173results about How to "Avoid misclassification" patented technology

A method of operating a point-of-sale (POS) system (1), the POS system comprising a POS terminal (3) having a software module (17, 21) thereon for enabling a retailer to process transactions within a transaction environment, and a peripheral device (5, 7) in communication with the POS terminal (3), the POS system (1) further comprising a driver software module (40) installed between the POS terminal (3) software module (17, 21) and the peripheral device (5, 7), the method comprising: receiving, at the driver software module (40), data sent between the software module (17, 21) and the peripheral device (5, 7) in communication with the POS terminal (3); communicating with a further device (44, 60, 64) in dependence on the data received at the driver software module (40); receiving modified data from the further device (44, 60, 64); and outputting the modified data.

The invention discloses a three-dimensional convolutional neural network (3D CNN) based video classifying method and belongs to the technical field of video processing. According to the method, a video is sampled at equal intervals to obtain a plurality of video segments, a video database is amplified, three-dimensional video segments are directly input into a 3D CNN, and time domain and space domain characteristics of the video are extracted, so that the limitation of a conventional video classifying method in manually selecting video characteristics and video modeling modes is improved. A parallel distributed 3D CNN multi-classification model lowers the complexity in learning the 3D CNN and enables a classification system to realize distributed parallel computation more conveniently. Relatively high identification rate can be achieved with only fewer video segments based on a 3D CNN multi-classification system, and videos not belonging to any type can be classified into new type, so that the classification error of the new type is avoided.

The invention discloses a method for constructing a face recognition model based on a ParaSoftMax loss function, which comprises the following steps: selecting a basic convolution neural network modelaccording to an application environment of a task; acquiring a face image marked with human face identity information in a specified number as a training data set; the decision edge parameters are obtained according to the difference of the class center angles between the difficult sample eigenvectors and the simple sample eigenvectors and the class center angles in the basic convolution neural network model. Obtaining a ParaSoftMax loss function according to the decision edge parameter; setting the loss function at the last layer of the basic convolution neural network model to form a face recognition model based on the loss function; input the training data set to the face recognition model, minimizing the loss function iterative training model parameters, and obtaining the optimal facerecognition model. Thus, the face recognition model of the present application can improve the accuracy of face recognition.

The invention belongs to the technical field of image processing, and discloses a method and system for extracting ground object spatial spectral features of a hyperspectral remote sensing image. Themethod comprises the steps of training and extracting spectral features through an auxiliary classifier generative adversarial network; performing band selection, and extracting spatial texture features with rotation invariance from a selected band; and forming spatial spectral features of the ground object through splicing the spectral features and the spatial texture features. Meanwhile, the invention discloses a hyperspectral remote sensing image classification system which adopts the ground object spatial spectral features is based on a convolutional neural network. The method and system verify that the ground object spatial spectral feature extraction technology disclosed by the invention not only can better characterize ground object information, but also can obtain higher classification accuracy with fewer labeled data sets.

The invention relates to an oil seal press fitting detection production line. The production line comprises a feeding device, an oil seal press fitting device, a height and parallelism detection device, an airtightness detection device, a separation device and a transfer device in sequence, wherein the feeding device, the oil seal press fitting device, the height and parallelism detection device, the airtightness detection device, the separation device and the transfer device are subjected to whole-process control of a computer respectively. According to the oil seal press fitting detection production line, computer integrated control is adopted, so that the automation degree is high, the detection accuracy is high, the machining efficiency is greatly improved, the labor intensity of workers is reduced, the parallelism, the height, the verticality and the airtightness of a product are guaranteed, and the defective rate of the product is reduced.

The invention is applicable to the technical field of image processing, and provides an image classification method, a device, a terminal device and a readable storage medium. The method comprises thefollowing steps: training a depth convolution neural network through a known class image to obtain a network training model; establishing a probability distribution model for each class of samples inthe known class image according to the network training model; correcting an activation value of the known class image according to the probability distribution model; obtaining an activation value of an unknown class image according to the activation value of the known class image data. The images are classified according to an activation value of the known category image and an activation valueof the unknown category image. The invention can reasonably and accurately classify the images other than the training set category in the known category images in practical application.

The invention discloses a deep neural network training method, belongs to the technical field of computers, and is used for solving the problem of relatively low performance of a trained neural network in a complex scene in the prior art. The method comprises the following steps: obtaining a plurality of training samples provided with preset category labels, and training a neural network model based on the plurality of training samples; wherein the loss function of the neural network model is used for carrying out weighting operation according to a first weight value in direct proportion to the distinguishing difficulty of each training sample, and determining the loss value of the neural network model. According to the deep neural network training method disclosed by the embodiment of theinvention, the importance of the training samples with higher difficulty in distinguishing in the training samples is adaptively improved, so that the samples with higher difficulty in distinguishingare prevented from being mistakenly classified by the neural network obtained by training, and the performance of the neural network is improved.

The invention discloses a zero sample learning method and system based on a semantic attribute attention redistribution mechanism, and the method comprises the steps: (1) building a neural network model based on the semantic attribute attention redistribution mechanism; (2) redistributing the weight between the semantic features by using the attention of the semantic attribute space; (3) traininga neural network model by using the image data set with the label; (4) calculating the similarity between the weighted semantic features of the images and the semantic prototypes of the unknown classes, calculating the similarity between the hidden layer features and the hidden layer feature prototypes of the unknown classes, and adding the two similarities to obtain the similarity between the test image and each unknown class; and (5) sorting according to the similarity with the types, and the type with the maximum similarity is selected as the type prediction of the image. According to the method, in the training process of zero sample learning, the semantic space and the hidden layer space can be more closely linked, and a joint classification result combining the two spaces is more robust.

A driver assistance system for motor vehicles, having a localization system for localizing objects in the surroundings of the vehicle and having a device for recognizing stationary objects by comparing the difference between the relative motion of the object and the inherent motion of the vehicle with a threshold value, wherein the device is embodied to vary the threshold value as a function of variables that influence the accuracy with which the relative and inherent motions are determined.

A driver assistance system for motor vehicles, having a localization system for localizing objects in the surroundings of the vehicle and having a device for recognizing stationary objects by comparing the difference between the relative motion of the object and the inherent motion of the vehicle with a threshold value, wherein the device is embodied to vary the threshold value as a function of variables that influence the accuracy with which the relative and inherent motions are determined.

The invention relates to the field of a two-dimension code application system, in particular to a system and a method for preventing misoperation in an electric power system by using two-dimension codes. The method for preventing the misoperation in the electric power system by using the two-dimension codes sequentially comprises the following steps that: (1) a two-dimension code database is manufactured; (2) two-dimension code pictures are extracted and manufactured; (3) the two-dimension code pictures are pasted to corresponding positions of equipment; (4) a mobile operation order system is built; (5) an operation order is manufactured; (6) two-dimension code information of the equipment is implanted into relevant items of the operation order; (7) the operation order is pushed to a mobile terminal; and (8) mobile terminal equipment is used for scanning the two-dimension codes of the equipment. The system and the method provided by the invention have the advantages that the functionality is high; the operation by operators is convenient; great importance is realized on safety risk elimination, personal injury and equipment damage avoidance and safe operation guarantee; and the system and the method are suitable to be widely popularized and applied.

The invention discloses a target tracking method based on fuzzy learning. The method mainly comprises the steps that firstly, a sparse representation algorithm is used to track a target in the aspect of a tracker; secondly, in the aspect of a detector, cascade classification detectors (a variance classifier, a collection classifier and a nearest classifier) are used to detect the target position; and finally, in the aspect of learning update, a fuzzy learner is used to integrate the output results of the tracker and the detector, and the final target position is acquired according to the membership of four constraints of time continuity, the spatial uniqueness , the similarity and the target size consistency. According to the target tracking method based on fuzzy learning, the real-time performance is ensured, and at the same time the adaptability to the target illumination change is great; and the discriminant ability of a learner is improved; the tracking accuracy and robustness of the algorithm are improved; and the method has important theoretical and practical significances for the research and practical application development of target tracking.

The invention discloses a Chinese question classification method. The invention aims to solve the problem that the classification accuracy is not high enough due to the fact that faulty wording, wrongly written characters, needless characters and the like exist in input questions and the inherent defect that an existing classification method is single. The Chinese question classification method comprises the steps: 1, obtaining Chinese question text data; 2, preprocessing the Chinese questions; 3, carrying out error correction by utilizing a language model; 4, vectorizing the Chinese questionsby using a word vector tool; 5, obtaining an intermediate semantic matrix vector containing semantic information by utilizing the bidirectional gating circulation unit layer; 6, generating an attention matrix vector by using a self-attention mechanism; 7, extracting a plurality of local features by utilizing a plurality of convolution kernels with different sizes, and obtaining a global feature matrix vector through pooling and splicing; and 8, outputting probability distribution of a corresponding category by utilizing a full connection layer and a normalization exponential function, and taking the category with the maximum probability value as a predicted category, namely, a result of Chinese question classification. The Chinese question classification method is applied to the field ofnatural language processing.

The invention relates to a wire cable automatic branching device which comprises a branching device, a wire pressing device, a wire stirring device, a CCD camera arranged on the outer side of a frame body and at least one die. The branching device, the wire pressing device and the wire stirring device are arranged on the frame body and are matched with each other. A branching disc is arranged on the branching device. A wire groove is formed in the side periphery of the branching disc. The wire groove can clamp a rotating wire harness and move the wire harness to any wire outlet which is communicated with the die. A wire pressing door is arranged on the branching device. The wire pressing door can press a plurality of wire harnesses to the side periphery of the branching disc. The wire stirring device can push the wire harnesses to the die from the wire outlets in a moving mode. Automatic branching is carried out by matching of the branching disc and the wire pressing door, the CCD camera is used for color distinguishing, the wire stirring device is used for pushing the wire harnesses into the corresponding wire groove, branching high efficiency and automation are achieved, and meanwhile branching errors caused by manual branching are avoided.

The invention discloses a mean shift based grey relation infrared imaging target segmentation method, which mainly solves the problems of over-segmentation phenomenon and low segmentation precision in the traditional similar method. The method comprises the following steps of: (1) performing the mean shift filtering on an original image, and determining whether to calculate the mean shift convergence value according to the similarity of a current pixel point and neighborhood points; (2) merging and marking pixels of the same type to obtain an initially segmented image; (3) respectively selecting background reference values and target reference values from regions of the initially segmented image; (4) calculating the grey relation coefficient between various regions and the reference values; and (5) searching for thresholds by the utilization of inflection points of a grey relation coefficient curve, and realizing the dynamic segmentation of the image by integrating three single thresholds. The method has the advantages of fast arithmetic speed, high segmentation precision, good stability and strong adaptability, and can be used in military or civil systems such as precise infraredguidance, target detection and fire control, optical remote sensing, night navigation and the like.

The invention discloses a 24 KV central-positioned switchgear that includes a cabinet (1), a relay instrument room (21) pisitioned on top of the handcart room (18), a busbar room (19) and a cable room (20) positioned on back side of the handcart room (18), the cable room (20) is positioned under the busbar room (19), the main busbar (10) is extracted with two branck busbars (9) that connected with an upside bush (8) by support of a support insulator (3), a downside electric-shock case (15) is connected with a current transformer (4) disposed in the cable room (20), the upper and lower metal valve corresponds to upper and lower electric-shock case (8) in the handcart room (18), a handcart (13) is disposed on guide rail (22), the upper and lower bush (16) on forehead of the handcart (13) implements disconnection and connection with the upper and lower electric-shock cases (8,15) along with movement of the handcart, an interlock structure is provided between the handcart (13) and a groundswitch operation shaft (7).

The invention belongs to the field of intravascular stent image segmentation, particularly relates to an intravascular stent image segmentation method and system based on a double attention mechanism,and aims to solve the problem that an intravascular stent cannot be accurately segmented from an intraoperative X-ray transmission image in real time in the prior art. The present invention comprises: an X-ray transmission to-be-detected video sequence is acquired, and a segmentation mask sequence for displaying an intravascular stent is generated through a lightweight upper attention fusion network based on deep learning based on the to-be-detected video sequence, and the to-be-detected video sequence is covered with the binary segmentation mask for displaying the intravascular stent to generate a video sequence for displaying the intravascular stent. According to the invention, the accuracy of intravascular stent image segmentation is improved by adopting the feature attention blocks and the associated attention blocks, model training is carried out by adopting the Dice loss function and the focusing loss function, wrong classification of edge pixels is avoided, and the performanceof an image classification network is improved.

The invention relates to a process method for automatically separating a cable. The method includes the following steps of firstly, clamping the cable between a cable pressing door and a cable separating disc; secondly, clamping cable bundles into a cable groove of the cable separating disc; thirdly, distributing the positions of the cable bundles according to shooting through a CCD camera; fourthly, moving the cable bundles to corresponding cable outlets through the cable separating disc and meanwhile making the positions of cable grooves in a die correspond to the cable outlets; fifthly, pushing the cable bundles into the corresponding cable grooves through a cable stirring device; sixthly, executing the second step, the third step, the fourth step and the fifth step repeatedly till the cable bundles of the cable are all distributed and clamped into the cable grooves of the die; seventhly, correspondingly transferring the cable bundles on the die to cable grooves in a circuit board one by one. According to the method, the cable is automatically separated through the cooperation of the cable separating disc and the cable pressing door, the color distinguishing is conducted through the CCD camera, the cable bundles are pushed into the corresponding cable grooves through the cable stirring device, and therefore the high efficiency and automation of cable separation are achieved, and meanwhile wrong separation caused by manual cable separation is avoided.

The invention designs a method for realizing unmanned aerial vehicle group formation reconstruction based on a genetic algorithm and a Dubins algorithm. The method specifically comprises the followingsteps: numbering unmanned aerial vehicles, establishing a position matching relation of each unmanned aerial vehicle in a new formation, and consequently completing coding of chromosomes; improving the Dubins algorithm, building an air route planning model, evaluating distance of completing reconstruction flight by a wing unmanned aerial vehicle; and allocating a reconstruction target position for each unmanned aerial vehicle based on the genetic algorithm. In the method provided by the invention, formation reconstruction is divided into task allocation and air route planning, relative to theexisting formation reconstruction algorithm, more stable air routes can be obtained, moreover, speed range and radius of turning circle of the unmanned aerial vehicles are considered, the air routesgenerated can be more rational and can be used in actual application more easily. In the method provided by the invention, by a mode of limiting variation and intersecting, each unmanned aerial vehicle is guaranteed to have a position allocated, situations of missing of allocation and allocating in mistake can be prevented, and quality of task allocation is improved further.

The invention discloses banknote-coin sorted processing device, comprising a pedestal, a banknote expansion unit and a coin collection sorting unit, wherein the banknote expansion unit and the coin collection sorting unit are fixed on the pedestal; a banknote-coin separation unit is arranged above the coil collection sorting unit; the top of the banknote-coin separation unit is provided with a coin feeding port; the bottom of the banknote-coin separation unit is provided with a banknote exit and a coin exit; the banknote exit is communicated with the entrance of the banknote expansion unit; and the coin exit is communicated with the entrance of the coin collection separation unit.The banknote-coin sorted processing device is high in automation degree and high in separation efficiency, greatly reduces the labor intensity, and saves cost. The banknote-coin sorted processing device is reasonable in design, good in stability, high in accuracy and strong in controllability.

A novel trash sorting and recycling method includes the steps of firstly, obtaining essential information of an article from a manufacturer database according to the ID if trash contains the automatic identification ID of any type, sorting and storing the ID, and adding a record to an Internet of Things database; secondly, searching the area surrounding a trash can for articles with the automatic identification ID through physical space relevancy scanning if trash does not contain automatic identification ID of any type, searching trash cans in a whole living quarter or a whole industrial district if necessary, and executing the third step if no automatic identification ID can be found. The identification and search technology under pervasive computing is adopted, and the trash sorting efficiency is greatly improved.

The present invention discloses a vegetation loss direction identification method based on a multi-remote-sensing index trend. The method comprises: calculating a temporal similarity of vegetation indexes between each year and a beginning year by using a JM distance to generate a track of temporal similarity of vegetation indexes; extracting a potential vegetation loss region according to a variation of the temporal similarity of the vegetation index, so as to define a region where the vegetation index is significantly decreased and an impervious surface index is significantly increased as a vegetation loss region; and on this basis, finally determining different vegetation loss directions such as urbanization, desertification and wetland formation according to a water body index and a bare soil index trend feature. In the method, the vegetation change region is determined by using the variation of the temporal similarity, and further, the vegetation loss direction is determined according to multiple remote sensing indexes, without depending on manual intervention for threshold setting, so that the method has the characteristics of high robustness, high classification precision, high automation and storing anti-interference ability, and so on.

The present invention relates to a chip testing and sorting method which comprises the steps of (S1) providing a tester and a multi-station sorter, (S2) connecting the tester and the communication interface of the sorter through a cable, and connecting the test port of the tester and the station of the sorter, (S3) reserving a station which does not pass a test of the sorter and closing other stations of the sorter, (S4) testing whether the cable connection relation of the currently reserved station is correct or not, showing that the currently reserved station passes the test if the connection relation is correct, repeating the step (S3) until all stations pass the test, adjusting the cable connection relation of the currently reserved station if the connection relation is wrong, and repeating the step (S4), and (S5) starting all stations of the sorter, and carrying out batch test sorting of chips. According to the chip testing and sorting method, the wrong classification problem caused by the wrong cable connection of multiple stations can be effectively avoided.

The invention relates to the technical field of power quality analysis and monitoring, and mainly relates to high-identification rate complex power quality disturbance signal identification method based on multiple features. Power quality signals acquired by monitoring equipment of a power quality monitoring point is used to serve as a to-be-identified disturbance signal type, that is, input of an automatic identification system, and the type of the disturbance signals is outputted through automatic identification. According to the method, multiple kinds of single disturbance existing in the power system can be identified, multiple kinds of complex disturbance can be precisely identified, an auxiliary decision is provided for management and governance on the power quality, and important practical significance is provided.

The invention discloses an intellectual property intelligent service method and system, and the method specifically comprises the steps: extracting a word vector from an input text of a user through amachine learning algorithm, and carrying out the entity labeling of the word vector; wherein the labeling result and the user intention form an association pair; and carrying out data training on theuser intention classification system by using a plurality of association pairs generated by manual judgment as a training data set, and generating a prediction model. The classification system specifically comprises: word vector extraction and entity recognition are conducted on user intentions, and the user intentions are distributed to different semantic processing systems for specific processing. According to the method, the input text processing problem in a composite user demand environment is solved, the input text of the user is distributed to different semantic processing systems through intention classification, optimal feedback is made, and the feedback accuracy is effectively improved.

The present invention discloses an open-pit mine typical ground object classification method based on a UAV image. According to the method, firstly the image is subjected to multi-scale segmentation to obtain an object layer suitable for different ground object extraction, then the features (including a spectrum feature, a texture feature, a morphological feature, and a customized feature) of a typical ground object are subjected to correlation analysis, a feature with large correlation is excluded, at the same time the dimension reduction of a feature space is carried out, thus a feature set with the most facilitation of classification is obtained, finally five features are selected from the feature set according to the concrete feature of each type of ground object, and a classification result is obtained and then postprocessing (category merging, edge smoothing and misclassification category adjustment) is carried out to optimize the classification result. The method has the advantages of high accuracy, high degree of automation and simple processing process, the bare soil and stope confusion problem in an open-pit mine can be effectively solved, and the method has a very important significance in open-pit mine ground object typical ground object feature classification.

The invention discloses an electrical load identification method based on an improved graph convolutional neural network, and belongs to the technical field of intelligent power utilization and intrusive load identification, and the method comprises the steps of collecting the power utilization data of a user at a power utilization side, and carrying out the standardized processing of the data; and taking the power consumption data of the user as a training set and a test set of the graph convolutional neural network for advanced training. then, evaluating the overall distribution characteristic, the local trend characteristic and the overall trend characteristic of the load curve of the electrical appliance by applying an Euclidean distance DTW to the acquired load curve, and performing weighted fusion on the three characteristic distribution weights by applying an entropy weight method; and then clustering the load curve of the electrical appliance by adopting a k-means clustering algorithm and applying a method for automatically generating a clustering number K value based on a DBI value as a measurement scale. And finally, taking the clustered electrical appliance load curve asan input set and inputting the input set into a graph convolutional neural network for electrical appliance identification. The trained graph convolutional neural network model identifies the corresponding load curve, and finally draws a probability density distribution curve of the applied electrical appliance.

The invention discloses an intelligent garbage classification garbage can, and belongs to the field of garbage classification. The intelligent garbage classification garbage can comprises a garbage can body, wherein four garbage bins are placed in the garbage can body; a garbage throwing port is formed in the top part of the front surface of the garbage can body; an automatic garbage door is slidably connected to the inner side of the garbage throwing port; and a second infrared object sensor is fixedly arranged on the inner side of the garbage throwing port and is located on the front surfaceof the automatic garbage door. According to the intelligent garbage classification garbage can provided by the invention, garbage can be automatically classified, so that the garbage classification error caused by the reason that people do not know the garbage types is avoided, the garbage can be conveniently classified and recycled, and the contribution is made for environmental protection; through shared power banks and shared umbrellas, the functions of the garbage can are enriched; and due to a solar panel fixedly arranged on the top part of the garbage can body, electric appliances in the garbage can body can be powered, so that the energy loss is reduced, and the device can be used in an environmental protection and conservation way.

The invention discloses a shape adaptive convolution deep neural network method for hyperspectral image classification. The method comprises the following steps: adopting a spatial structure information learning branch; using a shape self-adaptive convolution kernel based on a guide graph and the shape self-adaptive convolution kernel can be trained; a spectral dimension one-dimensional convolution layer and a spatial dimension two-dimensional convolution layer forming a space-spectrum feature extraction unit, and each unit having two inputs, namely a feature map and a guide map; wherein the deep network is formed by stacking a plurality of space-spectrum feature extraction units, and a skip layer connection is established between every two feature extraction units; wherein the network loss function is weighted cross entropy. through learning the spatial correlation between adjacent pixels in the space-spectrum data is learned, the receiving domain shape of convolution operation can beadaptively adjusted according to the spatial structure relationship between explicit definition pixels, the defect that anisotropic characteristics cannot be captured by fixed square convolution is overcome, and the method has excellent classification and generalization performance for hyperspectral images with different resolutions and different scene complexities.