351results about How to "Optimize detection results" patented technology

The invention relates to a cascaded neural network-based face key point detection method. The method includes the following steps that: a) a training-used face image set is established, and a key point position requiring detection is marked; b) a first-layer depth neural network is constructed and is used to train a face region estimation model; c) a second-layer depth neural network is constructed and is used to perform face key point preliminary detection; d) local region division is continued to be performed on an inner face region; e) a third-layer depth neural network is constructed for each local region respectively; f) the rotation angle of each local region is estimated; g) correction is performed according to the estimated rotation angles; h) a fourth-layer depth neural network is constructed for the correction data set of each local region; and i) any face image is given, and the above four-layer depth neural network model is adopted to perform key point detection, such that final face key point detection results can be obtained. With the cascaded neural network-based face key point detection method of the invention adopted, face key point detection can be improved, and especially the accuracy and real-time property of dense face key point detection.

An improved system and method for stirring suspended solids in a liquid media to enhance sample growth and improve sample detection results. The system and method employs a sample vessel holder which adapted to receive at least one sample vessel which contains the solids and liquid media and a stirrer, such as a ferrous metal filled stirrer, and maintain the sample vessel in a position such that the longitudinal axis of the sample vessel extends at an angle substantially less than 90 degrees with respect to the horizontal, such as within the range of about 15 degrees to about 25 degrees with respect to the horizontal. The system and method further employs a magnet driver, adapted to move a magnet, such as a rare earth magnet, proximate to an outer surface of the sample vessel to permit the magnet to impose a magnetic influence on the stirrer to move the stirrer in the sample vessel. Specifically, the magnet driver is adapted to move and, specifically, rotate the magnet such that the magnetic influence moves the stirrer along a side wall of the sample vessel. The magnet driver is further adapted to move the magnet away from said outer surface of the sample vessel to allow gravity to move the stirrer toward the bottom of the sample vessel. This technique therefore provides a more gentle and controlled stirring of the suspended solution.

The disclosed detector comprises: a driver, a magnetic-leakage probe, a magnetic-leakage signal process device with a rotary coder, a power, an ultrasonic probe, and a ultrasonic signal processor with a roller carried a milometer all connected by gimbals in turn. This invention integrates both magnetic-leakage and ultrasound, realizes precise positioning to defect in pipeline, and can detect the defect and hidden danger all round.

The invention discloses a traffic image multi-type vehicle detection method based on deep study. The traffic image multi-type vehicle detection method based on deep study firstly combines a nerve network characteristic with a region generation algorithm, realizes two processes-region generation and region determination-at the same time through using a nerve network conventional layer, uses a background model to determine a motion area of a discrete series image targeting a specific scene and provides an extra reference basis to region generation, and combines with a vehicle detection result to perform upgrading and correction on the background model according to different conditions. Besides, the traffic image multi-type vehicle detection method also brings forward a network model compression scheme to reduce model parameters and calculation time, brings forward a new detection result optimization means based on grouped error calculation to replace a conventional non-maximum-value inhibition scheme, and improves overall detection accuracy.

A solid-state detector for detection of neutron and alpha particles detector and methods for manufacturing and use thereof are described. The detector has an active region formed of a polycrystalline semiconductor compound comprising a particulate semiconductor material sensitive to neutron and alpha particles radiation imbedded in a binder. The particulate semiconductor material contains at least one element sensitive to neutron and alpha particles radiation, selected from a group including 10Boron, 6Lithium, 113Cadmium, 157Gadolinium and 199Mercury. The semiconductor compound is sandwiched between an electrode assembly configured to detect the neutron and alpha particles interacting with the bulk of the active region. The binder can be either an organic polymer binder or inorganic binder. The organic polymer binder comprises at least one polymer that can be selected from the group comprising polystyrene, polypropylene, Humiseal™ and Nylon-6. The inorganic binder can be selected from B2O3, PbO / B2O3 / , Bi2O3 / PbO, Borax glass, Bismuth Borate glass and Boron Oxide based glass.

The present invention encompasses a physical or virtual, computational, analysis, fusion and correlation system that can automatically, systematically and independently analyze collected sensor data (upstream) aboard or streaming from aerial vehicles and / or other fixed or mobile single or multi-sensor platforms. The resultant data is fused and presented locally, remotely or at ground stations in near real time, as it is collected from local and / or remote sensors. The invention improves detection and reduces false detections compared to existing systems using portable apparatus or cloud based computation and capabilities designed to reduce the role of the human operator in the review, fusion and analysis of cross modality sensor data collected from ISR (Intelligence, Surveillance and Reconnaissance) aerial vehicles or other fixed and mobile ISR platforms. The invention replaces human sensor data analysts with hardware and software providing two significant advantages over the current manual methods.

Provided is a computer malicious software detection novel method based on software control flow features. Static analysis is conducted on a control flow structure of a binary file, an operation code sequence is extracted automatically, a spatial vector model is utilized to convert the sequence to structuralized information, the structuralized information acts as a feature set of a file after being screened, a software classification rule is found from volume feature sets by using a data mining method, and the software classification rule is used for detecting malicious software. The computer malicious software detection novel method based on the software control flow features can not only effectively detect common zero-day malicious software, but also have a good detection result for malicious software utilizing a vague and polymorphic technology, and maintain good stability for calculated attack.

The invention discloses a remote sensing image cloud detection method based on a convolution nerve network, comprising steps of establishing a training sample set, generating a convolution nerve network classification model, extracting a super-resolution sub-area, generating a cloud probability graph, performing rough detection on a cloud area, and performing fine detection on the cloud area. Through the above arrangement, the remote sensing image cloud detection method can accurately detect the cloud area in the remote sensing image, and, for the remote sensing image having the complex background or the semitransparent cloud, can obtain a better detection result. As a result, the remote sensing image cloud detection method can solve problems of false determination and analysis which are caused by the cloud, and can facilitate the follow-up processing and analysis. The remote sensing image cloud detection method enables the cloud area in the cloud detection result to have a better rim, to have better robustness in a complex environment and can obtain a better detection result.

The present invention discloses an ultraweak fiber bragg grating array and Phi-OTDR combined optical fiber vibration sensing system. According to the system, the ultraweak fiber bragg grating arrays are accessed in a sensing optical fiber with the same spacing, and the spacing selection must be same with the spatial resolution of a Phi-OTDR optical fiber sensing system or must be integer multiples of the spatial resolution. By utilizing the reflection characteristics of the ultraweak fiber bragg gratings, the optical power of the detected backward rayleigh scattered light is improved, thereby improving the signal to noise ratio and the sensitivity of the Phi-OTDR optical fiber vibration sensing system.

Detection of a person disguised with one or more artificial materials includes detecting reflection from at least one portion of a head of a human body in at least a portion of an upper band of the near infrared spectrum. The presence of an artificial material associated with the head of the human body is determined based on the detected reflection.

The invention discloses a remote sensing image change detection method based on fusion and PCA kernel fuzzy clustering. The remote sensing image change detection method mainly solves the problems that in the prior art, the detection effect is not ideal, the accuracy of single-type difference image detection is low, and the application range is narrow. The method comprises the steps: (1) inputting two time phase remote sensing images X1 and X2 and conducting median filtering; (2) calculating a differential image, a logarithmic specific value image and a mean value ratio image of the two images after the filtering; (3) conducting fusion on the three images to obtain an image Xd after the fusion; (4) using a PCA method for conducting feature extraction on the images after the fusion, and obtaining a feature vector of each pixel to form a feature space matrix; (5) using a kernel-based fuzzy C mean value method for clustering the feature space matrix into two classes; (6) obtaining a final change detection result image according to the clustering result. The remote sensing change detection method has the better anti-noise performance and detection accuracy, the better effects of remote sensing images of different types can be obtained, and the remote sensing image change detection method can be applied to the field of environment monitoring and disaster evaluation.

The invention discloses a local outlier detection method based on density clustering. The method includes the steps: (a) acquiring data cluster number and clustering center of a to-be-detected dataset; (b) calculating the mean and standard deviation of description features of each data object in different data clusters; (c) detecting according to 3sigma criterion to obtain outliers of each data cluster. By adoption of the local outlier detection method based on density clustering, easiness in setting of parameters is realized, the method is applicable to datasets with different density regions and arbitrary shapes and can be used for detection of local outliers, accuracy of outlier detection results is high, algorithms are insensitive to selection of parameter values, and excellent robustness is achieved.

The invention provides a lung focus detection algorithm based on weak annotation data. The algorithm comprises the steps of data labeling, model training and image processing, wherein a partially labeled weak data labeling method is adopted for labeling, a labeling result is perfected by referring to the thought of difficult case mining in the data labeling process, and it is guaranteed that a high-quality labeled data set is provided for subsequent model training under the condition that only part of focuses are labeled. According to the method, model training is combined with a 'weak data 'marking mode, positive and negative sample anchor points for network training are acquired from positive and negative sample sets respectively, and the focal loss hyper-parameters are adjusted according to the source characteristics of the positive and negative sample anchor points, so that the detection model obtains a better detection effect.

A Collaborative Deep Network model method for pedestrian detection includes constructing a new collaborative multi-model learning framework to complete a classification process during pedestrian detection; and using an artificial neuron network to integrate judgment results of sub-classifiers in a collaborative model, and training the network by means of the method for machine learning, so that information fed back by sub-classifiers can be more effectively synthesized. A re-sampling method based on a K-means clustering algorithm can enhance the classification effect of each classifier in the collaborative model, and thus improves the overall classification effect. By building a collaborative deep network model, different types of training data sets obtained using a clustering algorithm are used for training a plurality of deep network models in parallel, and then classification results, on deep network models, of an original data set are integrated and comprehensively analyzed, which achieves more accurate sample classification.

The invention discloses a joint reasoning-based video multi-target tracking method in the technical field of video processing. The method comprises the following steps: firstly, reading a frame of image of a video file, carrying out image rasterization processing, and then calibrating a candidate position of a target by adopting an online detector and a KLT tracking algorithm as a tracker, respectively screening and then integrating the results; secondly, carrying out quantitative grading on the obtained candidate position result; finally, describing the target tracking condition by using a joint function, and taking the optimal solution based on the joint function as the position of the target in the frame, so as to achieve target tracking. By adopting the joint reasoning-based video multi-target tracking method, the processing of the treatment method combined with a detection tracking algorithm and multi-target relationship in the tracking technology under multi-target tracking can be achieved, and the multi-target relationship is described by using the joint function, so that not only is the result fusion of detection and tracking achieved, but also the relationship between the targets is integrated according to overall situation, so that a global optimal resolution is obtained.

The invention discloses a fabric surface defect detection machine set, and a detection method based on the fabric surface defect detection machine set. The fabric surface defect detection machine set comprises a machine frame, a feeding roller assembly, a foreign matter removing device, a flattening device, a deviation correcting device, a drawing roller assembly, a wrinkle removing device, an industrial camera, a marking machine, and a coiling machine. According to the detection method based on the fabric surface defect detection machine set, before detection, impurity removing, flattening, deviation correcting, and wrinkle removing are adopted so as to conveying flat fabric into a detection zone steadily; the fabric surface defect detection machine set can be used for on-line detection of fabric surface, increasing fabric surface detection efficiency and fabric detection result accuracy.

The invention relates to a driver fatigue detection method based on a long-short-term memory network. The method comprises the following steps: 1) acquiring a driver face video image by means of infrared acquisition equipment; 2) utilizing a multi-task cascaded convolutional neural network to carry out face detection and feature point positioning, and obtaining a driver eye image sequence according to a geometrical relationship among the feature points; and 3) designing an end-to-end convolutional recurrent neural network, extracting human eye spatial characteristics, analyzing context relationships between adjacent image frames, and judging whether the driver is in a fatigue state or not by combining time sequence changes of the human eye image characteristics within a period of time. Results show that the method can accurately extract eye features under the conditions of poor light conditions, sunglasses worn by a driver and the like, and compared with a fatigue detection method based on a CNN combined PERCLOS standard, higher fatigue detection accuracy is obtained, and prediction of the driving state video level of the driver is realized.

The invention provides a visual relation detection method, a visual relation detection device, a visual relation detection training method and a visual relation detection training device. The visual relation detection method includes the following steps that: a plurality of candidate regions for object detection are extracted from an image; a plurality of visual relation detection region sets areobtained on the basis of the plurality of candidate regions, wherein the visual relation detection region set contains a subject detection region, an object detection region, and a predicate detectionregion; and a subject item, an object item and a predicate relation are determined on the basis of the subject detection region, the object detection region, and the predicate detection region in each visual relation detection region set. With the methods and devices provided by the present invention adopted, a subject, a predicate and an object in a visual relation can be considered as a whole,the three parts are recognized simultaneously, and therefore, the accuracy of visual relation detection is improved.

The invention discloses an implementing method of FPGA (field programmable gate array) architecture of an HOG (histogram of oriented gradient) and SVM (support vector machine) based pedestrian detection system. The implementing method includes steps of inputting, gradient and direction computation, histogram generation, binaryzation, SVM classification and outputting. The invention further provides the FPGA (field programmable gate array) architecture of the HOG and SVM based pedestrian detection system. By the FPGA architecture and the implementing method thereof, the problem that an HOG and SVM based pedestrian detection algorithm is slow in computation on a PC(personal computer) is solved, transplant optimization of hardware is realized, the FPGA architecture of an embedded pedestrian detection system is implemented in the scheme with low power consumption, high detection efficiency and low resource consumption in real time, and thus, pedestrian detection is popularized and developed in the embedded field according to the scheme.

The invention relates to an automatic ship tracking method and system based on a deep learning network and mean shift. The method comprises the steps that monitoring video data including coastal areamonitoring video data under visible light are acquired, and each frame of image is extracted; preprocessing is carried out to extract the positive sample and negative sample of a ship target; throughan area-based convolutional neural network method, the samples of the ship target in a video are input into a neural network for model training; the initial frame data of the video are extracted, andship detection and probability density calculation are carried out on initial moment data according to a trained model; and the ship tracking result at the current moment is determined through the calculation result of the previous moment. According to the invention, the method has a great detection result for complex scenes such as clouds, cloudy days, rain and the like; the method has the advantages of high robustness and stability and fully automated tracking process; the stability and accuracy of the neural network method eliminate errors for a mean shift method; and a basis is laid for the tracking of an emerging target.

The invention discloses a target detection and motion state estimation method based on vision and laser radar, and the method comprises the following steps: 1, installing the positions of a camera andthe laser radar, carrying out the calibration between every two binocular cameras, and carrying out the combined calibration between a left camera in front of a vehicle and the laser radar; fusing and supplementing the sparse point clouds respectively generated by the stereoscopic vision system and the laser radar to form a global three-dimensional point cloud graph; 2, detecting the two-dimensional image and the three-dimensional point cloud by using a deep learning method, and fusing an image target three-dimensional motion state generated by stereoscopic vision with a laser radar three-dimensional point cloud target to obtain comprehensive target three-dimensional feature points; and 3, tracking the target by utilizing Kalman filtering according to the three-dimensional feature pointsof the target, and analyzing the motion state of the target. The key of the method lies in the fusion of vision and laser radar sensors, and improves the sensing capability of an intelligent automobile for the surrounding environment.

The device for detecting sample includes main body, detecting element housed inside the main body fixedly, and material of absorption and transfer. There are following parts setup on main body: gathering unit for collecting external sample, and docking area assorted; opening for application of sample in use for sample liquid to pass through, and result display window in use for observing tested result. Being located inside the opening for application of sample, the material of absorption and transfer keeps connectivity with detecting element so as to raise fluid velocity of detection liquid. Being processed by surfactant, the material of absorption and transfer promotes flow of detection liquid further, shortens detection time without disturbing detected result.

The invention discloses a method for detecting automatically a circular oil tank with a remote sensing image. The method comprises: first, executing MHC visual saliency transformation on the remote sensing image to obtain a visual saliency map, executing mathematical morphology enhancement to obtain a enhanced visual saliency map, and executing circle detection on the enhanced visual saliency map by means of hough transformation to obtain a suspected oil tank region; then, executing turbopixels over-segmentation on the remote sensing image, combining segmented blocks according to features, and obtaining a suspected oil tank region according to an approximately-circular feature; last, in conjunction with a hough detection result and an approximately-circular feature detection result, executing SVM classification by means of a relationship between a circle center and a radius of the oil tank and multiple features, and filtering out concentric circles and a non-oil tank region to obtain finally an oil tank region. Through a large number of experiments, it is proved that the method for detecting automatically a circular oil tank with a remote sensing image can obtain the higher precision ratio and recall ratio on the optical image having a large region and high resolution, and not only have a significant detection effect on the bright oil tank, but also have a certain detection effect on the darker oil tank.

The invention provides an expert knowledge constraint-based occlusion image ship target detection method and system, and the method comprises the steps: carrying out the preprocessing of sample data,and obtaining a preprocessed image; extracting candidate features by the RPN network, and generating a candidate region of the ship target; Expert knowledge extraction and feature dimension reduction:in order to solve hull occlusion, candidate regions generated by an RPN network are selected as a basis, and for each candidate region, expert knowledge features of an image block in the vertical direction are extracted in a sliding window form; CNN feature extraction: combining candidate regions generated by a region-based convolutional neural network and an RPN network to obtain feature maps ofthe candidate regions; Constructing a feature fusion network, and mapping the CNN features and expert knowledge to a unified feature space; Using the sample data to train a feature fusion network, and using the trained feature fusion network to realize shielding ship target detection. By applying the method, the ship detection result under the shielding condition can be optimized.

There is provided a method for establishing a parameters setup for inspecting a plurality of articles by an automatic inspection system. The method includes inspecting a first article by the inspection system, applying an automatic defects detection method according to a given set of inspection parameters, receiving an initial map of defects and sorting uncovered defects into defect types according to a predetermined set of defect types. While sorting defects, if new defects not recognized by the inspection system are detected, adding the new defects to the initial map to be sorted and automatically setting the inspection parameters by means of applying computational dedicated algorithms, using a heuristic approach, to form a modified parameters setup. The modified parameters setup is then used for obtaining a modified map of detected defects, and the modified parameters setup for inspecting other of the plurality of articles. A system for establishing a parameters setup for inspecting a plurality of articles is also provided.

The invention relates to a method for detecting repeated software defect reports. The method comprises the following steps of firstly, extracting a training sample set and a test sample set from a software defect report database, establishing a subject model of the training sample set, then applying the subject model to test samples to obtain a document-subject matrix, calculating the subject similarity between two test samples, extracting classification information of the test samples to calculate the classification information similarity, and multiplying the classification information similarity and the subject similarity to obtain LDA (local data area) similarity between the two test samples; secondly, extracting an N-gram sequence of the test samples to calculate N-gram similarity, performing weighted summation on the N-gram similarity and the LDA similarity to calculate the whole similarity between the two test samples; and finally, if the whole similarity is greater than or equal to a preset threshold value, indicating that the two test samples are the repeated defect reports. According to the method, the accuracy of a detection result is greatly improved; the repeated defect reports can be prevented from being dispatched to a developer as much as possible, and human resources are saved.