183results about How to "Reduce redundant calculations" patented technology

The invention discloses establishment of a virtual case library of cancer pathological images and a multi-scale cancer detection system based on a convolutional neural network. The system is based ona method of a convolution neural network, a cancer mass region is detected on a pathological full-scan section, and the system includes four modules: 1) a pathological section image preprocessing module; 2) a virtual case database construction module; 3) a high-scale cancer mass detection module; and 4) a small-scale cancer mass classification module. The multi-scale cancer detection system provided by the invention can make full use of the multi-scale information of a pathological image, on different scales, according to the characteristics of the image, different strategies are designed to detect a suspected cancer area, and at the same time, under the condition of insufficient training data, the virtual case library method established by the invention can provide more training data setsfor an existing data-driven deep learning method. The multi-scale cancer detection system based on a convolutional neural network has the characteristics of multi-scale detection, driving of a relatively small amount of data and the like, and has the characteristic of reducing computing resources required for one-time recognition and improving time efficiency of an algorithm on the basis of ensuring the overall recall rate and accuracy.

The invention discloses a low-decoding complexity rate matching polarization code transmission method based on a QUP method, and belongs to the technical field of the channel coding / decoding. The method comprises the following steps: step one, determining a punch position and a punch parameter; step two, constructing an information sequence; step three, performing polarization coding, and outputting a coding sequence; step four, outputting the punched coding sequence according to the punch parameter determined in the step one; step five, sending the coding sequence punched in the step four one bit by one bit by a sending end; step six, receiving the coding sequence sent in the step five through channel transmission by a receiving end; step seven, optimizing a decoder structure according to the punch parameter determined in the step one, and outputting the optimized SC decoder; and step eight, performing polarization code decoding by using the SC decoder optimized in the step seven. By using the method disclosed by the invention, the spatial complexity and the time complexity of the polarization code decoding are effectively lowered, and the saved time and time source is increased along the increasing of the punch number.

The invention discloses a collaborative virtual and actual sheltering treatment method in a shared enhanced real scene. The collaborative virtual and actual sheltering treatment method comprises the following steps of: generating an enhanced real scene; estimating the position of a point to be observed; projecting both a virtual object and a real object on an observation plane of the point to be observed, wherein the sheltering relationship between the virtual object and the real object is to be analyzed; taking the point to be observed as an initial point; taking a ray vertical to an imaging plane to scan the observation plane along the horizontal direction; carrying out sheltering encoding according to a successive sequence that the ray touches two sides of the virtual object and two sides of the real object when the ray scans the observation plane; determining whether the virtual object and the real object which are observed from the point to be observed are in either a sheltering relationship or a non-sheltering relationship according to the difference among sheltering codes; and specifically judging that the virtual object shelters the real object or the real object shelters the virtual object so as to reduce wrong judgment. According to the collaborative virtual and actual sheltering treatment method in the shared enhanced real scene, the quick judgment of the spatial sheltering relationship between the virtual object and the real object in a new video sequence region is achieved, and the accuracy in the judgment of the virtual and actual sheltering relationship in the enhanced real scene is improved.

The invention discloses an image style migration method combining a meta-learning mechanism and feature fusion. According to the image style migration method, feature fusion based on convolution calculation and a method for decoding a feature map by using the meta-learning mechanism are combined. Firstly, content features and style features are preliminarily fused through convolution calculation,weighted summation operation is carried out on a preliminarily fused feature map and a preliminarily fused content feature map, and the stylization degree is controlled by adjusting the weight; and then the fused feature map is decoded into a stylized image by using a meta-learning mechanism, and secondary learning is carried out on the style in the decoding process, so that full expression of style features is ensured. According to the invention, the quality of the stylized image is improved, so that the style of the synthesized image is fairer than the original style; controlling the degreeof stylization based on the characteristics of the content image and the style image; a meta-learning mechanism is used to simultaneously carry out style secondary learning and feature map decoding operation so that style migration time is shortened and stylization of any image is rapidly realized.

The invention discloses a method for detecting thermal imaging defects of a high pressure vessel on the basis of variable step size segmentation of lines. The method comprises the following steps: selecting the maximum pixel value point from an eddy current pulse thermal image, calculating a correlation coefficient according to the maximum pixel value point by utilizing a Pearson correlation coefficient method, and further obtaining the transformed step size; calculating transient thermal response of each pixel point in combination with the transformed step size, performing classification processing, and performing defuzzification on all the pixel points by utilizing a membership degree maximization criterion so as to obtain the category to which each pixel point belongs; finally, performing feature extraction after dimensionality reduction by adopting a Canny operator algorithm, so as to extract the defect features of the eddy current pulse thermal image.

The invention discloses a method for verifying the integrity of data in hybrid cloud storage. The method is characterized in that based on a PDP algorithm, a secret key is distributed for the file anda label is generated by means of a chameleon signature algorithm and a private cloud and is stored in a public cloud; when the selected file is verified, the private cloud generates verification information and sends the verification information to the public cloud, the public cloud obtains the secret key and the label to form evidence, the private cloud serves as a platform, the evidence collides with the to-be-verified file block, and whether the to-be-verified file block is complete or not is judged. According to the data integrity checking method and device, by means of chameleon hash andpublic cloud and private cloud architecture, splitting of the label generation stage in the data integrity checking scheme is achieved, the online waiting time of a user is shortened, the user experience is improved, redundant calculation in the dynamic updating stage is reduced, and the calculation overhead is reduced.

The invention relates to a digital circuit, and discloses a memory optimization type static time sequence analysis method and system. According to the method, on the premise that the subgraphs are notexpanded, in the traversing and scanning process, false paths are identified and discarded in real time, the influence of the false paths on downstream nodes is reduced, redundant calculation is reduced, meanwhile, multiple time sequence constraints on each node are dynamically and differentially processed in the traversing process, dichotomy coverage is correspondingly carried out, and the number of labels on each node is reduced. According to the application, occupation of system memory space can be greatly reduced, the system operation efficiency is improved, and the system performance isimproved.

The invention discloses a lung parenchyma CT image segmentation method based on a weighted full convolutional neural network, and belongs to the field of medical image processing. The method comprisesthe following steps: selecting a public lung data set for preprocessing, and extracting a lung parenchyma boundary in a labeled image as a semantic category; designing an improved network structure based on a standard full convolutional neural network framework, and establishing an overall structure framework of the pulmonary parenchyma segmentation convolutional neural network by using a principle that a standard path structure for encoding and decoding simultaneously comprises jump connection, expansion convolution and batch normalization; adopting a weighted loss function layer; dividing the data set; carrying out offline model training out to acquire model weight parameters; inputting a test image and outputting a segmentation result by an output layer through layer-by-layer feedforward of a network. According to an existing lung parenchyma segmentation method, a segmentation missing phenomenon is prone to occurring in a focus area in lung parenchyma, and correct segmentation of the focus area in lung parenchyma segmentation can be effectively improved through enhancement processing on important pixels.

The invention discloses a kernel sparse representation-based fast remote sensing target detection and recognition method. The method includes the following steps that: S1, four RGB characteristic channels are created; S2, the four-phase Fourier transformation of the four characteristic channels of a given image is calculated, a phase spectrum is extracted, the images of the four characteristic channels are reestablished through inverse Fourier transformation, and a saliency map can be generated; S3, binaryzation division is performed on the saliency map obtained in the step S2, and candidate regions of interest are extracted; S4, a search box is scanned through an effective sub-window search algorithm, so that image blocks to be detected are obtained, so that a remote sensing target image block training set is obtained; S5, SIFT features are extracted from the remote sensing target image block training set, and a sparse dictionary is generated; S6, a spatial pyramid is adopted to map the SIFT features; S7, kernel sparse representation is obtained; S8, the kernel sparse representation is solved; S9, the space pyramid vector representation of a target is performed; and S10, a linear support vector machine classification algorithm is used in combination to complete a recognition task.

The invention relates to a face recognition method based on an elastic context relation loss function. According to the method, how to overcome the influence of massive unbalanced data on face recognition training is studied; according to the method, the combined data blocks can be effectively mined, useless redundant calculation is reduced, the whole training process is accelerated, the designedtarget function based on the elastic context does not need to introduce extra training parameters for each class, the influence of long tail classes in large-scale training is reduced, and meanwhile,the face recognition accuracy can be improved through soft spacing.