262results about How to "Reduced ability to recognize" patented technology

The invention discloses a three-dimensional point cloud reconstruction device and a three-dimensional point cloud reconstruction method based on a multi-fusion sensor. The device comprises a multi-fusion sensor module, a system data processing module and a three-dimensional point cloud reconstruction module. The method comprises the following steps: acquiring point cloud data and a video image byusing a multi-fusion sensor device, and processing the video image to obtain a target image including target posture, position and texture information; acquiring target point cloud corresponding to the original point cloud according to the acquired posture position information in the original point cloud and the target image; filtering the point clouds to remove messy point clouds; wherein the point cloud coarse registration comprises key point quality analysis; point cloud fine registration; and inputting the registered point cloud into a point cloud reconstruction unit to realize three-dimensional point cloud fusion and reconstruction. According to the method, the feature points corresponding to the target image reconstruction point cloud corresponding to the original three-dimensional point cloud are extracted to obtain the target three-dimensional space position information, texture and color information. Therefore, the point cloud registration precision is improved, and the falseidentification and tracking loss probability is effectively reduced.

The invention belongs to the technical field of infrared imaging, and specifically relates to a high-precision non-uniformity correction method for dynamic adjustment of integration time. The method of the invention comprises the following steps: determining linear relation between a detector response value and the integration time span under a fixed temperature; performing non-uniformity correction; calculating a non-uniformity correction parameter; performing self-adaptive adjustment of the integration time span according to an radiation temperature, determining the section that the integration time span is in, generating the non-uniformity correction parameter in real time in the current integration time, and completing the non-uniformity correction process. The method of the invention realizes continuous adjustability of the integration time in the working process of an infrared detector, solves the problem of inadaptation between the integration time span and the non-uniformity correction parameter of the infrared detector when the change range of the temperature of the scene is large, reduces influence of switching of the integration time on detector performance and recognition tracking, and can give full play to the detection performance of the detector in the working process.

The invention discloses a mobile phone wearing mode-oriented falling event detection system and a mobile phone wearing mode-oriented falling event detection method. The system comprises an acceleration sensor installed on a mobile phone, and a data acquisition module, a falling pre-judgment module, a characteristic extraction module and a falling judgment module, which are sequentially connected with the data output end of the acceleration sensor. The method comprises the following steps that: firstly, the data acquisition module continuously acquires a wearer motion variation sensed by the acceleration sensor, squares components of three directions of acceleration, summates squares, extracts a sum, calculates a module of an acceleration vector, and simultaneously maintains a cache for recording historical acceleration information; secondly, the falling pre-judgment module periodically reads the data cache maintained by the data acquisition module, and judges whether falling possibly happens or not; thirdly, the characteristic extraction module acquires acceleration variation data a plurality of seconds before and after an event from the data acquisition module, and calculates characteristic components; and finally, the characteristic extraction module generates a characteristic vector by using the obtained characteristic components, and inputs the characteristic vector into the falling judgment module, and the falling judgment module which is obtained according to distinction motion training required by the use of the mobile phone judges whether the event belongs to a falling event or not.

The invention discloses an RFID multi-label read-write identification method and device based on multiple channels. The label read-write method comprises the steps that a plurality of initial communication channels are set in advance, radio frequency trigger signals are sent out and comprise the number of the communication channels and label read-write state information of all communication channels, received reflection carrier signals are decoded, at least one set of label feedback data in at least one communication channel are obtained, when label response information of one set of label feedback data meets a preset recognition condition, product electronic code information corresponding to the set of label feedback data is read, and label recognition information related to the product electronic code information is sent out. According to the method and device, accuracy of label identification and identification efficiency are improved.

A personal identification system, which uses a vein pattern of a finger, optimizes the amount of light of a light source based on a captured finger image and emphasizes the vein pattern during image processing for identification.

A method and apparatus for providing multi-domain control over a digital data item via a first domain security policy assigned to the digital data item at a first domain, the data item being transferred from the first domain to a second domain, the second domain being autonomous from the first domain in respect of security policies. The method comprises assigning the security policy to the digital item within the first domain; transferring the digital items to the second domain together with data defining the first domain security policy; analyzing the first domain security policy within the second domain; and distributing and / or allowing usage of the digital items within the second domain in accordance with analyzed first domain security policy, and / or reporting breaches or attempted breaches of the policy.

The invention discloses an infrared camouflage material, which is characterized by the following: the infrared hidden paint is compound of launching rate less than 0.63 green visible dyes, 0.76-0.78 yellow earth visible dye and more than 0.92 deep green visible dyes. The paint possesses mechanic property, which is compatible with visible camouflage property.

In a mobile wireless communication system automatic speech recognition is performed in a distributed manner using a mobile station based near or front end stage which extracts and vector quantizes recognition feature parameters from frames of an utterance and an infrastructure based back or far end stage which reverses the vector quantization to recover the feature parameters and subjects the feature parameters to a Hidden Markov Model (HMM) evaluation to obtain a recognition decision for the utterance. In order to conserve network capacity, the size (Sz) of the codebook used for the vector quantization, and the corresponding number of bits (B) per codebook index B, are adapted on a dialogue-by dialogue basis in relation to the vocabulary size |V| for the dialogue. The adaptation is performed at the front end, accomplishes a tradeoff between expected recognition rate RR and expected bitrate BR by optimizing a metric which is a function of both.

The invention relates to a target positioning method and device and an unmanned aerial vehicle. The method comprises the steps of acquiring a target image through an image acquisition device; acquiring position information of the target in the target image according to the target image; obtaining an original point cloud of an environment in front of the unmanned aerial vehicle through a depth sensor; obtaining a point cloud corresponding to the target in the original point cloud according to the position information of the original point cloud and the target in the target image; and acquiringposition information of the target in a three-dimensional space according to the point cloud corresponding to the target. According to the invention, the depth information of the target is obtained bydetermining the position information of the three-dimensional space of the target; a motion estimation model with higher stability and precision can also be provided for target tracking, the probability of false recognition and tracking loss is reduced, more accurate visualization of three-dimensional path planning and real-time three-dimensional reconstruction can be achieved, and meanwhile a three-dimensional map with a target object can be used for obstacle avoidance.

The embodiments of the invention provide a license plate identification method and system. The license plate identification method comprises the following steps: obtaining a license plate image, and determining a license plate character identification result of a currently identified license plate; when the obtained license plate image satisfies a preset condition, through the license plate character identification result of the currently identified license plate and the license plate character identification result of the last already identified license plate, determining whether the currently identified license plate and the last identified license plate are similar license plates; and if the two are similar license plates, obtaining a first running locus of a corresponding vehicle of the currently identified license plate and a second running locus of a corresponding vehicle of the last identified license plate, calculating the coincidence degree of the first running locus and the second running locus, if the coincidence degree of the first running locus and the second running locus is smaller than a certain threshold, determining that the currently identified license plate and the last identified license plate are license plates of different vehicles, and otherwise, determining that they are the license plates of the same vehicle.

The invention discloses an anomaly query recognition method based on a recognition model, a recognition device and a medium, which are applied to the technical field of artificial intelligence. The method comprises the following steps: when receiving a policy inquiry request sent by a target terminal, acquiring target operation data on the target terminal within a preset time range; Inputting thetarget operation data into a preset abnormal behavior recognition model to obtain a target analysis result of a query operation corresponding to the insurance policy query request; According to the corresponding relationship between the preset analysis result and the control strategy, the control strategy corresponding to the target analysis result is determined, and the target terminal is controlled according to the determined control strategy. The application is helpful for improving the efficiency of abnormal query identification of insurance policies and reducing the identification overhead and labor cost.

The invention discloses a ground guiding type unmanned aerial vehicle flying landing system and method based on an LED dot matrix. The system comprises an unmanned aerial vehicle and an unmanned aerial vehicle taking-off and landing station. The unmanned aerial vehicle taking-off and landing station comprises a carrying table, a three-axis pan-tilt is installed on the carrying table, and a camera is installed on the three-axis pan-tilt; a lens of the camera is controlled by the three-axis pan-tilt to be always kept upward and unchanged in direction, and the camera is connected with a ground computer; and the ground computer controls the camera to carry out photographing. The ground guiding type unmanned aerial vehicle flying landing system and method based on the LED dot matrix have the beneficial effect that an all-weather centimeter-level precise landing function of the multi-rotor unmanned aerial vehicle is achieved through the color and shape recognition algorithm of the computer vision technology and the multi-section PID control technology.

The invention discloses an image stitching and deduplication method based on video frame extraction. The method comprises the following steps: acquiring video data of a target area, and extracting frames from the video data to obtain a key frame image set of a group of video data; inputting the key frame image set into a preset stitching library to obtain a panoramic image corresponding to the keyframe set and a homography transformation matrix of each key frame image in the key frame set; inputting the key frame set into a preset target detection model to obtain a target detection result ofeach key frame image in the key frame image set; and performing coordinate transformation and deduplication on the target detection result of each key frame image in the key frame image set, and outputting the panoramic image and the target detection deduplication result of the panoramic image. According to the invention, the manpower consumption for shooting a plurality of scene photos can be avoided by automatically extracting the key frame, and meanwhile, the time and the manpower cost can be saved and the identification efficiency and the identification precision can be improved by identifying the key frame and automatically de-duplicating the identification result.

The invention relates to the technical field of target attribute identification, and provides a small target vehicle attribute identification method based on feature fusion. The method comprises: firstly, constructing a small target vehicle attribute recognition network based on feature fusion, comprising a feature pyramid network, a regional nomination network and a small-size target cascade network; inputting a traffic image to be detected into the feature pyramid network, generating a feature map containing low-level edge detail information, middle-level stacking fusion scale information and high-level semantic information, and stacking and fusing the feature map to obtain a multi-scale feature map; inputting the multi-scale feature map into a regional naming network to generate a candidate box containing a target; inputting the multi-scale feature map and the candidate box into a small-size target positioning network at the same time, outputting target coordinate information, and cutting a target according to the information; and finally, inputting the sheared target into a small-size target classification network, and identifying and outputting the target and the category thereof. According to the invention, the accuracy of small-size target attribute identification can be improved, and the false identification rate and the missing identification rate are reduced.

The invention, which relates to the pedestrian re-identification technology, provides a method for realizing a pedestrian re-identification system. The method comprises: step one, offline modeling iscarried out on a monitoring video; to be specific, g pedestrian detection and pedestrian feature extraction are carried out; and step two, online retrieving is carried out; to be specific, when onlineretrieving is carried out, two kinds of sorting is carried out; pedestrians in each video are sorted based on similarities with a retrieving target, and then a possibility value of containing a retrieving target is calculated for each retrieving video, and the videos are sorted. A method combining deep features with manual features is employed in practical application. Deep feature learning is carried out by using a network structure with fusion of two kinds of loss functions including a classification loss function and a ternary-set loss function; and because of the fusion of the two kinds of loss functions, tag information of a data set and similarity information of an image pair are utilized fully, so that effective distinguished pedestrian features are obtained under the condition ofthe small data set. With the improved ternary-set loss function, the training effect is optimized.

The invention discloses a mapping method for drawing a nuclear power plant terminating map. The mapping method comprises the steps of S1, importing preset terminating data into a mapping system, S2, selecting cables and apparatuses in the mapping system and then establishing a logical connection relationship between the apparatuses by virtue of connection of the cables, S3, after allowing the preset number of terminals for the apparatuses, automatically distributing the terminals of a control cabinet according to the preset terminating data and establishing a logical connection relationship between the terminals of the cables and the apparatuses and the terminals of the control cabinet, S4, completing the drawing of the minimum unit map of the nuclear power terminating map, and S5, automatically copying the minimum unit map and completing the drawing the the nuclear power terminating map. The invention also provides a mapping system for drawing the nuclear power plant terminating map, and the mapping system is used for implementing the method. The mapping method and the mapping system for drawing the nuclear power plant terminating map are quick in mapping and high in accuracy.

The invention discloses an intelligent application network, which is of a multi-stage structure. The network comprises at least one application server located at the uppermost stage and at least one intelligent terminal node located at the lowermost stage, and also comprises a plurality of routing nodes located at the intermediate stages, wherein the communication connections are sequentially established between the application server and the routing nodes, between the routing nodes, and between the routing nodes and the intelligent terminal node; the network delivers in the format of frame messages determined through the negotiation between nodes; data items of each frame message comprise a message identifier, an application target identifier and a node address; each intelligent terminal node corresponds to one anode address; and the delivery direction of the frame messages can be judged according to message identifiers in the frame messages. The intelligent application network has the advantages of simple structure, uniform format of communication messages, convenience for system expansion, and convenience for application in a complex large-scale network.

The invention discloses an intelligent tractor field obstacle avoidance control system and method, and belongs to the technical field of agricultural planting control systems. The intelligent tractorfield obstacle avoidance control system comprises a central processing module, an image collection module, a accelerator control valve extinguishing control assembly and an electric brake control assembly, wherein the central processing module is bidirectionally and electrically connected with the image collection module, the accelerator control valve extinguishing control assembly and the electric brake control assembly. According to the intelligent tractor field obstacle avoidance control system and method, a real-time camera device matched with a simple depth ranging algorithm can be used for replacing all expensive sensor devices, the production and design cost of the whole tractor obstacle avoidance system is greatly lowered, the purpose of intelligent recognition and avoidance of obstacles in front of the tractor with low cost and high precision is achieved, meanwhile early warning is achieved to a driver before tractor emergency braking avoids the obstacles, and thus abrasion ofthe tractor driving components caused by multiple times of emergency braking is lowered.

The invention discloses a model combination type speech recognition method based on GMM (Gaussian mixture model) noise estimation. According to the method, a GMM containing fewer Gaussian units is used for real-time estimation of noise parameters in noisy testing speech and monitoring change of noise. The noise parameters are estimated according to specific time intervals and are updated once at every time interval, and mute segments are processed as noisy speech. Except for use for model combination, the estimated noise parameters are stored in an internal storage to be used for making noise change judgment of next time interval. The noise monitoring includes firstly, reading the noise parameters of last time interval from the storage; then, combining the noise parameters with a clean speech GMM so as to obtain a noisy speech GMM, subjecting noisy testing speech of current time interval to probability calculation, comparing an output average log likelihood value with an average log likelihood value outputted by a noise parameter estimation submodule, considering that noise changes if the likelihood value is larger than a threshold value, and considering that noise is unchanged if not.

The disclosure relates to an image processing method and apparatus and an electronic device. The method may include the following steps: obtaining an initial gray value of each pixel point on an image to be processed; and according to gray value difference degrees among pixel points in a neighborhood corresponding to each pixel point, processing the initial gray value of each pixel point into an adjusted gray value, and obtaining adjusted images, wherein when the gray value difference degree corresponding to any pixel point is increasingly lower, the adjusted gray value of any pixel point increasingly approaches to an intermediate gray value of the corresponding adjusted image. Through the technical scheme of the disclosure, images can be preprocessed effectively to enable edges and texture details of objects to be shot to be improved, which is helpful in identifying objects and realizing other subsequent functions.

The embodiment of the invention provides a target detection and tracking method and device, a storage medium and an electronic device. The method comprises the following steps: collecting image information of a target monitoring area; performing first target detection on the target visible light image to obtain a first target detection result of the target visible light image; under the conditionthat the target infrared image meets the first condition, performing second target detection on the target infrared image to obtain a second target detection result of the target infrared image; according to the first target detection result and the second target detection result, a third target detection result of the target monitoring area is obtained, and the third target detection result comprises position information and category information of an object of the target monitoring area. The method achieves the effects of reducing the probability of false detection and missing detection andimproving the detection precision.

The invention discloses a wafer warping elimination method and a composite substrate. The method comprises steps that S1, a warped wafer (1) is selected; S2, a stress compensation film (2) internallyhaving the stress is deposited at a back surface of the wafer (1), and the internal stress of the stress compensation film (2) is made to offset the internal stress of the wafer (1) to acquire a flatcomposite substrate; and the wafer (1) is a silicon carbide based silicon carbide wafer or a silicon-based gallium nitride wafer; the composite substrate comprises the wafer (1) and the stress compensation film (2), wherein the stress compensation film (2) covers the back surface of the wafer (1), the internal stress of the stress compensation film (2) is made to offset the internal stress of thewafer (1), and the wafer (1) is the silicon carbide based silicon carbide wafer or the silicon-based gallium nitride wafer. The wafer warping elimination method is advantaged in that the stress compensation film internally having the stress is deposited at the back surface of the warped wafer, the internal stress of the stress compensation film (2) is made to offset the internal stress of the wafer (1), and the warping phenomenon is eliminated.

The invention discloses a security defense method and system for a cloud environment and a computer readable storage medium. The system comprises a first virtual security network element module, a first virtual interactive network element module and a first detection module; the first virtual security network element module comprises one or more virtual security network element groups, and each virtual security network element group comprises a main virtual security network element and a standby virtual security network element; the first virtual interactive network element module carries outflow forwarding between the cloud platform and a first defense service chain, and the first defense service chain is a security defense loop which is connected with each virtual security network element group according to a specified sequence; the first detection module detects whether each main and standby virtual security network element is available or not through a first detection service chain, and the first detection service chain is an available test loop for connecting the first detection module and each virtual security network element. According to the application, the main virtual security network element and the standby virtual security network element are arranged, and the node fault is detected by detecting the service chain, so that the high-availability serial defense capability and the automatic and low-delay fault identification and recovery capability are realized.

The invention discloses a light sensing circuit and a driving method thereof, a display panel and a display device, belonging to the technical field of display. The light sensing circuit comprises that: the gate of a first transistor is electrically connected to a first control terminal, the first pole is electrically connected to a first voltage terminal, and the second pole is electrically connected to a first node; the gate of a second transistor is electrically connected to the first node, the first pole is electrically connected to a second voltage terminal, and the second pole is electrically connected to the first pole of a third transistor; the gate of the third transistor is electrically connected to a second control terminal, and the second pole is electrically connected to a data line; the gate of a fourth transistor is electrically connected to a third control terminal, the first pole is coupled to the first node, and the second pole is electrically connected to a third voltage terminal; the first transistor and the fourth transistor comprise active layers formed of metal oxide, and the fourth transistor is multiplexed into a light sensing element; and the second transistor and the third transistor comprise active layers formed of polysilicon. Compared with the prior art, the invention is advantageous to reduce the process steps and thin the display panel.

The present invention provides an augmented neural network architecture and its training method, and a computer-readable storage medium. The augmented neural network architecture includes a first neural network model and a second neural network model. The first neural network model has been trained using samples. The model of the second neural network model is a model without sample training, and the input terminals of the first neural network model and the second neural network model are connected. The training method includes the following process: input new samples, the first neural network model and the second neural network model The neural network model outputs the first result and the second result respectively based on the new sample; the expected result minus the first result is used as the target value of the second neural network model; and the second neural network model is trained based on the target value. The present invention not only guarantees the ability to effectively identify both original samples and new samples, but also expands the range of samples that can be recognized by the augmented neural network architecture as a whole.

The embodiment of the invention discloses a vehicle passing record determination method and device, equipment and a storage medium. The method comprises the steps of obtaining a vehicle body image ofa target vehicle; determining feature information of the target vehicle according to the vehicle body image, the feature information comprising a CNN feature, the CNN feature being a feature of a target area of the target vehicle in a preset convolutional neural network model, and the preset convolutional neural network comprising at least ten convolutional layers and at least three pooling layers; and based on the feature information, searching a picture containing the target vehicle in a vehicle passing record image library to obtain a vehicle passing record of the target vehicle. Accordingto the technical scheme provided by the embodiment of the invention, the target vehicle can be accurately identified, the accuracy of the determined vehicle passing record is improved, and the probability of occurrence of the situation of false identification or missed identification of the vehicle is reduced.

The invention discloses a face exchange method, which comprises the following steps of: inputting a source image and a target image, and acquiring face feature points by using a detector; adjusting the shape of the 3D face through the obtained face feature points; performing face segmentation on the adjusted 3D face shape by adopting an FCN detection model after sample training, and segmenting faces in the source image and the target image from background and occlusion; the face of the target image is effectively covered with the face of the source image, and fusion is carried out; and outputting the picture after face exchange. According to the method, a standard segmentation network obtained by training on rich and diverse samples is used; the FCN realizes accurate and rapid segmentationof the face from the background and the occlusion, and efficiently processes face alignment, face segmentation, 3D shape estimation, expression estimation and the like without any constraint on the image, so that face exchange is more real, and the face change and the recognition rate are reduced.

The invention belongs to a geomagnetic field signal compensation technology, and specifically discloses an environment magnetic compensation method for a distributed magnetic anomaly detection system.The environment magnetic compensation method comprises the steps of performing preprocessing on magnetic field data, wherein the preprocessing comprises main magnetic field compensation, crust magnetic field compensation, linear trend compensation and low-pass filtering processing, then performing correlation analysis, calculating a transfer function between two groups of magnetic field data by using a cross-correlation spectrum and an autocorrelation spectrum, calculating the environment interference magnetic field intensity through Fourier transform, and performing effective compensation onthe obtained environment interference magnetic field intensity to acquire a submarine magnetic anomaly signal. According to the invention, real-time analysis, processing and compensation for environment magnetic interference signals are realized, influences imposed on submarine detection and recognition by the environment interference magnetic field are reduced, and the detection ability of the distributed magnetic detection system is improved.