16results about How to "Stable deployment" patented technology

The invention relates to a multi-agent cooperation-based language model reasoning resource scheduling method and system, and the method achieves the intelligent simulation of multi-view iterative thinking in a complex reasoning task through the construction of a multi-agent system which is clear in division of labor and is provided with a special knowledge base. View limitation and decision deviation of a single model in long-sequence and multi-step reasoning are effectively overcome; depending on the fusion of the general capability of the large language model and the task special knowledge base, the professionality and accuracy of the reasoning result are improved. Meanwhile, the problems of resource waste, redundant calculation and unstable convergence caused by cognitive overload in the cooperation process are solved by combining multi-round dynamic cooperation with a convergence mechanism of cognitive load perception and a dynamic token number limitation and low-confidence branch pruning strategy; therefore, on the premise that the reasoning quality is guaranteed, the utilization efficiency of computing resources is remarkably optimized, peak value occupation is reduced, the overall reasoning time delay is shortened, and reliable technical support is provided for efficient and stable deployment of a large language model in a complex task.

The invention discloses a polarization degree and light intensity image fusion method based on bidirectional cross attention. The method belongs to the technical field of polarization imaging, computational imaging and image processing. The technical problems that in the prior art, dynamic balance between structural detail keeping and polarization saliency enhancing is generally difficult to achieve, and particularly under the condition of a complex background or a low signal-to-noise ratio, texture missing or noise artifacts are prone to occurring in a fusion result are solved. A fusion mechanism capable of establishing a two-way information interaction relationship between a light intensity image and a polarization degree image is provided, and further processing is carried out aiming at a noise problem, so that more stable and more reliable polarization fusion imaging is realized.

The utility model provides a monitoring and early warning device for a debris flow gully in a canyon fall section, and relates to the technical field of natural disaster early warning. The device comprises a supporting column, a connecting frame, a sliding rail, a debris flow monitoring module and a rainfall monitoring module. Wherein the supporting columns are embedded and fixed on the two sides of a debris flow gully; the connecting frame is fixedly connected to the upper end of the supporting column. The sliding rail is fixed to the lower end of the middle of the connecting frame and extends in the same direction as the debris flow gully. The debris flow monitoring module is slidably connected to the sliding rail and can follow the debris flow state in the debris flow gully when the debris flow disaster occurs. And the rainfall monitoring module is fixed at the upstream position of the debris flow gully. According to the scheme, the debris flow monitoring module can move along the channel for monitoring through the sliding track, dynamic data collection is achieved in combination with the rainfall monitoring module, the problem that traditional fixed-point monitoring cannot adapt to complex canyon terrains is solved, and the accuracy and timeliness of early warning are improved.

The present application relates to the technical field of GPU computing optimization, and in particular to a sparse attention computing method, device and medium for GPU, wherein the method realizes the high efficiency of long context reasoning through the geometric perception sparse attention framework of ball hashing, and combines a large-scale parallel hashing optimization algorithm and a load adaptive computing kernel. Compared with the existing sparse attention methods based on heuristics or gradient learning, the present application realizes higher retrieval recall rate, lower preprocessing overhead and efficient hardware adaptation to irregular sparse patterns.

The present application relates to the technical field of computer vision and road engineering, and particularly relates to a pavement crack lightweight segmentation and quantification method based on enhanced feature fusion. The method comprises the following steps: decoding and reading pavement images and performing verification, then performing grid division and cutting, and performing pixel normalization processing on the effective sub-images after cutting; a lightweight crack segmentation network is constructed; the preprocessed standard images are input into the lightweight crack segmentation network, and feature extraction, feature fusion, target detection and pixel segmentation are sequentially completed, and a crack detection frame and a binary segmentation mask are output; the lightweight crack segmentation network is trained, the optimal model is saved according to the iteration optimization of the verification set index; the segmentation mask output by the model is denoised, the crack geometric size is calculated, and the final detection quantification result is output. The method has the advantages that the small-scale crack recognition capability is improved, the crack length and width can be automatically and accurately measured, the crack size measurement error is small, and the method is suitable for large-scale pavement inspection operation.

This utility model discloses a telescopic blast-proof baton, relating to the field of police equipment technology. The telescopic blast-proof baton includes a handle, a switch assembly, a central tube assembly, and a small tube assembly. The switch assembly is located on one side of the handle, and the central tube assembly is located inside the handle. The central tube assembly includes a central tube, a central tube outer wall retaining ring, a central tube narrow damping ring, a central tube O-ring, a central tube end wide damper, and a metal drum. The central tube narrow damping ring employs an arc-shaped contact surface design consistent with the curvature of the central tube outer wall, significantly improving the contact with the central tube outer wall. Combined with optimized hardness selection, it avoids the problem of uneven damping force caused by insufficient contact with the central tube outer wall. When the damping ring hardness matches the contact, it avoids excessive friction due to high hardness and insufficient damping effect due to low hardness, ensuring that law enforcement officers can quickly deploy the baton in emergency situations.

The utility model belongs to the technical field of vehicles, specifically relates to a folding pipe type structure and a vehicle comprising the structure, the folding pipe type structure comprises: a vertical pipe, a folding pipe, a rotating mechanism and a locking mechanism, wherein the vertical pipe is rotatably connected with the folding pipe through the rotating mechanism, the locking mechanism locks the vertical pipe and the folding pipe when the folding pipe rotates to the folding position, and the locking mechanism releases the folding pipe through the compression of the spring when the folding pipe rotates away from the vertical pipe under the force. The mechanism provided by the utility model is simple and exquisite, the elastic reversible buckle structure is used to realize the stable locking structure of folding and unfolding, the appearance change of the vehicle body is minimized, and the utility model has a wide range of application and is less modified for the existing vehicles.

The invention discloses a fish body length measuring method and system. The method is realized based on geometric measurement and statistical model fusion. According to the method, stable and reliable fish body length measurement under the conditions of complex fish body postures and segmentation errors is realized by constructing a two-channel length estimation method of a geometric measurement and area statistics model and combining a shape quality index-based adaptive fusion mechanism. Specifically, geometric measurement length is obtained through fish body contour extraction and skeleton analysis, a statistical length estimation model is constructed based on fish body area and shape features, and area estimation length is obtained. In order to improve the stability of the system in a complex scene, a self-adaptive fusion mechanism based on a shape quality index is further constructed, indexes such as fish body contour integrity, contour smoothness and a region filling rate are evaluated, and a measurement credibility weight is calculated; and carrying out adaptive fusion on the geometric measurement result and the area estimation result according to the weight to obtain a final fish body length estimation value.

The application discloses a multi-fingered hand operation reinforcement learning method based on a human motion prediction model and belongs to the field of humanoid robot dexterous hand body operation. A human motion prediction model is trained by using data with human hand and object three-dimensional pose annotation, which is used to generate three-dimensional trajectories of human hand and object interaction in a generalizable manner. A general reward function is constructed by using the trajectory, which is used to train a dexterous hand reinforcement learning strategy in simulation. Then, the migration ability of the strategy from simulation to reality is improved through simulation parameter domain randomization and curriculum learning, and the trained reinforcement learning strategy is deployed on a real-world robot. The scheme introduces the understanding of the future pose of the object to be operated, predicts the interaction process of the human hand and the object as a whole, and constructs a unified, simple, efficient reward function based on this, which is suitable for different operation tasks and different types of dexterous hands, and has the advantages of strong generalization, high success rate and strong reward function universality.

The utility model provides an easy to place and recycle's self -balancing settlement formula seabed base observation system belongs to marine observation technical field. Including integral type recovery part and disposable chassis part, integral type recovery part includes base, four side prismatic shell, flow guide hole, self -floating type universal instrument cabin etc., disposable chassis part includes combined steel frame, anti -sinking plate, barbed foot etc. The utility model discloses utilize flow guide hole to realize self -sinking type placement, utilize integral type recovery part buoyancy point to be higher than gravity point, and the self -floating sea surface is floated, and disposable chassis need not consider recovery simultaneously, and the operation process of offshore placement and recovery is simplified, and the consumption of manpower and material resources is reduced, and self -floating type universal instrument cabin can still make the equipment detection vertical direction sea surface in the placement state of not more than 45 inclination angle on the sea bottom, and the combined chassis foot is high, and the anti -sinking plate and barbed foot are used, so that the sea current is not blocked in the seabed base chassis empty space, and the seabed base siltation and the bottom trawl salvage overturning are prevented.

The application discloses a verifiable privacy search method and system, a server maps an original database into a lattice matrix and calculates an auxiliary index matrix, a client generates a hash digest of the auxiliary index matrix and strips the hash digest to the server for depositing; the client constructs privacy query ciphertext based on a linear privacy search mechanism of a lattice difficulty problem; the server calculates a homomorphic response vector and constructs a proof tensor; the client generates a challenge vector, and the server generates a compressed proof according to the proof tensor; the client performs consistency verification, and if the verification is passed, the client restores target data through a stripping signal, otherwise, the client triggers an alarm; and the client inversely maps the target data and delivers the target data to an application layer. The system is deployed in a client-server architecture and supports coexistence of stateful and stateless clients. The application realizes privacy search and response verification, is lightweight and suitable for resource-limited devices, can be applied to various structured database scenes, and has high security and implementability.

The application provides an atomizer and an electronic atomization device. The atomizer comprises a shell, a sealing element and an atomization assembly. The shell has opposite first and second ends and a side wall, and the side wall is provided with a connecting port. The sealing element is arranged in the shell and comprises a sealing body and a supporting part. The sealing body is provided with a first through hole, and the supporting part is arranged between the sealing body and the second end. The atomization assembly is arranged between the sealing body and the first end and comprises an atomization tube and an electric heating element arranged in the atomization tube. The supporting part is provided with a connecting channel, and the atomization tube, the first through hole, the connecting channel and the connecting port are sequentially communicated. The connecting channel intersects with the extension direction of the first through hole, so that the lead wire of the electric heating element can be led out of the atomization tube to the connecting port for electrical connection with a power supply device. By integrating the sealing body and the supporting part provided with the connecting channel in the sealing element, stable arrangement and reliable connection of the lead wire are realized while good sealing performance is ensured, and the safety and stability of the atomizer are improved.

The application discloses a large-span photovoltaic support arranged on a steep collapsible loess mountain slope, and solves the problem of how to construct a support pile foundation system suitable for the collapsible loess mountain slope; a slope bottom PVC corrugated pipe pile is arranged at the slope bottom of the collapsible loess mountain slope (1), and a truss is hoisted on the slope bottom corrugated pipe pile, a middle pile foundation and the slope top corrugated pipe pile; the lower end of the truss is connected to a first triangular connecting piece (9), the upper end of the truss is connected to a second triangular connecting piece (16), and a middle triangular connecting piece (22) is connected to a large-span photovoltaic support body (24) through two clamping plates (23); X-shaped arrangement of a cable-stayed anchor cable is arranged between two adjacent trusses, and an inclined cable-stayed anchor cable is arranged between each pile foundation and the top end of the truss opposite to the foundation; and the X-shaped arrangement of the cable-stayed anchor cable and the inclined cable-stayed anchor cable enable the whole large-width support to form a stable box frame structure.

The present application relates to a computer vision-based real-time evaluation method for PCB hole wall copper plating uniformity. The core is to use a multi-angle, multi-polarization state pulse ring LED array light source and a high-resolution camera to cooperate, collect and preprocess the hole wall reflected light intensity sequence, combine with known parameters such as Young's modulus, establish a fine response model representing the correlation between hole wall micro-deformation and reflectivity, and generate a three-dimensional local physical fingerprint accordingly, realize automatic segmentation and texture feature extraction in nano, sub-micron and micron multi-scale sensitive domain. Through graph convolution embedding and time series adaptive regression mapping, a dynamic mapping of texture-process parameters is established, the process response delay is compensated, and adaptive optimization of copper plating process fluctuation is effectively realized. The application has the beneficial effects of high microstate detection precision, fast process response and strong model adaptive ability.

The utility model discloses a patient position adjusting support, which belongs to the technical field of sickbed improvement, and comprises a main support, the top end of the main support is connected with an adjusting frame, two sides of the main support are connected with auxiliary support frames, the adjusting frame comprises a cantilever frame, one end of the cantilever frame is provided with a counterweight groove, and the other end of the cantilever frame is provided with an auxiliary support frame. The bottom end of the cantilever frame is connected with a servo motor, and the other end of the cantilever frame is connected with a telescopic frame. According to the body position adjusting support, telescopic adjustment is achieved through the telescopic frame connected with the cantilever frame, the rotation angle can be adjusted in cooperation with the servo motor of the cantilever frame body, after a certain limb of a patient is carried by the hanging frame, comfort can be provided while stable clamping is kept in cooperation with the air bag cushion, and vertical adjustment is achieved through connection of the winding frame; a rotation angle and a lifting and telescopic adjusting process are formed, so that a patient is assisted to move and adjust a certain limb, the legs of the patient can be conveniently lifted for a long time, and medicine application or examination is facilitated.

The present application relates to the field of ship collision avoidance control, and particularly relates to a multi-ship cooperative collision avoidance method based on multi-agent reinforcement learning and imitation learning, comprising: building a multi-agent deep reinforcement learning architecture based on a deep Q network, the deep Q network comprising a decision network and an evaluation network, and simultaneously modeling a multi-intelligent ship navigation environment; training the reinforcement learning architecture according to the navigation environment and using a training set to obtain a trained reinforcement learning architecture, the training comprising imitation learning and reinforcement learning, the imitation learning comprising generating an action space of a demonstration state based on an expert strategy; the reinforcement learning comprising generating an action space of an exploration state by each intelligent ship based on the reinforcement learning architecture; in a collision avoidance decision phase, the intelligent ship generates ship motion parameters by using the trained reinforcement learning architecture based on observation information, and generates and executes a collision avoidance action based on the ship motion parameters. The present application realizes implicit cooperation between multi-ships under the condition of distributed execution without explicit communication.