection layer by each mode sub network; in the full-connection portion of the multi-layer sub network, the multiple mode depth characteristic vectors are fused according to the weight ratio to form a fusion characteristic vector containing the information of each mode, and characteristic expression with lower dimensions and better expressive force can be acquired through multi-layer network training. The method is advantaged in that the characteristic expression can be applied to the identification and classification fields.