35 results about "Cascade algorithm" patented technology

The invention relates to a cascaded depth neural network-based face attribute recognition method. The method includes the following steps that: 1) a cascaded depth neural network composed of a plurality of independent convolution depth neural networks is constructed; 2) a large number of face image data are adopted to train networks at all levels in the cascaded depth neural network level by level, and the output of networks of previous levels is adopted as the input of networks of posterior levels, such that a coarse-to-fine neural network structure can be obtained; and 3) the coarse-to-fine neural network structure is adopted to recognize the attributes of an inputted face image, and final recognition results can be outputted. According to the cascaded depth neural network-based face attribute recognition method of the invention, a cascade algorithm system is adopted based on depth learning, and therefore, training time can be accelerated; and a cascaded coarse-to-fine processing process is realized, and the performance of a final network can be improved by networks of each level through utilizing information of networks of upper levels, and therefore, the speed and the accuracy of face attribute recognition can be effectively improved.

A computer-implemented method for de-identifying data by creating tokens through a cascading algorithm includes the steps of processing at least one record comprising a plurality of data elements to identify a subset of data elements comprising data identifying at least one individual; generating, with at least one processor, a first hash by hashing at least one first data element with at least one second data element of the subset of data elements; generating, with at least one processor, a second hash by hashing the first hash with at least one third data element of the subset of data elements; creating at least one token based at least partially on the second hash or a subsequent hash derived from the second hash, wherein the token identifies the at least one individual; and associating at least a portion of a remainder of the data elements with the at least one token.

The invention relates to the technical field of communication, and particularly relates to the safety certification problem among a tag, a reader and a back-end database in a mobile RFID system. The method comprises the following steps: (1) the reader sends the request information to the tag; (2) the reader sends the tag response request information and the own information to the database; (3) thedatabase authenticates the reader and the tag; (4) the reader authenticates the database; (5) the tag authenticates the database. According to the invention, the protocol of the system adopts a cyclic checking function and a simple exclusive-or and cascade algorithm to encrypt the to-be-transmitted information. In this way, the anonymity of the tag and the bidirectional authentication of the system are realized. Meanwhile, the calculation amount of the database and the tag is effectively reduced. A tag random number is generated by a back-end database random number generation mechanism and then is forwarded to the tag or automatically updated through the cyclic checking function. Therefore, the hardware cost of the tag is reduced. According to the invention, a dynamic ID and a cyclic checking function mechanism are utilized, so that the system can resist various attacks.

The invention discloses a method of realizing video target tracking by adopting a two-layer cascading Boosting classification algorithm. The method is characterized in that image pretreatment is used to acquire target samples and image blocks in target areas; characteristic values of positive samples and negative samples can be acquired by adopting characteristic extraction of filters; the two-layer Boosting cascading algorithm comprising the on-line characteristic selection and the weight training is provided; and a classifier is used to detect the target tracking. A plurality of filters are used for the characteristic extraction of the image blocks, and the characteristics of the image blocks can be expressed in a refined manner, and then the problems such as single template formed by adopting Haar-like characteristics, loss of image colors and texture details can be solved; and in addition, by adopting the two-layer cascading structure, the filter types and the image block positions can be selected respectively, and then the selected characteristics are more suitable for the tracking task.

The invention relates to a linear programming and minimum sum cascading decoding method for an LDPC (low-density parity-check) code and belongs to the technical field of communication. The method is applied to the decoding module of a communication system and is used for increasing the property of the decoding module by cascading two decoders on the basis of linear programming and minimum sum decoding, so as to efficiently improve the communication quality of the communication system within a wider scope of signal to noise ratio. A cascading factor is related to a cascading part. The method comprises the following steps: firstly acquiring a generalized fraction solution which also can be an integer; and then choosing the corresponding cascading factor from a factor table which is acquired by performing numerical calculation. Compared with the single linear programming decoder, the cascading algorithm related to the invention has the characteristic of self-adaptively increasing the decoding property. Compared with the single linear programming and minimum sum decoding method, the improved decoding method can be used for efficiently increasing the error-correcting capability of the decoding module at a channel receiving end and improving the whole communication quality of the communication system within a wider scope of signal to noise ratio.

The invention discloses a pedestrian distance detection method and device. The method comprises the following steps: calculating a rectangle feature of a pedestrian image shot by a monocular camera, and training the rectangle feature to obtain multiple weak classifiers; combining the multiple weak classifiers into a strong classifier via a cascade algorithm; acquiring pixel coordinates of a pedestrian in the pedestrian image via the strong classifier; and calculating the distance between the monocular camera and the pedestrian according to the parameter of the monocular camera and the pixel coordinates of the pedestrian. The method can improve the instantaneity and robustness of pedestrian distance detection.

The invention discloses a face real time detecting method based on video and device thereof, belonging to the field of multimedia. In the invention, a video image frame is divided into a full detection frame and a predetermined frame; the type of the current input video image frame is judged according to the frame interval between the current input video image frame and the former full detection frame; and a corresponding detection way is adopted. As video sequence has predictable characteristics in a time-space domain, parts of the video image frame are detected by searching different regions by different step lengths, so that the method can efficiently reduce the searching time in the detecting process, avoid the shortage that the general algorithm based on AdaBoost cascade algorithm searches all the regions by fixed step length, greatly reduce the calculating complexity in the detecting process and guarantee the accuracy of the detection.

The invention provides a pedestrian detection method based on the multi-granularity feature, and belongs to the technical field of video image processing. The method comprises that S1) the multi-granularity feature vector in a training sample set is extracted; S2) the multi-granularity feature vector is trained on the basis of a Gentle Adaboost cascade algorithm to obtain a pedestrian cascade classifier; S3) the pedestrian cascade classifier is used to carry out pedestrian detection on input images obtained in a self-monitoring area; and S4) pedestrians detected in the S3) are tracked and counted in the EKM algorithm. According to the invention, the multi-granularity feature vector in the training sample set is extracted, the pedestrian cascade classifier is utilized in an optimized manner to detect the pedestrians in the monitoring area, and the pedestrian counting efficiency and accuracy in the common area are effectively improved.

The invention relates to a low-delay video coding adaptive Qp cascade method. Firstly a video sequence is divided into multiple image groups, the layered structure of the image groups is constructed, and each image group is composed of four frames which are arranged at different layers; then a rate distortion optimization model is derived according to the dependency between different layers and the dependency between the image groups; then a correlation model is introduced, the rate distortion optimization model is converted into the problem of solving the optimal Qp, and the model is enabled to be adaptive to different video sequences by updating the model parameters; and finally the optimal Qp of each layer of the model is solved by introducing the Lagrangian multiplier ^ and the Newton-Raphson method, the Qp of each frame is reasonably allocated and thus rate distortion optimization can be realized. The adaptive cascade algorithm for low-delay video coding is put forward for mainly solving the problem of allocation of low-delay video coding.

The invention provides an image compression method based on a multilayer convolutional neural network, belongs to the technical field of image processing. A cascade algorithm system on the basis of the deep neural network is introduced, the training time is accelerated, the training effect is improved, the working efficiency is greatly improved, so that simple and tedious work is automatic, and the time is saved.

The invention provides a face detection method based on multiple classifiers, comprising the following steps: selecting positive sample images and negative sample images, and clustering the positive sample images through a K-Means clustering algorithm to get clustered positive sample sets; training the clustered positive sample sets and the negative sample images through a main classifier and sub classifiers by use of a Cascade algorithm to get a trained main classifier and trained sub classifiers; and inputting an image, detecting the input image according to the trained main classifier and the trained sub classifiers, and outputting a detection result. Compared with the prior art, a face can be detected from a complex image, and the detection accuracy is high.

The invention discloses an error code negotiation method of a quantum key distribution system, solves the problem that the negotiation efficiency is reduced due to collision in the error code negotiation process of a conventional Cascade algorithm, and belongs to the technical field of quantum communication. The method comprises the following steps: S1, in the first round, carrying out parallel processing on a binary lookup odd-even error correction process; S2, in the second round, obtaining blocks containing odd errors, placing the block with the highest collision probability in a processingset, and placing the rest blocks in an error set; S3, calculating the collision probability of blocks in the error set and the processing set, putting n blocks with the minimum collision probabilityin the error set a into the processing set, conducting one-step binary search odd-even error correction on the blocks in the processing set in parallel each time, when the error blocks are found, removing the blocks from the processing set, determining the positions of the error blocks in the previous round, and if the blocks are in the error set, removing the blocks are removed, and otherwise adding the blocks are added into the error set; S4, repeating S3 until the error set is 0, and turning to S5; and S5, carrying out all the following rounds according to the S2 to S4.

The invention provides a fluorescence immunochromatography quantitative image peak searching algorithm based on deep learning. The fluorescence immunochromatography quantitative image peak searching algorithm comprises the following steps: collecting a large amount of fluorescence immunochromatography quantitative image data; labeling peak point positions in the collected fluorescence immunochromatography quantitative images to obtain label information of the image data; performing standardized preprocessing on the label information, and establishing an algorithm training set; establishing a first-layer convolutional neural network of a cascade algorithm, and positioning a peak point in a very small error range; a second-layer convolutional neural network of the cascade algorithm is established, so that the result is more accurate; and performing standardized preprocessing on the data of the test set, establishing the test set, inputting the test set into the trained algorithm network,and testing the peak searching accuracy of the fluorescence immunochromatography image. According to the fluorescence immunochromatography quantitative image peak searching algorithm based on deep learning, correct peak points can be recognized, and accurate peak point coordinate data can be output.

The invention provides an FBG optical fiber sensing demodulation system for marine environment monitoring. The system comprises a signal input source and a signal sensing module, and further comprisesa demodulation technology module and a signal processing module. The demodulation technology module is used for receiving and demodulating a reflected optical signal output by the signal sensing module, outputting a demodulated transmitted optical signal, and converting the transmitted optical signal into an electrical signal, the signal processing module collects the electrical signal output bythe demodulation technology module, and information of demodulated monitored variable is obtained through a signal processing algorithm. The system is based on a tunable optical fiber FP filter, adopts a cascade algorithm combing a peak-detection algorithm and an adaptive filtering improvement algorithm to improve the demodulation accuracy and improve the system performance in the premise of not changing hardware facilities.

An object detection method can begin with receipt of an ensemble classifier previously trained upon a binary dataset of images for an object. The ensemble classifier can be defined by a list, W, of N weak classifiers, wn for n=1, . . . , N. The ensemble classifier can be trained as a detector for the object using a training dataset of images; each image can have a set of candidates having one or more known ground truth candidates. As a result of training, the weak classifiers can be reordered, based upon performance, into a new list, W′, and a rejection threshold on candidate rank can be assigned for each detection stage. A cascading algorithm can be performed using the trained detector to detect the object in an image having known object candidates. The rejection threshold for a stage of the cascading algorithm can be applied to a rank of the object candidates.

The invention discloses an optimization algorithm for extracting independent sift key points based on principal component analysis. The optimization algorithm comprises the processing steps of firstlyperforming histogram matching on histograms of sift feature points obtained by a sift algorithm and surrounding pixels, and enabling the histograms to be in Gaussian distribution; secondly, when thehistograms are enabled to serve as inputs of a pca algorithm, indicating that principal components output by pca are mutually independent; and finally, mapping the principal components to histograms of original sift feature points and surrounding pixels through eigenvalues, and finishing the screening of the sift feature points, wherein the obtained sift feature points are mutually independent. The optimization algorithm has the beneficial effects that for the problem of a matching error caused by similarity of the sift feature points, the method for extracting the independent sift key pointsby utilizing the pca is proposed; the screening of the sift feature points is finished; and the to-be-matched key points are mutually independent, so that the probability of the matching error is reduced.

The invention provides a multi-energy system optimization scheduling method based on a knowledge migration Q learning algorithm. According to the method, a joint optimization scheduling framework of amulti-energy system is built based on an energy center modeling method; a typical multi-energy system joint optimization scheduling model considering the energy supply cost of a valve point effect and a carbon emission target is built; and based on the constructed model, a cascaded algorithm composed of a knowledge migration Q learning algorithm and an interior point method is provided for solving, wherein a multi-energy system determined after upper layer Q learning takes the active power of an unit as an action variable and lower layer takes an interior point method to solve the active power of the unit is used for optimize the model, and the solving efficiency is improved through knowledge migration.

The invention discloses an apple picking machine. The apple picking machine comprises a base, a first motor base, a storage bin and a mounting rack, the first motor base is arranged at the bottom inside the base, a servo motor is arranged on the first motor base, and a driving shaft is arranged at the end of the servo motor. The apple picking machine has the advantages that color obtaining is conducted on apple trees through a color sensing camera, after imaging is conducted through a color imager, track information is obtained through a cascade algorithm processor, and then a jaw is driven tomove to a specified working position, and the claw is driven by an air pump to act to pick apples and convey the apples into the storage bin. A traditional manual picking mode is replaced, the workload of workers is reduced, and the efficiency is improved; meanwhile, a gravity sensor is arranged on the bottom surface of the storage bin, gravity information of the apples in the storage bin can beobtained in real time, the gravity information is converted into weight information through a control main unit to be displayed on a display screen to allow a user to conduct observation and counting,subsequent carrying, weighing and counting do need to be manually conducted, and the quality of the apples is guaranteed.

The invention discloses a time-to-digital converter system and a multiple delay phase-locked loop comprising the same. The system comprises a first-level time-to-digital converter, a first-level digital-to-time converter, a first-level time amplifier, a second-level time-to-digital converter, a second-level digital-to-time converter, a second-level time amplifier, a third-level successive approximation register type analog-to-digital converter and a digital-to-analog-to-digital converter. The time-to-digital converter system provided by the invention and the multiple delay phase-locked loop capable of lowering in-band quantization noise use a cascade algorithm similar to Delta-Sigma, the accuracy of the time-to-digital converter applied to the multiple delay phase-locked loop is effectively improved, the quantization noise is lowered, and the generation of the stray of the multiple delay phase-locked loop is eliminated.

The invention discloses a client evaluation platform of garment factory products. The platform comprises a base, a driven sliding rail and a workbench, the driven sliding rail is arranged on the uppersurface of the base, and a lifting arm is arranged on the driven sliding rail; a supporting rack is arranged on the side surface of the lifting arm, and a drive sliding rail is arranged at the end ofthe lifting arm. The client evaluation platform of the garment factory products has the advantages that a grading area is arranged on a sensing plate, when a client conducts grading on the grading area, a distributed optical fiber sensor can sense the change of the stress at a certain position, then grading calculation is conducted by a cascade algorithm processor according to corresponding evaluation of the position where the stress is changed, scores are displayed on a display screen in real time, an evaluation result of the client is visually displayed, and the platform is convenient to operate; meanwhile, a lifting mechanism which is composed of the drive sliding rail, the lifting arm and the driven sliding rail is arranged, the height of the workbench can be freely adjusted, the platform adapts to clients of different heights, and the human-friendly design is achieved.

The invention discloses a generalized load random characteristic-considered power transmission and distribution coordinated planning method, which comprises the following steps of decomposing a transmission and distribution coupling system into a plurality of small systems according to a physical structure, and then constructing a transmission and distribution iterative interaction distributed decision framework, firstly, fitting a probability density function of generalized loads containing distributed new energy according to measured data, randomly sampling generalized load values by using a Monte Carlo sampling method, and sequentially accessing the generalized load values into a power distribution network with new energy power generation, and independently operating each power transmission system and each power distribution system, and solving the boundary power between power transmission and power distribution. According to the method, the mean value and the standard deviation of the boundary power are used as shared variables, the probabilistic target cascade algorithm is used, information transmission between the power transmission system and the power distribution system is achieved, the target function with the Lagrange penalty function is listed with the minimum total operation cost of the system as the target, and therefore the calculation precision is not reduced, and the operation speed can be greatly improved.

The invention discloses a method for realizing video target tracking by using a two-layer cascaded Boosting classification algorithm. The method includes: image preprocessing, obtaining target samples and image blocks in the target region; extracting features through filters to obtain positive and negative The feature value of the sample; the two-layer Boosting cascade algorithm including online feature selection and weight training; the target tracking of classifier detection and other steps. The present invention uses multiple filters to extract features from image blocks, expresses the features of image blocks more finely, and solves the problem of using a single Haar-like feature to construct a template and losing image color and texture details; in addition, using two-layer cascaded structure, the filter type and image block location are selected separately, so that the selected features are more suitable for tracking tasks.

Disclosed in the embodiments of the present invention are a method and relative device for upgrading cryptographic algorithm used for improving security in the process of upgrading cryptographic algorithm. The method in the embodiments of the present invention comprises: a server transmits a upgrading signal of the cryptographic algorithm to a device to be upgraded, chooses, from a target algorithm set, one or more target algorithms to perform cascade to generate a cascade algorithm when determining that all the algorithms in the target algorithm set are disabled algorithms after receiving a connection request sent from the device to be upgraded, and upgrades the cryptographic algorithm for the device to be upgraded by the upgrading packet of the cascade algorithm and cryptographic algorithm. As the cascade algorithm can improve the strength of algorithm security effectively, the security during the process of upgrading cryptographic algorithm can be improved effectively.

The invention discloses a power transmission line damper deformation defect detection method based on a cascade R-CNN algorithm, and the method comprises the steps: carrying out the preprocessing of a damper image obtained through the aerial photographing of an unmanned plane, and obtaining a damper image data set after the preliminary preprocessing; based on the preprocessed image data, constructing a vibration damper target detection training set, a verification set and a test set; using data in the training set to carry out iterative training on the Cascade R-CNN model; inputting the test set into the Cascade R-CNN model to test the model precision; judging whether the vibration damper is deformed or not and marking the finally obtained deformation defect area of the vibration damper to be detected. The invention has the beneficial effects that the problems of overfitting caused by an overlarge IoU threshold value and mismatching of the IoU threshold value and a threshold value in a trainer in the existing detection technology are solved, the capabilities of detecting the vibration damper of the power transmission line and identifying the deformation defect of the vibration damper are improved, and the inspection working efficiency is improved.

The invention provides a bottled white spirit detection method based on a trajectory cascade algorithm. The method comprises the following steps: establishing a suspicious object target detection deeplearning model file, and initializing a suspicious object target detection algorithm; acquiring a suspicious object target detection video image; deep learning suspicious object target detection; deep learning suspicious object target tracking; and carrying out trajectory cascade analysis on suspicious object impurities. The invention further provides a detection device. The device comprises a target detection deep learning algorithm unit, a target tracking deep learning algorithm unit and a trajectory cascade algorithm unit. According to the method, deep learning algorithms and the track cascade algorithm are combined, and the problems that in the prior art, bottled liquor detection hardware investment and equipment maintenance cost are high, moving bubbles are prone to being detected bymistake, and the deep learning algorithms have limitation are solved. Through a new track cascade analysis algorithm, the influence of bottle body traces, bubbles and the like is accurately eliminated, the accuracy of white spirit impurity detection is greatly improved, and the cost performance of the product is very high.

The invention discloses a common English vocabulary query device for computer software programming. The device comprises a box, a damp-proof mat, a box door and an object placement frame, the bottom surface of the box is provided with the damp-proof mat, the side surface of the box is provided with the box door, one side of the box door is provided with a hinge, the other side of the box door is provided with a latch, and the surface of the box door is provided with a display screen. The common English vocabulary query device has the advantages that the inner side of the box is provided with ashielding layer made of graphite, the bottom surface of the box is provided with the damp-proof mat made of EVA foamed plastic, and therefore the device can be shielded from external electromagneticinterference and protected from dampness; annotations and other information of common English vocabularies are transmitted to a storage device for storage through a USB interface, then a user inputs query vocabularies through an input keyboard, the annotations and other information of the vocabularies are displayed on the display screen after a cascade algorithm processor conducts matching, and the operation is convenient; the heat dissipation function is achieved through the arrangement and combination of a heat dissipation net, a servo motor and fan blades, and the situation can be avoided that the use of the device is affected due to heat accumulation in the device.

The invention discloses a computer scoring display device for a golf practice court. The device comprises a base, an outer frame and a ball body, wherein mounting holes are arranged in the surface ofthe base, and a support post is arranged in the middle of the upper side of the base. The device has the beneficial effects that a wireless signal transmitter is used for transmitting wireless signals, a wireless signal receiver is used for receiving the wireless signals, a cascade algorithm processor is used for carrying out cascade algorithm processing, a scoring result is displayed on a displayscreen in real time for observation and statistics of a practicer, the traditional manual field statistics method is replaced, thus the workload of people is alleviated, and the accuracy of the statistics result is also guaranteed; meanwhile, a solar thin film panel is arranged, the solar energy can be converted into electric energy which is provided for the electricity utilization parts of the device for use, the self power supply mode is adopted, and thus the device is environmentally friendly; and moreover, an electricity amount detector is arranged, so that the electricity amount of a storage battery can be detected in real time, then when the electricity amount is insufficient, a control main machine controls a buzzer to send out alarm sounds for reminding a user to carry out timelytreatment, and the use is convenient.

The invention belongs to the video image processing technical field and provides a time-space movement profile feature-based pedestrian detection method. The method includes the following steps that: S1, a 3D Haar filter is utilized to extract the 3D Haar feature vectors of a time-space body in a training sample set; S2, the 3D Haar feature vectors are trained based on a Gentle Adaboost cascade algorithm, so that a pedestrian cascade classifier can be obtained; S3, the pedestrian cascade classifier is utilized to perform pedestrian detection on input images acquired by a self monitoring region; and S4, pedestrians detected in the step S3 is tracked and counted based on an EKM algorithm. According to the method of the invention, the 3D Haar feature vectors in the training sample set are calculated, and the optimized pedestrian cascade classifier is utilized to detect pedestrians in the monitoring region, and therefore, the efficiency and accuracy of counting of pedestrians in a public region can be effectively improved.