1993 results about "Small target" patented technology

A pattern from a patterning device is applied to a substrate. The applied pattern includes device functional areas and metrology target areas. Each metrology target area comprises a plurality of individual grating portions, which are used for diffraction based overlay measurements or other diffraction based measurements. The gratings are of the small target type, which is small than an illumination spot used in the metrology. Each grating has an aspect ratio substantially greater than 1, meaning that a length in a direction perpendicular to the grating lines which is substantially greater than a width of the grating. Total target area can be reduced without loss of performance in the diffraction based metrology. A composite target can comprise a plurality of individual grating portions of different overlay biases. Using integer aspect ratios such as 2:1 or 4:1, grating portions of different directions can be packed efficiently into rectangular composite target areas.

The invention discloses a small target detection method based on feature fusion and depth learning, which solves the problems of poor detection accuracy and real-time performance for small targets. The implementation scheme is as follows: extracting high-resolution feature map through deeper and better network model of ResNet 101; extracting Five successively reduced low resolution feature maps from the auxiliary convolution layer to expand the scale of feature maps. Obtaining The multi-scale feature map by the feature pyramid network. In the structure of feature pyramid network, adopting deconvolution to fuse the feature map information of high-level semantic layer and the feature map information of shallow layer; performing Target prediction using feature maps with different scales and fusion characteristics; adopting A non-maximum value to suppress the scores of multiple predicted borders and categories, so as to obtain the border position and category information of the final target. The invention has the advantages of ensuring high precision of small target detection under the requirement of ensuring real-time detection, can quickly and accurately detect small targets in images, and can be used for real-time detection of targets in aerial photographs of unmanned aerial vehicles.

An apparatus and method for the insertion of a medical tool, for example a needle, within the human body. The apparatus and method are particularly useful in prostate brachytherapy and in prostate biopsy. The apparatus comprises a telescoping guide universally coupled to a first and second positioning means that are used to automatically and / or manually position the guide at a desired needle insertion trajectory. Automatic positioning of the guide is accomplished with reference to three-dimensional transrectal ultrasound images that can also be used to show needle insertion in real-time. The apparatus may be manually positioned in the approximate insertion trajectory and then a computer interconnected with the apparatus may be used to achieve the final trajectory based upon the ultrasound images. The apparatus is particularly useful in cases where multiple needles are to be inserted into a small target area and in cases where pubic arch interference prevents direct access to the target area.

An apparatus and method for electrostatic spray deposition of small targets, such as medical devices like stents. The apparatus includes a target holder which applies a first electrical potential to the target, and an electrostatic dispensing nozzle which applies a second potential sufficient to attract the coating fluid from the nozzle toward the target. Because the entire dispensing nozzle is conductive, the coating fluid may receive a greater charge than may be obtained with internal electrode-type nozzles. Electrostatic attraction of the coating fluid to the target is enhanced by the combination of higher charge density imparted to the coating fluid by the conductive nozzle, and application of a momentary voltage spike to the target to provide consistent conductivity between the target and its holder, thereby ensuring the target is presents the full first potential applied to the holder. The voltage spike may also be used independently of the conductive nozzle.

Pseudo-random numbers (PRNs) generated by a PRN generator are mapped to predefined number ranges or target ranges. The target range may be smaller or larger than the range of the PRN generator. Mapping to a smaller target range may include generating PRNs (e.g., integers) from a particular bit-input stream (e.g., 32-bit) having a uniform distribution across the range of numbers; selecting an optimal subset of the generated PRNs to map; and mapping the selected PRNs to a corresponding number in a target range such that the mapped numbers are uniformly distributed across the target range. Mapping to a larger target range may include generating uniformly distributed PRNs; applying a generation function to the PRNs to generate uniformly distributed packed numbers; and applying a mapping function to map selected packed numbers to the target range such that the mapped numbers are uniformly distributed.

The invention provides a multi-mode and multi-target precise tracking device and the method, wherein, a digital servo-platform is used as the support platform; a CCD video camera and an infrared sensor are arranged on the digital servo-platform for receiving image information; the received image information is processed through a comprehensive information process platform for obtaining the tracking information of the target; the tracking information is compressed and transmitted to a control center through a transmission equipment for monitoring. The means that acquiring the target tracking information from the image information is the core of detecting and tracking weak and small targets in complicated background. The method divides and selects the targets through target adaptive threshold based on the binomial distribution judgment rule after pre-processed the image under low signal-to-noise ratio, and then improves the detecting probability of the target and reduces the false alarm probability through data fusion of infrared sensor and visible light sensor, and finally detects and estimates the movement of the selected target to acquire the tracking information of the target. When the target shape is changed, the feature invariant is searched through shape identification of edge feature normalization to realize precise tracking to the target.

The invention provides a method for tracking a small target with high precision under a complex background and a low signal-to-noise ratio (SNR). The method comprises pre-treating an image under the complex background and low SNR, portioning and extracting the target based on a target adaptive threshold of the binomial distribution judgment rule, adopting a curve fitting algorithm based on Kalmanfilter thinking improvement to carry out motion prediction on the target, using data fusion of infrared and visible light transducers to improve detection probability of the target and reduce false-alarm probability, and when the shape of the target alters, using a shape of which edge characteristics are normalized to identify and seek characteristic invariables so as to achieve precise tracking of the target.

The invention discloses a design method of a target detection network based on fused image segmentation features, which is effective for large-scale targets. Based on the general target detection framework Mask RCNN and image segmentation feature fusion, the target segmentation feature and a ResNet-101 convolutional network are integrated into the rpn module, an RoI Pooling module and an RoI Alignmodule, the experiments show that the method is effective for large targets, and the image segmentation algorithm for small targets can be improved completely if the image segmentation effect is ideal.

The invention discloses a small target rapid detection method based on a deep convolution neural network. The deep convolution neural network is improved by the following steps: selecting the sliding windows on the convolution feature map of the last shared convolution layer of a VGG16 network as candidate boxes, wherein the sliding windows adopted are half-pixel precision sliding window; deleting a fifth pooling layer, and retaining other convolution layers and pooling layers; adding a convolution layer with a 3*3 convolution kernel; and using two convolution layers with 1*1 convolution kernels to replace all full-connection layers in the network to get the network adopted in the invention, training the network using collected data to get a small target classification model, and using the model to detect small targets. By using the method, the computational complexity is reduced, and the detection rate of small targets is improved.

An apparatus comprising an acoustically resonant ultrasound launcher is provided for cleaning of optical fiber connectors and other optical fiber parts. The resonant ultrasound launcher is designed to transfer and to focus ultrasonic energy from an ultrasound transducer to a relatively small target area with high intensity. Specially designed fluid flow channels allow this apparatus to efficiently clean optical fiber connectors with an exposed end surface or with an end surface concealed in a connector adapter. Circuitry is provided to automatically track the frequency to enhance the ultrasound generation. Another circuitry is provided to program the sequence of the cleaning process, including washing, rinsing and drying.

The invention discloses a dual-channel convolution neural network semantic segmentation method sensitive to small targets. The method comprises the following steps: a Caffe depth learning frame is used to build a non-weighted learning network and a weighted learning network; for the two-channel network, the corresponding semantic segmentation model is obtained by two-stage training. The output scoring charts of two channels are obtained by two semantic segmentation models, and the output scoring charts of two channels are fused by different model fusion algorithms, and the optimal model fusionalgorithm is selected according to the specific evaluation index. The test image is segmented according to the semantic segmentation model and the selected optimal model fusion algorithm. The invention can ensure that on the premise that the overall segmentation accuracy of the data set is better, the invention is more sensitive to the small target area existing in the image.

The invention provides a method for constructing a dense small target detection model, the model and the detection method. Based on the information fusion of the target midpoint context, through cutting the picture with high resolution, the method avoids the picture sampled under the input network from losing too much image information, and affects the network feature extraction. The residual pyramid feature extraction network is used to fuse the features of different scales, which improves the detection accuracy of the network for different size targets, especially for small targets. A RoIAlign layer instead of RoIPooling layer is used to solve the position deviation of candidate frames caused by feature mismatch of candidate regions. Because the small target features are easily lost in the network transmission, the center point context features are fused with the original RoI features, so that the network can make full use of the target context information, ensure the network runningspeed, more accurately locate and identify the dense small target, and improve the network performance.

The invention discloses a small target semantic segmentation method combined with target detection. Attention semantic segmentation network, training the network to get the whole semantic segmentationmodel; making small target detection dataset and small target semantic segmentation dataset; training the small target detection network based on YOLOv2 through the small target detection data set; asmall target semantic segmentation network is designed and trained by using the small target semantic segmentation data set to obtain the small target semantic segmentation model. In the testing phase, the test image is used as the input of the whole semantic segmentation model and the small target detection network, and the segmentation result and the small target boundary box of the whole imageare obtained, which is modified by the small target semantic segmentation model. The invention can greatly reduce the segmentation difficulty of the small target, thereby effectively improving the segmentation performance of the small target.

The invention relates to a countermeasure system of a small unmanned aerial vehicle. The countermeasure system comprises a low-altitude small target monitoring radar, a photoelectric tracker, a high-energy microwave orientation jammer, a display control bench and a controller. The low-altitude small target monitoring radar is used for searching and finding a small unmanned aerial vehicle in a protection region and finding a target. The photoelectric tracker is used for identifying, tracking, monitoring, aiming at and locking a threatening target under the guidance of the low-altitude small target monitoring radar. The high-energy microwave orientation jammer is used for carrying out suppression and interference on the threatening target tracked and locked by the photoelectric tracker, and damaging a measuring, controlling and navigation system of the threatening target. The display control bench and the controller are used for carrying out comprehensive and intelligent control on the confrontation system, and displaying orientation, tracks and GIS information of the threatening target. The countermeasure system is used for a system defense method for searching, monitoring, tracking and disturbing a small unmanned aerial vehicle flying above a safety protection place, and preventing terrorists from endangering public safety by use of the air vehicle.

The invention discloses a small target detecting method based on R-FCN, wherein the method relates to the field of image processing. The method comprises the steps of introducing a to-be-detected image into a convolutional network, successively performing characteristic extraction on a to-be-detected image through M network layers according to a sequence from a topmost layer of M network layers to a downmost layer and according to a sequence from the downmost layer of the M network layers to the topmost layer, generating characteristic mapping graphs with different scales, selecting an N characteristic mapping graphs into an RPN for performing foreground-and-background classification, determining the coordinate of a foreground area, processing a characteristic mapping block which corresponds with the coordinate of the foreground area for obtaining a characteristic vector; inputting each characteristic vector into a classifier for performing secondary classification, detecting whether the kind to which the characteristic vector is affiliated corresponds with a to-be-detected small target and outputting a detecting result. According to the small target detecting method, a manner of combining a top-down characteristic pyramid and a down-top characteristic pyramid is utilized for performing small target detection on the characteristic mapping graphs with different scales, thereby reducing report omission for the small target and improving detecting precision.

The invention discloses a vehicle-mounted video target detection method based on deep learning, and the method comprises the steps: employing an improved Faster R-CNN algorithm to realize target detection in a complex traffic environment, to provide a driving safety auxiliary function. An existing target tracking algorithm has a serious problem of small target missed detection. The depth information channel is added, the depth information channel is connected with an original color image channel in parallel, fusion is carried out in the channel dimension, candidate frame extraction and targetdetection are carried out on the fused feature image, the detection rate of a small target is improved, in addition, training of a difficult sample is added in training, and the overall target recognition rate of the algorithm is improved. According to the method, a problem of small target missed detection by using the Faster R-CNN algorithm is fully considered. The accuracy of vehicle recognitionin a complex traffic scene is improved through depth image feature fusion and a difficult sample mining method.

The invention discloses a detection and tracking method of a moving small target in an aerial shot video. The method includes the steps of 1, collecting images, 2, extracting SURF feature points, 3, carrying out grouped matching on the images, 4, obtaining an affine matrix, 5, obtaining a difference image, 6, carrying out opening operation, 7, extracting a target area, 8, determining a target template, 9, determining a target detection area, 10, extracting and matching the feature points, 11, determining a registering center position of a target, 12, determining a target central position, and 13, determining the length and the width of the target. The method has good real-time performance and robustness in the target tracking process, and can obtain smooth target moving tracks in the target tracking process.

The invention relates to a video monitoring system and method for target detection and tracking. The system comprises a video collection device, a target detection device, an information processing device, an information transmission device and an image collection device, wherein the image collection device is composed of a front end image collection device and a video decoding device and is usedfor obtaining a real-time video stream of a target object; the target detection device analyzes the position and the size of the target object in a video image; the information processing device performs real-time processing on related information of the target object within a continuous time to obtain a moving speed, a trajectory and direction information of the target object, and judges the number, the location and the size information of the target object in advance; the information transmission device sends the related information of the target object to the image collection device; and the image collection device controls the target object to always locate in a middle area of the screen and tracks the target object in real time. According to the video monitoring system and method, thetracking of numerous small targets is facilitated according to the deep neural network with multiple features, the recognition rate is improved, and it is beneficial to the good running of the tracking algorithm, and real-time tracking of the target object is achieved.

The invention relates to an unmanned aerial vehicle (UAV) small target detection method based on motion features and deep learning features, which belongs to the technical field of image processing and computer vision. The method includes the following steps: processing an input video data set through a video image stabilization algorithm to make compensation for the motion of a camera; analyzingdetected moving candidate target regions in images; dividing the video data set into two parts, and carrying out training by using a training data set to get an improved candidate region generation network model; generating a candidate target for the video images of a test set through a candidate region generation network based on depth features obtained from training; fusing the candidate targetregions; carrying out training by using the training data set to get a deep neural network model based on dual channels, and obtaining an identification result by using the model; and applying a target tracking method based on multilayer depth features to the identification result in the previous step to get the final position of a UAV. A UAV in a video image can be accurately detected, and thus,support can be provided for the subsequent research in fields related to UAV intelligent monitoring.

The invention provides a target detection method based on DenseNet and multi-scale feature fusion. The target detection method comprises the following steps: S1, constructing a feature extraction network model; S2, training the feature extraction network model, and obtaining an optimal target detection model through multiple times of iterative training; S3, inputting the to-be-detected image datainto the optimal target detection model for detection, and marking the position and the category of each object on the to-be-detected image by using a rectangular frame. According to the feature extraction network model, a DenseNet network serves as a basic network, the network hierarchy is deepened, the feature quality is improved, meanwhile, a feature fusion module is used, context information is introduced, six feature maps used for final prediction are obtained, and the feature extraction network model has rich semantic information and high resolution. According to the method, the model scale can be reduced on the basis of ensuring the detection speed, and the detection precision of a small target is improved.

The invention discloses a convolutional neural network-based target detection method and system. The method comprises the steps of constructing a data set of a detected target; dividing the image datain the data set of the detected target into a training set, a test set and a verification set according to a preset proportion, and marking images in the training set, the test set and the verification set; constructing a network structure of a convolutional neural network model, wherein the convolutional neural network model adopts different feature scales to predict an object; loading a training set into the convolutional neural network model for training; in the training process, a verification set is loaded, and parameters of the convolutional neural network model are optimized through amulti-verification method; carrying out performance test on the convolutional neural network model through the test set, and detecting the generalization capability of the convolutional neural networkmodel; and carrying out target recognition is carried out by adopting a convolutional neural network model with the generalization capability meeting requirements. The convolutional neural network model obtained through training can quickly and accurately identify small targets, compact and dense targets or highly overlapped targets in a shopping mall.

The invention discloses a hyperspectral remote sensing image small target detection method based on spectrum saliency and belongs to the field of hyperspectral remote sensing images. When the method is used for target detection, local saliency is calculated with an improved Itti model by means of spectrum information and spatial information extracted from a hyperspectral image, and a local saliency map is constructed; then global saliency is calculated with an improved evolutionary programming method, and a global saliency map is constructed; finally, the local saliency map and the global saliency map are combined in a normalized mode to obtain an overall vision saliency map which is taken as the final target detection result. According to the method, a saliency model suitable for the hyperspectral image is established according to the spectrum saliency, image interested target detection is achieved based on comprehensive analysis of the spectral signature and spatial signature of the hyperspectral image, main contents of the image are highlighted, and image processing and analyzing complexity is reduced.

The invention discloses a traffic identifier detection method based on multi-scale circulation attention network. The method comprises the following steps: firstly, building a traffic identifier detection model, wherein the traffic identifier detection model is formed by compounding a convolutional neural network model feature extraction model for carrying out image feature extraction and a multi-scale circulation attention network model for improving small-target detection accuracy; then training the traffic identifier detection model by utilizing a reasonable training sample so as to acquirea trained traffic identifier detection model; and inputting to-be-detected images into the trained traffic identifier detection model during testing so as to acquire a detection result. According tothe method disclosed by the invention, by applying an encoder / decoder structure, the acquired features are enhanced, small targets are detected by using a multi-scale attention structure, and referring to a residual difference structure, the problems of gradient disappearance and gradient explosion are solved. Compared with the other advanced traffic identifier detection methods, the method disclosed by the invention has the advantage of competitiveness.

The invention provides a method of detecting small target nucleotide sequences, in particular, small RNA species that are present in a sample. The method generally comprises a poly-A polymerization step or a ligation step to add a universal sequence to the 3′-end of all RNA molecules, followed by a universal primer-mediated cDNA synthesis, solid-phase selection, assay oligo annealing, extension and PCR amplification / labeling. The method of the invention can be practiced to amplify and label a small amount of miRNA or other ncRNA. The resulting amplification product can be read out on a universal array or an array with miRNA-specific or ncRNA-specific probes. The invention has multiple embodiments, including methods, compositions, and kits. In general, the nucleic acids, compositions, and kits comprise materials that are useful in carrying out the methods of the invention or are produced by the methods, and that can be used to detect small target nucleic acid sequences present in samples, in particular, small RNA species.

The invention discloses a small target detection and recognition method for enhanced feature learning, belongs to the field of image processing, pattern recognition and computer vision, and solves theproblems of low small target detection and recognition task detection precision and low network efficiency in the prior art. The method comprises: sequentially constructing a basic network module, afeature extraction module, a candidate box generation module and a prediction output module to serve as a small target detection and identification network; preprocessing the extracted small target sample image data based on the extracted small target sample image data; inputting the preprocessed small target sample image data into the small target detection and recognition network with initialized parameters for training to obtain a trained small target detection and recognition network; and inputting a to-be-predicted small target image into the trained small target detection and recognitionnetwork, and outputting the prediction box position and category information of the small target end to end through forward propagation. The method is used for small target detection and recognition.

The invention discloses a convolutional neural network-based small target detection method and a convolutional neural network-based small target detection model, and the method comprises the steps: carrying out the marking of small targets in a training set image, and building a small target data set; construction a training platform with caffe-ssd as the bottom layer; constructing a single-step detector model ELFSSD for enhancing low-layer feature fusion; adopting a model through pre-trained VGG-16 model initializing, inputting a small target data set in an lmdb format, carrying out iterativetraining; and detecting a small target in the detection set image by using the trained ELFSSD model. The method enhances the low-level characteristics, removes the high-level redundant characteristics, simplifies the detection process, improves the detection speed, accurately detects the small target in the image in real time, and solves the problem of poor detection effect of the small target inthe prior art.

The invention discloses a multi-scale target detection method based on a self-attention mechanism. By adopting a bottom-to-top and top-to-bottom multi-scale feature fusion mode based on a self-attention feature selection module, low-level features and high-level features of a target can be combined, the representation capability of a feature map and the capability of capturing context informationare enhanced, and the stability and robustness of a target detection stage are improved. Moreover, the self-attention module is used for re-calibrating the features, the calculated amount is smaller,the detection precision and speed are both considered, and the method has important significance for solving the detection problems of dense objects, small targets, shielded targets and the like in target detection.

The invention discloses a target detection method which comprises the following steps of: acquiring a to-be-detected image, wherein the to-be-detected image is subjected to multi-layer convolution extraction of features in a neural network to generate a feature map; loading modified structural parameters in a neural network model, and generating corresponding anchor box coordinates on the basis ofthe structural parameters, wherein the preset structural parameters comprise a reference dimension of an anchor box, an anchor box scale and a length-width ratio of the anchor box; generating candidate box coordinates on the basis of a region nomination subnet, taking a corresponding region on the feature map according to the candidate coordinates, and by pooling of a ROI (Region Of Interest), obtaining corresponding features; and on the basis of the features, determining prediction box coordinates, and on the basis of the prediction box coordinates, determining a target object position. Theinvention further discloses a target detection device and a computer readable storage medium. According to the invention, a case of generating an optimized prediction box to determine a target objectis implemented, a small target can be detected, and a detection rate for the target is improved.