4031 results about "Detection performance" patented technology

Presented are systems and methods that provide a unified end-to-end detection pipeline for object detection that achieves impressive performance in detecting very small and highly overlapped objects in face and car images. Various embodiments of the present disclosure provide for an accurate and efficient one-stage FCN-based object detector that may be optimized end-to-end during training. Certain embodiments train the object detector on a single scale using jitter-augmentation integrated landmark localization information through joint multi-task learning to improve the performance and accuracy of end-to-end object detection. Various embodiments apply hard negative mining techniques during training to bootstrap detection performance. The presented are systems and methods are highly suitable for situations where region proposal generation methods may fail, and they outperform many existing sliding window fashion FCN detection frameworks when detecting objects at small scales and under heavy occlusion conditions.

Face detection techniques are provided that use a multiple-stage face detection algorithm. An exemplary three-stage algorithm includes a first stage that applies linear-filtering to enhance detection performance by removing many non-face-like portions within an image, a second stage that uses a boosting chain that is adopted to combine boosting classifiers within a hierarchy “chain” structure, and a third stage that performs post-filtering using image pre-processing, SVM-filtering and color-filtering to refine the final face detection prediction. In certain further implementations, the face detection techniques include a two-level hierarchy in-plane pose estimator to provide a rapid multi-view face detector that further improves the accuracy and robustness of face detection.

Methods for using statistical analysis to monitor performance of new network infrastructure and applications for deployment thereof. A method monitors a release of executing software applications or execution infrastructure to detect deviations in performance. A first set of time-series data is acquired from executing software applications and execution infrastructure. A first statistical description of expected behavior is derived from the first set of acquired data. A second set of time-series data is acquired from the monitored release of executing software applications and execution infrastructure. A second statistical description of behavior is derived from the second set of acquired data. The first and second statistical descriptions are compared to identify instances where the first and second statistical descriptions deviate sufficiently to indicate a statistically significant probability that an operating anomaly exists within the monitored release of executing software applications and execution infrastructure.

The invention discloses a transmission method of downlink control signaling, comprising the followings: a base configures a downlink transmission mode based on two new additional antenna ports M and N, and configures the corresponding relation between the downlink transmission mode and a downlink control information format (DCI format); and sends downlink control signaling to user equipment (UE) according to the configured corresponding relation. The invention further discloses a transmission device of downlink control signaling, thereby guaranteeing the detection performance and backward compatibility of physical downlink control channel (PDCCH) by adding or amending the existing seven types of downlink transmission modes and determining the corresponding relation between the corresponding DCI format and downlink transmission mode.

Adaptive vision-based vehicle detection methods, taking into account the lighting context of the images are disclosed. The methods categorize the scenes according to their lighting conditions and switch between specialized classifiers for different scene contexts. Four categories of lighting conditions have been identified using a clustering algorithm in the space of image histograms: Daylight, Low Light, Night, and Saturation. Trained classifiers are used for both Daylight and Low Light categories, and a tail-light detector is used for the Night category. Improved detection performance by using the provided context-adaptive methods is demonstrated. A night time detector is also disclosed.

A small and light automotive radar having a high detection performance by preventing the road clutter and its in-vehicle positioning is optional is provided. The automotive radar comprises an antenna 1, 2a, 2b equipped with at least one radiating element which radiates linear polarized radio waves; a slit plate 7 which is a metal plate in which a plurality of slits are defined, placed in front of the surface of the antenna; radio wave absorbers 5 provided between the antenna and the slit plate; and a transceiver device which supplies transmit signals to the antenna to radiate radio waves and, from signals acquired by receiving reflection waves which are returned waves of the radio waves reflected by an obstruction, detects a direction in which the obstruction exists.

The invention relates to an aluminum material surface defect detection algorithm based on deep learning, and the algorithm comprises the steps: (1) employing a camera to shoot the surface of an aluminum material, obtaining a related data set, employing a labelImg tool to label an image, and obtaining label information; (2) dividing the image into a training set and a test set, and performing dataenhancement on the training set; (3) inputting a defective image, a non-defective image and label information of the defective image into the network at the same time every time to carry out model training; and (4) inputting the test image into the trained aluminum material surface defect detection model, and obtaining the position and the corresponding category of the defect. According to the method, a defective image and a non-defective image can be effectively utilized; the generalization ability and the detection precision of the model are improved, the detection performance is further improved by fully utilizing context information around the candidate region, the detection performance of dense small defects can be improved by utilizing a soft non-maximum suppression algorithm, and the method is an efficient aluminum material surface defect detection algorithm.

Performance of an intrusion detection system is enhanced with the addition of rule optimization, set-based rule inspection, and protocol flow analysis. During rule optimization, rule sets are created and selected in such a way that for every incoming packet only a single rule set has to be searched. Set-based rule inspection divides rules into content and non-content type rules. Only search patterns of content type rules are initially compared to a packet. Rules containing matched search patterns are validated with a parameterized search against the packet. Matches are recorded as events. Non-content rules are searched against a packet using a parameterized search. These matches are also recorded as an event. At least one event is selected per packet for logging. Protocol flow analysis determines the direction of flow of network traffic. Based on the direction of flow and the protocol, portions of packets can be eliminated from rule inspection.

Providing: quickly brining a vapor cell 119 to a desired temperature when retaining the heat of the vapor cell 119 to enhance the magnetic field detection performance of an optically pumped magnetometer; preventing adherence of atoms in the vapor cell 119 to a laser irradiation light passing-through part of the vapor cell 119; downsizing the periphery of the vapor cell 119; and suppressing the effect of a magnetic field from a heater used to retain the heat of the vapor cell 119. The present invention includes: a transparent film heater 118 provided to a laser irradiation light passing-through part of a vapor cell 119, the vapor cell 119 being a magnetic detection part of the optically pumped magnetometer; a temperature detector 115 provided at a center part of a side of the vapor cell 119; a temperature regulator 111 that sets a desired temperature for heat retention of the vapor cell 119 and compares the desired temperature and the actual temperature of the vapor cell measured by the temperature detector 115; an operation unit 112 that upon receipt of a PID control signal for temperature control from the temperature regulator 111, performs a temperature adjustment and switches on/off, in a pulsed manner, current applied to the transparent film heater 118 after the desired temperature is reached; and a heater power supply 113 that upon receipt of an operation signal from the operation unit 112, applies current to the transparent film heater 118.

System and method for improving the detection performance of a wirelessly transmitted signal. A preferred embodiment comprises specifying a desired response for missed channel detection and false alarm probabilities for a plurality of signal qualities and determining if a channel detection threshold is based on missed channel detection or false alarm probabilities for the plurality of signal qualities. A signal quality estimate of a channel can be inferred from a signal quality measurement of a second channel, wherein the channel and the second channel are sourced by a single transmitter. The channel detection threshold can be adjusted based upon the previously determined required response and the signal quality estimate of the channel.

The invention discloses a CFAR (constant false alarm rate) detection and depth learning-based SAR (synthetic aperture radar) target detection method. The method includes the following steps that: (1) SAR images are obtained; (2) a training sample set is expanded; (3) the network structure of a Faster-RCNN model is constructed; (4) a trained RPN model is obtained; (5) a trained Fast-RCNN model is obtained; (6) a finely-adjusted RPN network is obtained; (7) a trained Faster-RCNN model is obtained; and (8) target detection is carried out. With the CFAR (constant false alarm rate) detection and depth learning-based SAR (synthetic aperture radar) target detection method of the invention adopted, end-to-end image-level detection is realized, and major problems in existing SAR target detection technologies can be solved. The method has good detection performance under complex conditions.

The invention belongs to the technical field of radar remote sensing, and relates to an attention pyramid network-based SAR image multi-scale ship detection method. On the basis of an existing featurepyramid network, the invention provides a multi-scale feature extraction method with adaptive selection of significant features, namely the feature pyramid network based on a dense attention mechanism, and the method is applied to SAR image multi-scale ship target detection. Prominent features are highlighted from global and local ranges through the channel attention model and the spatial attention model respectively, and better detection performance is obtained; meanwhile, the attention mechanism is applied to the multi-scale fusion process of each layer, the characteristics can be enhancedlayer by layer, false alarm targets can be effectively eliminated, and the detection precision is improved.

The invention relates to a method for improving target detection performance by improving target classification and positioning accuracy. The method is characterized by extracting image characteristics according to a convolution neural network framework, selecting the first M layers of convolution layers to be output and carrying out characteristic fusion so as to form a multi-characteristic characteristic graph; carrying out grid division on the convolution layer M, and predicting target candidate frames with the fixed number and size in each grid; mapping the candidate frames to the characteristic graph for cutting, and carrying out multi-characteristic connection on cutting results; and passing the above results through a whole connection layer, classifying the image characteristics through the Softmax classification algorithm and using the overlapping area loss function to carry out online iteration regression positioning so as to obtain the final target detection result. According to the invention, the method is properly designed; characteristics are extracted through the convolution neural network, and multilayer fusion is performed on the image characteristics; and finally, the Softmax classification algorithm is used for classifying the image characteristics and the overlapping area loss function is used to carry out positioning, so a good target detection result is acquired.

The invention discloses a natural scene text detection method based on full convolution neural network, This method uses CNN network to extract the feature representation of text, and adjusts the feature representation by feature fusion module, at the same time fuses the semantic feature of high-level feature map and the position information of low-level feature map, so that the extracted featurehas stronger representation ability, and combines with text prediction module to directly predict the candidate text object. This method adopts end-to-end training and prediction process, The processing flow is simple, and no multi-step hierarchical processing is needed. Finally, the final detection result is obtained by simple NMS operation, which has high accuracy and strong robustness. The multi-directional and multi-size text object in the natural scene image with complex background can be detected well, and the detection performance of the natural scene text is excellent.

The invention discloses a list successive cancellation decoding method for polar codes, belonging to the technical field of wireless channel coding. In searching of a code tree, a candidate path list is updated as follows: each candidate path in a previous layer of candidate path list is expanded, and a reliable measurement value of each expanded path is computed; then, expanded paths, the reliable measurement values of which are less than the reliable measurement threshold of the current layer, are eliminated, it is determined whether the quantity of left expanded paths is greater than a preset maximum list length lmax, if yes, expanded paths among the left expanded paths, the reliable measurement values of which are the minimum, are eliminated, and the quantity of the last remaining expanded paths is not greater than lmax; and finally, the paths are selected adaptively. Loose ranking of the measurement values of the paths is combined with existing self-adaption polar decoder, so that the apparatus is guaranteed to have the same detection performance as conventional decoders, and decoding complexity is greatly lowered.

The invention provides an image object detection method. The image object detection method is capable of achieving good detection performance in a large image data set. The method includes the steps that multiple sample images are marked according to the priorities of information amounts, so that corresponding marked images are obtained; regions containing objects or regions highest in probability of object existence are extracted from the marked images, and candidate windows are generated; feature expressions of the candidate windows are extracted from a convolution neural network and form a candidate set, the candidate set is fitted through semi-supervised learning, and therefore a target function of an image detection model is obtained; the region containing a target object or the region highest in probability of target object existence is extracted from an image to be detected, a window to be detected is generated, the feature expression of the window to be detected is extracted and detected, and the candidate window highest in the probability of target object existence is obtained. The method can rapidly and accurately detect the target object in the large data set.

The repetitive structure of a preamble signal is exploited to enhance timing synchronization performance and frame start detection performance under adverse channel conditions. Received values are cross-correlated in time against a known noise-free version of the preamble. The presence of peaks in the cross-correlation output indicates presence of a frame. The peak locations provide symbol timing. Further cross-correlation processing and/or non-linear processing can be used to enhance the signal to noise ratio of the peaks.

The invention discloses a target detection method based on Keystone and matched filtering and belongs to the radar weak signal detection technology field. The method comprises steps that firstly, a radar is utilized to emit a chirp signal, pulse compression for a received target echo signal is carried out, and first-order range migration caused by no fuzzy speed of a target is corrected through utilizing Keystone transformation; secondly, a fuzzy speed multiple and acceleration of the target are estimated through utilizing combined two-dimensional matched filtering processing of a distance frequency domain, and first-order range migration caused by the fuzzy speed of the target and second-order range bending and Doppler migration caused by acceleration are corrected and compensated; and lastly, phase-coherent accumulation of target energy is realized through fast Fourier transform, and a peak value of phase-coherent accumulation is utilized to carry out target detection. The method is advantaged in that amplitude and the phase information of target echoes are utilized to carry out long time phase-coherent accumulation, so a radar echo signal-to-noise ratio can be effectively improved, and target detection performance of the radar can be improved.