Abstract: The present invention discloses a scene recognition method, a training method and a device based on pyramid attention, belonging to the field of computer vision. The method includes: pyramid layering a color feature map and a depth feature map respectively, calculating the corresponding attention map of each layer; taking the output of the attention of the last layer as the output; taking the attention output of the last layer as the final feature map, for the remaining layers, adding the result after upsampling of the final feature map of an upper layer to the attention output of this layer as the final feature map of this layer; scaling the attention map and the final feature map, using the average of two new attention maps as the final attention map, mapping the largest k position in the final attention map to the final feature map of this layer.

Abstract: A miniature imaging lens for close-range imaging includes: a first lens group, an aperture, and a second lens group sequentially arranged in a direction from the object side to the image side of an optical axis; the first lens group and the second lens group have positive focal power, an object-side clear aperture of the first lens group is larger than an image-side clear aperture of the first lens group, and an object-side clear aperture of the second lens group is less than an image-side clear aperture of the second lens group, and specific process parameters are provided. A sandwich structure lens configuration composed of a first lens group, an aperture and a second lens group can obtain a high close-range imaging effect under the condition of miniaturization, and can effectively reduce aberrations of close-range imaging, especially distortion and chromatic aberration.

Abstract: The invention discloses a bionic wrist joint based on an asymmetric 3-RRR parallel mechanism, including: an asymmetric 3-RRR parallel mechanism and a drive unit. The asymmetric 3-RRR parallel mechanism includes: a moving platform, a first static platform, and three asymmetrically distributed parallel branch chains, wherein each branch chain includes a passive rod and an active rod. An end of the active rod is connected to the first static platform via the revolute pair, and another end thereof is connected to the passive rod via the revolute pair. The axes of the revolute pairs at two ends of the active rod form an axis included angle. Three axis included angles are different, the passive rod and the moving platform are connected by the revolute pair, and three axis included angles corresponding to the passive rods are different. The drive unit is configured to drive the asymmetric 3-RRR parallel mechanism to move.

Abstract: The invention provides a method and a product for manufacturing a titanium alloy dual-structure turbine disk based on partial hydrogenation, which includes the following steps: coating a glass coating on the partial surface of a titanium alloy billet where hydrogen-blocking is required, and sintering the titanium alloy billet coated with the glass coating; performing hydrogenation treatment on the titanium alloy billet, such that the hydrogen concentration at the hydrogenation-required portion reaches the predetermined level; removing the glass coating from the titanium alloy billet; preheating the titanium alloy billet, and then performing high temperature die forging in the forging dies; performing vacuum dehydrogenation treatment on the forged turbine disk to remove hydrogen element inside the forging, so that the hydrogen content is 0.015 wt. % or less.

Abstract: The present invention belongs to the field of processing residual stress, and discloses a method for calculating processing parameters for residual stress control by parameter inversion. This method comprises: (a) extracting a characteristic index reflecting the residual stress distribution characteristic from a residual stress distribution curve; (b) respectively presetting initial values of processing parameters for residual stress control, calculating an initial value of the characteristic index, and drawing curves of the characteristic index over the respective processing parameters to obtain respective fitted curves; (c) respectively establishing a relation formula between respective characteristic index increment of the processing parameters and the fitting curve; and (d) assigning the values and performing inversion calculation to obtain the required processing parameters.

Abstract: A wide stripe data storage and a constructing, repairing and updating method thereof are provided. In the wide stripe data storage, a stripe is (n, k, r, z) and comprises local groups and a global group comprising a rack for storing global parity blocks; each local group comprises r + 1 c racks, each rack has different storage nodes for storing data blocks or local parity blocks, the total number of the data blocks and the local parity blocks in each rack is c; one local parity block and r data blocks are stored in each local group; the number of the storage node in each rack is greater than or equal to c; c is smaller than or equal to the node fault-tolerant capability f of the wide stripe based on fault-tolerant attribute of the rack; the number of the global parity blocks is f?1; z is the total number of the racks.

Abstract: A supercritical CO2 boiler capable of realizing uniform combustion, corrosion resistance and coking resistance, and a boiler system are provided. The supercritical CO2 boiler includes a main combustion chamber, an upper furnace, a furnace arch and a flue, wherein a cross section of the main combustion chamber is circular or oval, or is of an N-sided shape, where N>4; at least four burner groups are disposed on the main combustion chamber, each group of burner nozzles corresponding to each burner group includes a recirculating air nozzle, a primary air nozzle and a secondary air nozzle; lateral recirculating air nozzles symmetrically distributed are respectively disposed at two sides of the primary air nozzle, the recirculating air nozzle and the lateral recirculating air nozzle are configured to feed recirculating flue gas or a mixed gas of the recirculating flue gas and secondary air into the main combustion chamber.

Abstract: Disclosed are a pulse magnet device based on magnetic flux compression, and a high-flux measurement method. The device includes a diamagnetic block, reinforcing plates, screw rods and a magnet coil. The diamagnetic block and the magnet coil are concentrically arranged in the axial direction; the reinforcing plates are arranged at ends of the magnet coil and the diamagnetic block and are connected by means of the screw rods. The diamagnetic block is used for inducing the induction current opposite the coil current during the discharge of the magnet coil, and for compressing the magnetic field to the area between the diamagnetic block and the magnet coil. The intensity and uniformity of the magnetic field around the magnet coil are improved by means of increasing the magnetic flux density.

Abstract: The present invention discloses a pedestrian trajectory prediction method and system based on a multi-interaction spatio-temporal graph network, which belong to the field of pedestrian trajectory prediction.

Abstract: The disclosure belongs to the technical field related to on-line measurement in manufacture of integrated circuit, which discloses a snapshot type overlay error measuring device and a measuring method thereof. The measuring method includes: the detection light is subjected to polarization and retardation in sequence to obtain measurement spectrum; Fourier analysis is performed on the measurement spectrum to obtain the frequency-domain signal of the measurement spectrum, and sub-channel frequency-domain analysis is performed on the frequency-domain signal to obtain the linear combination of the non-diagonal Mueller matrix elements of the overlay error sample to be tested; the linear combination of the non-diagonal Mueller matrix elements are processed to obtain the overlay error of the overlay sample under test. This disclosure does not need to measure all 16 Mueller matrix elements, the measurement is carried out on only a few non-diagonal Mueller matrix elements which are sensitive to overlay error.

Abstract: The disclosure discloses a three-dimensional (3D) convolution operation device and method based on a 3D phase change memory, which includes a 3D phase change memory, an input control module, a setting module, and an output control module. By using the 3D phase change memory to perform 3D convolution operation, the phase change units on the same bit line constitute a convolution kernel. Based on the multilayer stack structure, the upper and lower electrodes of the 3D phase change memory serve as the information input terminal, and they are convolved after passing through the respective phase change unit arrays, and the result of the convolution operation is superposed on the middle electrode in the form of current, thereby obtaining the sum of the convolution calculation results of the input information of the upper and lower electrodes, such that the 3D convolution operation is completed in one step.

Abstract: The disclosure discloses a reconfigurable hardware acceleration method and system for Gaussian pyramid construction and belongs to the field of hardware accelerator design. The system provided by the disclosure includes a static random access memory (SRAM) bank, a first in first out (FIFO) group, a switch network, a shift register array, an adder tree module, a demultiplexer, a reconfigurable PE array, and a Gaussian difference module. In the disclosure, according to the requirements of different scenarios and different tasks for the system, reconfigurable PE array resources can be configured to realize convolution calculations of different scales. The disclosure includes methods of fast and slow dual clock domain design, dynamic edge padding design, and input image partial sum reusing design.

Abstract: A point cloud reconstruction method and apparatus based on a pyramid transformer, a device, and a medium are provided. The method includes: after obtaining point feature data corresponding to each view, inputting the point feature data into a dual-channel pyramid network; when passing through a first channel, inputting an updated output of an attention mechanism of each layer to a next layer, when passing through a second channel, splicing and inputting the updated output of the attention mechanism of each layer and the updated output of the attention mechanism of a corresponding layer of the first channel in a same layer as the next layer to the next layer; converting the updated output of the attention mechanism of a bottommost layer of the second channel into point cloud coordinates to obtain a point cloud result, and taking an intersection of the point cloud result to characterize a target object.

Abstract: A compression-molding method for a permanent includes: providing a drive coil to generate an electromagnetic force when a transient current is passed into the drive coil, so as to apply a molding compression force to magnetic powder under compression, and providing an orientation coil to generate an orientation magnetic field when a transient current is passed into the orientation coil, thereby providing the magnetic powder under compression with an anisotropic property; and synchronously passing the transient currents to the drive coil and the orientation coil to synchronously generate the electromagnetic force and the orientation magnetic field, thereby completing compression-molding of the permanent magnet, wherein a magnitude of the electromagnetic force and an intensity of the orientation magnetic field are respectively changed by changing peak values of the transient currents.

Abstract: A superlattice phase-change thin film with a low density change, a phase-change memory and a preparation method. The superlattice phase-change thin film includes first phase-change layers (7) and second phase-change layers (8) that are alternately stacked to form a periodic structure; during crystallization, the first phase-change layer (7) has a conventional positive density change, and the second phase-change layer (8) has an abnormal negative density change, therefore, the abnormal density reduction and volume increase of the second phase-change layer (8) during crystallization can be used to offset the volume reduction of the first phase-change layer (7) during crystallization.

Abstract: A multi-degree-of-freedom spherical motor includes a primary frame, a secondary frame, a tertiary frame, a terminal rotor, a first rotating shaft, a second rotating shaft, a third rotating shaft, permanent magnets and electromagnetic coils. The electromagnetic coils are arranged on an inner surface of the primary frame. Two sides at a top portion of the primary frame are symmetrically connected to the secondary frame through the first rotating shaft. The tertiary frame is arranged inside the secondary frame in a non-mechanical contact manner. The terminal rotor and the tertiary frame are connected through the third rotating shaft. The permanent magnets are arranged on one side of the terminal rotor close to the primary frame. The secondary frame and the tertiary frame are connected through the second rotating shaft. The first rotating shaft, the second rotating shaft and the third rotating shaft are perpendicular to each other.

Abstract: A system of electron irradiation includes an electron accelerator and an electron beam focusing device. The electron accelerator emits and accelerates a beam of electrons. The electron beam focusing device is located at a rear end of the electron irradiation and includes a beam restraining rail and 2n+1 sets of magnetic poles. The beam restraining rail forms a beam restraining channel through which the beam of electrons are to pass. The 2n+1 sets of magnetic poles are installed on the beam restraining rail and distributed at different locations of the beam restraining channel. An nth set of magnetic poles thereof are arranged for performing, on the beam of electrons, focusing in a first direction. An (n+1)th set of magnetic poles thereof are arranged for performing, on the beam of electrons, focusing in a second direction. The second direction is perpendicular to the first direction. The n is a positive integer.

Abstract: A method for synthesizing an aliphatic polyester block copolymer regulated by carbon monoxide. The method uses an organic cobalt metal complex as a convertible catalyst and includes: first regulating an anionic ring-opening copolymerization reaction between an aliphatic acid anhydride and an epoxy compound; then using carbon monoxide as a conversion agent for conversion to obtain a catalyst having a new catalytic active site; and regulating a vinyl monomer to perform active free radical polymerization to obtain an aliphatic polyester block copolymer having a controllable structure. The described synthesis method uses a convertible catalyst and a conversion agent, combines two controllable polymerization reactions which have different but compatible mechanisms, and obtains a block copolymer by means of a “one-pot” reaction.

Abstract: The disclosure belongs to the field of microelectronics, and specifically, relates to a method of inducing crystallization of a chalcogenide phase-change material and application thereof. To be specific, a dielectric material is brought into contact with an interface of the chalcogenide phase-change material. The dielectric material is in an octahedral configuration, and the dielectric material provides a crystal nucleus growth center for the crystallization of the chalcogenide phase-change material at the interface between the two, so as to induce the phase-change material to accelerate the crystallization. The method is further applied in a phase-change memory cell. Among all the dielectric material layers in contact with the chalcogenide phase-change material layer, the dielectric material structure of at least one side of the dielectric material layer is an octahedral configuration.

Abstract: The disclosure discloses a Gm-APD array lidar imaging method under strong background noise, comprising following steps: respectively acquiring two sets of cumulative detection data of the Gm-APD array lidar at two different opening times of a range gate of the Gm-APD array lidar under strong background noise; respectively performing a statistic operation on the two sets of cumulative detection data of the Gm-APD array lidar with respect to all pixels, to obtain two cumulative detection result histograms of the Gm-APD array lidar; determining a range interval of the imaging target according to the two cumulative detection result histograms; and acquiring a lidar image by a peak discrimination method in the range interval of the imaging target. The Gm-APD array lidar imaging method according to the present disclosure is capable of improving the laser image quality by eliminating the interference of strong background noise in other range intervals.