857results about How to "Achieve compression" patented technology

The invention discloses a method and a device for processing a video stream based on a cloud desktop. The method comprises the steps of monitoring the network state between a cloud desktop server and a terminal to determine network parameters, receiving the running performance parameters of a terminal and determining the running status of the terminal, adjusting a video frame compression ratio and a frame rate according to the running status of the terminal and the network parameters, and compressing video data based on the video frame compression ratio and the frame rate to obtain the video stream matched with the present running status of the terminal and the network state. According to the method for processing the video stream based on the cloud desktop, the video frame compression ratio can be adjusted in real time according to the load condition of the current network and the frame rate of the video frames is adjusted according to the running status of the terminal, and therefore, the compression method of cloud desktop video transmission is optimized and the requirement of a cloud desktop system on bandwidth is reduced.

The invention discloses a compression and decompression method based on a hardware accelerator card on a distributive-type file system, which belongs to the technical field of a distributive-type file system. The method comprises the steps of firstly fragmenting a file block when a client writes data to a data node, compressing fragments through the hardware accelerator card, and transmitting the compressed fragments to the data node; and firstly retrieving all fragments containing reading data from the data node when the client reads the data from the data node, calling the hardware accelerator card to decompress and combine the fragments, and transmitting the combined data to a superior application. The superior application can independently utilize the hardware accelerator card to compress or decompress the data in a data flow form. The hardware accelerator card is used in the distributive-type file system and the superior application, so that the data compression or decompression can be realized only through small cache, a great amount of central processing unit (CPU) resource consumed by a traditional compression method can be unloaded, and complete transparency for users of the system can be realized.

The invention discloses a depth convolution neural network acceleration and compression method based on parameter quantification, and the method comprises the steps: carrying out the quantification of parameters of a depth convolution neural network, and obtaining a plurality of sub-codebooks and indexes corresponding to the sub-codebooks; and obtaining an output characteristic graph of the depth convolution neural network according to the plurality of sub-codebooks and indexes corresponding to the sub-codebooks. The method can achieve the acceleration and compression of the depth convolution neural network.

A calcaneus plate provided with two vertically-spaced plate sections, with corner-located screw holes. An intermediate wall connects the plate sections. An arcuate protrusion extends from the intermediate wall and contains an internal sliding mechanism for a compression screw. The sliding mechanism includes an arc-shaped compression hole with an elongated opening that includes a ridge upon with the compression screw head can slide.

A spread spectrum waveform generator has a photonic oscillator and an optical heterodyne synthesizer. The photonic oscillator is a multi-tone optical comb generator for generating a series of RF comb lines on an optical carrier. The optical heterodyne synthesizer includes first and second phase-locked lasers, where the first laser feeds the multi-tone optical comb generator and the second laser is a single tone laser whose output light provides a frequency translation reference. At least one photodetector is provided for heterodyning the frequency translation reference with the optical output of the photonic oscillator to generate a spread spectrum waveform. A receiver pre-processor may be provided to operate on the spread spectrum waveform.

The present invention provides a device for several blocks combined compressing video data and a method thereof, the device comprises a camera head, an image division processor, several single-block compressing treatment modules, a microprocessor and a memorizer, wherein, the single-block compressing treatment modules are connected with the image division processor and the microprocessor respectively, the memorizer is connected with the microprocessor and each single-block compressing treatment module respectively, and the image division processor is connected with the camera head and the microprocessor respectively. The present invention realizes the compression of high-definition video data and can be effectively applied in any high-definition image sampling system, especially in a patching display wall system of large screen DLP beitou TV; the system can not only effectively accomplish high-definition image sampling, but also can compress and memorize the data in real time and for a long period; and it has notable effects on enhancing the quality of multimedia products and on improving the compressing ca[ability of the high-definition video.

The invention discloses a multi-field coupled environment simulating and online monitoring / observing system. The system comprises: a main body test box provided with a quartz observing window and an infrared window; a mechanical loading module comprising two coaxially arranged mechanical loading shafts and respectively connected with the main shafts of a universal testing machine and a sample clamping device; a multi-field service environment module arranged outside the main body test box and comprising an infrared laser rapid heating device, an X-ray radiation device, a corrosion gas box and a cooling system and respectively used for realizing heating, irradiating, corroding and cooling environments; and a damage online monitoring / observing module comprising a surface / interface strain real-time observation system, an in-situ fracture behavior observation system, a damage signal collecting system and an infrared thermodetector which are arranged outside the main body test box, and damage signal detectors arranged at the upper end and the lower end of a tested sample, and wherein devices contained in the damage online monitoring / observing module are respectively used for real-time capturing of strain information, in-situ fracture image information, damage information and temperature information, and synchronously run through a cynchronous control system.

The invention discloses an image retrieval database generating method, an image retrieval method, a corresponding image retrieval database generator, an image retrieval device and an image retrieval system. When a database is generated, sample image feature description information is classified and dimensionally reduced, and a label datum is formed for each feature point. Sample image content data and label data are stored in the database in a classified mode. During image retrieval, a target image to be retrieved is processed through a similar method, and a plurality of labels are formed for each feature point. During matching, traversal matching is carried out on the labels of the target image and all labels in a corresponding classified index in the database, and a matching value of the target image and sample images in the database is calculated. According to the image retrieval database generating method, the image retrieval method, the corresponding image retrieval database generator, the image retrieval device and the image retrieval system, the description content is compressed, and data transmission efficiency can be effectively improved; in addition, due to the fact that the sample image content data and the label data are stored in the classified mode, the image retrieval waiting time can be effectively shortened, and real-time retrieval can be possibly achieved in the large database.

The invention discloses an anomaly detection method based on a time series. The anomaly detection method based on the time series comprises intra-node data correlation analysis and inter-node data correlation analysis. The intra-node data correlation analysis describes time correlation of sensor node data and is to compute the data correlation between current data and historical data. The inter-node data correlation analysis describes spatial data correlation between different sensor nodes and is to compute the data correlation between the sensor node data and data of neighbor nodes of sensor nodes. The anomaly detection method based on the time series can be used for effectively detecting abnormal data, and good performance can still be kept under the influence of wrong data. Meanwhile, an approximation method of the time series data is applied so that the time series data can be effectively compressed, a method for measuring the similarity between the compressed time series is put forward and used for clustering analysis of the time series, and consequently anomaly detection based on the time series is achieved.

The invention relates to a fixed large-bale straw baler comprising a feeding and conveying device, a power driving device, a transmission mechanism, a reciprocating type compressing device, a shredding device, a material precompression chamber, a knotting mechanism, a knotting needle, a compression chamber and a bail discharging part, wherein the feeding and conveying device is connected to a feed inlet at the lower part of the material precompression chamber through the shredding device, a discharge outlet at the upper part of the material precompression chamber is arranged between the reciprocating type compressing device and the compression chamber, the front end and the back end of the compression chamber are respectively arranged on the reciprocating type compressing device and the bail discharging part, materials fed by the material precompression chamber is formed into material bales by the reciprocating type compressing device, the knotting mechanism and the knotting needle are respectively arranged above and below the compression chamber, the knotted material bales are pushed to slide out of the bale discharging part by the reciprocating type compressing device, and the power driving device is respectively connected with the reciprocating type compressing device, the shredding device, the knotting mechanism and the knotting needle through the transmission mechanism. The invention can realize the continuous operation of straw shredding, filling, compression, automatic knotting and automatic bale discharge and replaces manual operation.

The invention discloses a deep network compression method based on dimensional adaptation adjusting Tucker decomposition. The method comprises a dimensional adaptation adjusting process and a dimensional adaptation weight tensor decomposition process. Through adaptively adjusting a size of each dimensions of a tensor, a new any order tensor is generated. Through a learnable nuclear tensor and a transfer matrix, tensor decomposition is realized so that a purpose of network optimization compression is reached. Compared to an existing low rank compression method, by using the method of the invention, under the condition that network performance is maintained, a network parameter quantity has a greater compression multiple and a higher compression multiple can be acquired. Simultaneously, a guiding position of a nonzero element does not need to be stored, an index does not need to be recorded and a storage space can be effectively used.

The invention discloses an improved method for compressing a neural network. The method has the advantages that a pruning technology of the neural network is integrated into the original training process of the neural network, so as to reduce the training iteration frequency, thereby shortening the training cycle of a sparse neural network; the pruning technology can also be integrated into the compression retraining process of a trained model, so that the neural network can be compressed at the premises of ensuring the model accuracy.

The invention discloses a compressive-sensing-based digital image watermark embedding and extraction method, and belongs to the technical fields of information hiding and image processing. The method comprises the following steps of performing sparsification processing on binary digital image watermark information to obtain sparse one-dimensional watermark information, and constructing a measurement matrix which is used as a key; measuring the one-dimensional watermark information to realize the compression and the encryption of the one-dimensional watermark information by using the key; performing discrete cosine transform on original carrier image information, blocking the transformed discrete cosine transform-domain information, and embedding the processed watermark information into a carrier image to obtain a watermark-containing image. According to the method, a compressive sensing method is used for processing watermarks, so that dual functions of watermark compression and encryption are realized, the number of the embedded watermarks is increased, the invisibility and the security are enhanced, and the overall performance of the watermarks is remarkably improved; an SBHE (scrambled block hadamard ensemble) matrix is applied, so that the method is small in memory space, high in calculation speed and simple in hardware; sparsity is introduced into a compressive sensing reconstruction process, so that the reconstruction accuracy and effects are greatly improved.

The states associated with a programmable state machine are reordered to compress the storage of transitions which define the state machine. To reorder the states, a score is computed and assigned to each of the states. Next, the states are sorted according to their computed scores. In some embodiments, to compute the score for each current state based on the received input symbol, the number of times that the input symbol causes transition to similar states is added. The sum of the scores in each row of the table is representative of the score for the associated current state associated with that row. The states are sorted according to their score and a new state transition table is generated in accordance with the reordered states.

The invention discloses the field of dynamic characteristic research for interaction of laser and atmosphere in spectrum detection for trace gas and free-space optical communication, and particularly relates to an automatic tunable system with multiple optical paths. The automatic tunable system comprises angle rotating devices, length adjusting devices, supports, motor supports, stepper motors, a DSP (digital signal processor) controller and a cylindrical main body. The main body can keep vacuum and can bear certain high pressure. Signals are inputted into the DSP controller according to requirements of the optical paths, rotational angles of concave reflectors and required voltage are computed by the DSP controller, piezoelectric transistors (PZT) can drive the concave reflectors to rotate by the rotational angles after being supplied with the required voltage, signals are outputted for triggering and driving to realize automatic angular rotation, simultaneously, the stepper motors are driven, corrugated pipes are driven to stretch or contract, and distances d among reflecting pools are horizontally increased or reduced. The automatic tunable system with the multiple optical paths has the advantages of high precision and sensitivity, simplicity in operation and automatic control function.

A method for compressing test data of block-labeled system chip utilizes both complete block of data block to be coded and with zero label in compatible to reference data block as well as complete block of data block to coded and with ten label in anti-phase compatible to reference data block simultaneously to dynamically update reference data block in order to increase compatible probability or anti-phase compatible probability of reference data block and data block to be coded.

The invention discloses a data compression, transmission, receiving and uncompressing method and a corresponding device, can achieve IQ data compression with low complexity and small error, and can achieve transmission of IQ data compressibility factors with fewer resources. The data compression method comprises reading data to be compressed, finding I data and Q data with maximum absolute values from a group of data to be compressed, determining an effective bit number M of the I data with the maximum absolute value and an effective bit number N of the Q data with the maximum absolute value, generating a first digital automatic gain control (DAGC) factor for compressing the I data according to the effective bit number M and a compressed target bit wide number X, generating a second DAGC factor for compressing the Q data according to the effective bit number N and the compressed target bit wide number X, using the first DAGC factor to compress each I datum in the group of data to be compressed, and using the DAGC factor to compress each Q datum in the group of data to be compressed.

The invention discloses a network transmission method for a large extensible markup language (XML) document. The method comprises the following steps of: 1, reading the XML document; 2, preprocessing all node tags in the XML document; 3, processing redundant tags in the XML document; and 4, finishing the compression processing and fragmentation transmission of the XML document. The step 4 further comprises the following steps of: 41, setting an initial fragmentation size to be Initial_Size; 42, extracting a plurality of pieces of data to form a data packet for the compression processing and the network transmission, wherein the size of the data packet is Initial_Size; 43, judging a current network load condition, and obtaining a value of a maximum transmission unit; 44, re-determining a fragmentation size to be New-Size according to the current network load condition and the value of the maximum transmission unit; 45, re-extracting a data packet with the size of New-Size for the compression processing and the network transmission; and 46, repeating the steps 43 to 45.

The invention provides an image classification method and device, and relates to the field of artificial intelligence, in particular to the field of computing vision. The image classification method comprises the following steps: obtaining a convolution kernel parameter of a reference convolution kernel of a neural network and a mask tensor of the neural network, and performing Hadamard operationon the reference convolution kernel of the neural network and the mask tensor corresponding to the reference convolution kernel to obtain a plurality of sub-convolution kernels; and performing convolution processing on the to-be-processed image according to the plurality of sub-convolution kernels, and classifying the to-be-processed image according to the convolution feature map finally obtainedby convolution to obtain a classification result of the to-be-processed image. Due to the fact that the mask tensor occupies a smaller storage space relative to the convolution kernel, some devices with limited storage resources can deploy a neural network comprising the reference convolution kernel and the mask tensor, and then image classification is achieved.

The invention discloses a tension-shear coupling characteristic tester of an asphalt concrete material. The tester comprises a temperature control box, vertical uploading control system for shearing a load, a transverse uploading control system for tensioning or compressing the load, and a coupling characteristic testing centre for controlling the temperature of the temperature control box, the vertical uploading and the transverse uploading and obtaining a tensioning stress and a shearing stress in the horizontal direction according to fitting of a tensioning-compressing displacement and a shearing displacement; a station platform for placing the asphalt concrete material to be tested is arranged in the temperature control box; the vertical uploading control system comprises a vertical oil cylinder, a vertical servo valve, a vertical displacement sensor and a vertical uploading controller; the transverse uploading control system comprises a transverse oil cylinder, a transverse servovalve, a horizontal displacement sensor and a transverse uploading controller; and the vertical uploading controller and the transverse uploading controller both are connected with the coupling characteristic testing centre. The tension-shear coupling characteristic tester of the asphalt concrete material can test the tension-shear coupling characteristic at controllable temperature.

The invention relates to a ghost imaging method and a ghost imaging system based on a bionic vision mechanism, and belongs to the photoelectric imaging field. A laser source, a collimating lens, a rotary frosted glass, and a spectroscope are respectively disposed on the same optical path sequentially. The laser source, the collimating lens, and the rotary frosted glass are used to generate parallel pseu-dothermal light required by the ghost imaging. The spectroscope is used to divide the pseu-dothermal light into two optical paths, and reflection light is a reference arm optical path, and transmission light is a detection arm optical path. The light intensity distribution of the reference arm optical path is received by a bionic detector array, and pseu-dothermal light source two-dimensional light intensity distribution information is completed. After the light of the detection arm is irradiated on a target, the light is reflected, and then the light is reflected by the spectroscope, and the reflection light total light intensity is received by a barrel detector, and the target reflection light total light intensity information acquisition is completed. A related arithmetic unit is used for the operation of the information acquired by the bionic detector array and the barrel detector. The bionic variable resolution detector array is adopted, and large visual field, high resolution imaging and fast imaging are realized at the same time.

A calcaneus plate provided with two vertically-spaced plate sections, with corner-located screw holes. An intermediate wall connects the plate sections. An arcuate protrusion extends from the intermediate wall and contains an internal sliding mechanism for a compression screw. The sliding mechanism includes an arc-shaped compression hole with an elongated opening that includes a ridge upon with the compression screw head can slide.

The invention provides a compression method based on layer-by-layer network binarization. The compression method based on layer-by-layer network binarization includes the steps: constructing a floating point type deep convolutional neural network; according to the opposite sequence of the hierarchy depth of the deep convolutional neural network, performing layer-by-layer binarization on the parameters in the network from deep to shallow until binarizing all the hierarchies in the deep convolutional neural network, and obtaining a binarized deep convolutional neural network; and performing pedestrian detection through the binarized deep convolutional neural network. Therefore, the compression method based on layer-by-layer network binarization realizes compression and acceleration of the network, and can effectively solve the problem of great precision loss caused by network quantification.

The invention relates to a quantification strategy determination method and device of a neural network, an image recognition method and device, computer equipment and a readable storage medium. The method comprises the following steps: acquiring a plurality of preset quantization strategies for a target neural network; wherein each quantization strategy is used for representing the value of the operation attribute of each network layer in the target neural network, and the target neural network has corresponding model parameters under each quantization strategy; and selecting at least one quantization strategy from a plurality of quantization strategies corresponding to the target neural network as a target quantization strategy according to the test precision of the target neural networkfor testing the test sample by adopting the model parameters under different quantization strategies. By adopting the target quantification strategy determined by the method, the problem that the prediction precision of the target neural network is seriously reduced can be avoided while the target neural network is compressed.

The embodiments of the invention provide a data packet processing method and device; the method comprises the following steps: receiving or sending compressed configuration information of an application layer data packet; performing compression processing or decompression processing on the application layer data packet according to the compressed configuration information. By means of the method,compression and decompression processing of the application layer data packet can be realized, the overhead of data packet transmission is reduced, and the utilization rate of network resources is improved.

The invention provides a compensation data processing method for an OLED display panel. By conducting wavelet transform on compensation data, discarding a detail matrix obtained through wavelet transform, only storing an approximate matrix obtained through wavelet transform, then estimating the detail matrix by means of the stored approximate matrix, reconstructing wavelet transform and conducting inverse wavelet transform to obtain compensation data again, compression and decompression of the compensation data can be achieved. System storage space occupied by the compressed compensation data is reduced greatly, so that time for data transmission and burning on a production line is shortened. Furthermore, the compressed data have a fixed compression ratio, the error between the compensation data obtained after decompression and the original compensation data is extremely small, and the display quality of the OLED display panel is not affected.

The invention discloses a manipulator artificial potential field space path planning method combined with a spherical tree model. The manipulator artificial potential field space path planning methodcombined with spherical tree model comprises the following steps of (1) constructing a sphere octree bounding box of a barrier; (2) trimming a hollow sphere, and after clipping the hollow sphere, obtaining a model with a rough shape of the barrier; (3) trimming a redundant sphere, and after trimming the redundant sphere, obtaining a model capable of accurately reflecting the shape of the model; (4) building an artificial potential field model, and calculating the attractive force of a target point, and the repulsive force and the resultant force of the barrier suffered by a robot; (5) judgingwhether the robot is trapped in a local minimum value or a vibration area or not, and if yes, building a virtual target point, and calculating the attractive force of the virtual target point, and therepulsive force and the resultant force of the barrier suffered by a current point; (6) carrying out iterative computation, computing to obtain an optimal solution, and planning an appropriate obstacle avoidance path; and (7) smoothing a track.

The invention discloses a parameter stereo coding, decoding method based on inter-channel transfer function. The parameter stereo coding, decoding method is characterized in that 1) a lower mixed signal can be generated by a coding end using a left track signal and a right track signal; 2) after the respective framing of the input lower mixed signal and the single track signal, the inter-channel transfer function coefficient of every frame can be extracted, and then the extracted transfer function coefficients can be used to form a two-dimensional matrix, and in addition, the single track signal can be the left track signal or the right track signal; 3) the two-dimensional compression, the run coding, and the entropy coding of the two-dimensional matrix can be carried out sequentially, and the coefficients of the two-dimensional matrix can be compressed to form the side information, and then the side information and the lower mixed signal can be used to form the bit stream. The lower high-power space information of the signals can be retained, and the inner-frame compression and the inter-frame compression can be realized at the same time.

The invention provides a multi-bit plane reversible data hiding method for an encrypted image with small cloud storage overhead. The operation is mainly that difference images are obtained by using apixel difference prediction method, the difference images are classified by bit, and information is embedded in a manner of replacing variable bits; the operation of embedding the information is independent of the positions of pixels, so that an image XOR scrambling or a DES encryption operation can be used for the encrypted image, which improves the security of the encrypted image; the pixel difference prediction method effectively compresses an original image, so that a large amount of space is generated by high planes of a part of pixels, and a proper threshold range is found adaptively; abit plane embedding method is adopted for variable pixels within the threshold value, so that multi-bit information can be embedded into one pixel; a sparse matrix compression algorithm is used to compress a pixel classification matrix, which effectively improves the steganographic capacity; and the encrypted image of the corresponding size is generated according to the user's requirements, whichsaves the user's cloud storage overhead and expands the application range of the algorithm.