89results about How to "Solve the loss" patented technology

The invention discloses a large model knowledge base construction method based on multi-granularity retrieval. The method comprises the steps of obtaining a target text and analyzing a document structure, splitting and reorganizing text data according to three levels of titles, fragments and sentences, and constructing three levels of knowledge of the titles, the fragments and the sentences; using a large language model to construct a generative index as question-level knowledge according to the sentence-level knowledge; using a vector model to generate a first vector representing index keys of fragment-level knowledge and / or sentence-level knowledge, and a second vector representing index keys of title-level knowledge and / or question-level knowledge; and respectively storing the first vector and the second vector into a vector database, and storing the index values and the index types corresponding to the first vector and the second vector into a relational database. According to the method, multi-granularity document knowledge base construction is achieved, the expression modes of knowledge under different granularities are expanded, the accuracy of user query and knowledge matching is improved, and the problem that information is lost or incomplete in an existing method is solved.

A three-dimensional target detection and tracking method fusing multi-sensor information belongs to the technical field of computer vision, and can make up for the defect of a single mode by collecting camera and radar image information and fusing camera images and laser radar point clouds. Multi-target tracking is carried out by using a multi-data coupling method of 2D + 3D data, a remote object is identified through a camera, when the object enters an induction range of the laser radar, the laser radar can carry out more accurate positioning, and position estimation of the target in a 2D image is corrected by using a more accurate 3D position provided by the laser radar. The method is used for solving the problems of inaccurate target identification, target shielding and the like under complex road conditions; a traditional single-mode three-dimensional target detection algorithm cannot effectively fuse the advantages of multi-sensor data and is difficult to accurately identify objects in a distance or under a complex road condition.

The invention provides a helmet standard wearing detection method based on improved YOLOv8, and belongs to the technical field of computer vision target detection. The technical problem that an existing detection algorithm cannot judge whether the safety helmet is worn normatively or not is solved. According to the technical scheme, the method comprises the following steps: S1, collecting sample data; s2, a variable convolution Deform ConvolutionV2 is adopted to replace a traditional convolution, and a variable convolution is adopted to replace the traditional convolution; s3, DCNV2 improvement based on MPCA is carried out; s4, an improved multi-scale feature fusion module; and S5, training the improved YOLOv8 network model by using the helmet standard wearing data set to obtain a trained model. The beneficial effects of the invention are that the method not only helps a manager to effectively supervise the wearing condition of the safety helmet on the working site, but also provides a new idea and example for the target detection technology based on deep learning in the field of safety helmet detection.

The invention discloses a diseased fish detection method and device, and belongs to the technical field of computer vision detection. In order to solve the problems that in the prior art, diseased fish disease detection efficiency is low, accuracy is not high, and real-time detection cannot be achieved, a diseased fish detection model is constructed, and the diseased fish detection model comprises a backbone network, a neck network and a detection head network; the method comprises the following steps: inputting a diseased fish image into a backbone network based on a spatial deep convolutional neural network for feature extraction to obtain a visual feature map, and then inputting the visual feature map into a contrast-enhanced attention transfer structure of a neck network for feature fusion processing to obtain a fused feature map; and the fusion feature map is input into a detection head network, and a sick fish target detection result is obtained through detection of a multi-head prediction small target detection algorithm, so that fish school disease detection is performed in real time, the detection efficiency and detection precision of fish diseases are effectively improved, and the method has relatively high practicability and application universality.

The invention discloses a lithium battery health state and residual life prediction method and system. The method comprises the following steps: preprocessing original charging data, eliminating noise and abnormal values, and obtaining normalized data; and performing structured processing on the normalized data in a differential mode to generate structured data. And in combination with a convolution trunk line module, a channel sparse attention residual module, a space sparse attention residual module and a global attention residual module, extracting a global time sequence feature map from the structured data. The feature maps are fused with historical data of the health state of the lithium battery to form a fused feature map, prediction is performed based on the fused feature map, the health state and the residual life of the lithium battery are obtained, through an optimized difference mode, the information amount of data can be enriched, the model can sensitively capture subtle state changes in the continuous cycle period, and meanwhile, the model can be used for predicting the health state of the lithium battery. Potential time sequence features in the charging data can be effectively extracted through the attention residual module, and the prediction accuracy is improved.

The invention discloses a database construction method and device for nonlinear data based on a large model. The method comprises the steps that a node is selected in an initial graph structure to input request data of a user, a retrieval result of the request data is retrieved and generated through an RAG system, and reply data of the request data is generated through the large model; writing the request data, the reply data and the retrieval result into a new node, determining a nearest node associated with the new node in the existing nodes, and forming a branch path with topicality by taking the nearest node as a father node of the new node; storing the new node in a database; wherein all the branch paths form a tree-shaped structure. According to the method, each dialogue node has a traceable father-son relationship by constructing a database storage mode of a multi-branch tree structure, and a user can branch a new dialogue path from any historical node. And the user can freely select different topic paths for dialogue, so that the dialogue context content of the large model is more accurate.

The invention discloses a rock mineral content detection system and a rock mineral content detection method for geological survey, which are characterized in that after image features and spectral features of rocks and ores are extracted, internal generality of the image features and the spectral features of the rocks and ores is mined by using a data processing technology based on deep learning, and the internal generality of the image features and the spectral features of the rocks and ores is used as a mask; according to the method, unique information in rock and ore image features and spectral features is emphasized, and the general information and the unique information of the rock and ore image features and the spectral features are comprehensively utilized through compensation type interactive fusion, so that accurate identification of rock and ore categories is realized. In this way, the problem of information loss in the multi-modal feature fusion process can be effectively solved, and therefore the accuracy of rock and ore classification and the calculation precision of the mineral content are improved.

The invention discloses a connecting piece welding quality detection method and system, a medium and electronic equipment, and relates to the technical field of industrial automation and machine vision. The method specifically comprises the following steps: controlling a moving mechanism carrying an imaging device to move to a detection position, and obtaining a to-be-detected image containing a connecting sheet; inputting the to-be-detected image into a pre-trained target detection model, and obtaining a prediction bounding box of a connection sheet region; performing size expansion based on the prediction bounding box, and generating a region of interest covering the deformation region of the connecting sheet; in the region of interest, extracting a target feature point coordinate of a connecting sheet and a centroid coordinate of a welding spot by using an image processing algorithm; and calculating a distance value between the coordinates of the target feature point and the coordinates of the mass center of the welding spot, and judging the welding quality of the connecting piece based on the distance value. The invention aims to realize high-precision and high-efficiency welding quality detection of the connecting piece under the conditions of complex background interference and uncertain target form.

The invention discloses an unsupervised multi-source information domain self-adaption method and device for rotating machine migration diagnosis, a medium and a product, and relates to the field of mechanical fault diagnosis. The method comprises the following steps: acquiring multi-source information; multi-source information is fused and compressed into a single fusion sample; constructing an information fusion enhanced domain adaptive self-attention network transfer learning model; and inputting the single fusion sample into the information fusion enhanced domain adaptive self-attention network transfer learning model to obtain a transfer diagnosis result of the rotating machinery. According to the method, the problems of negative migration, information loss, high calculation burden and the like in a known multi-source information domain adaptive method can be solved.

The invention relates to a multi-modal medical image fusion method based on image denoising and illumination enhancement. The method comprises the following steps: inputting a multi-modal medical image into an illumination and contrast enhancement network for illumination intensity correction to generate an enhanced image; adding random noise to the multi-modal medical image to obtain a noise image; inputting the noise image into a denoising and reconstruction network to carry out denoising and reconstruction tasks; using the enhanced image and the multi-modal medical image as labels to jointly supervise and train a denoising and reconstruction network to obtain a feature encoder and a feature decoder with denoising and illumination enhancement functions, using the trained feature encoder to respectively extract features of different modal images to obtain modal feature maps, and using the modal feature maps to obtain the multi-modal medical image; and carrying out feature fusion on the modal feature maps of the different modal images, and reconstructing the fused feature maps by adopting a trained feature decoder to obtain a multi-modal medical image fusion result. The problems that image fusion is interfered by noise, illumination is degraded and fusion efficiency is low are solved.

The invention discloses a textile defect detection method which comprises the following steps: step A, image acquisition: performing real-time and / or timed acquisition on a textile part of a textile needle by virtue of a visual sensing system to acquire images; step B, constructing a feature extraction network: based on a YOLOv5 backbone network, introducing a multi-branch discrete cosine attention mechanism module (MDCA) to form a new backbone network SCNet, and drawing an output feature map of the new backbone network SCNet according to the acquired image; and step C, adopting an MUPool pooling method: a new backbone network SCNet outputs a feature map, important node features are acquired and screened through multi-view map convolution, a cascade structure and a rear-end fusion module, and finally a multi-scale feature map is generated, so that the problems of false detection and missing detection caused by oil stains, black oil needles, bright oil needles and dirty yarns can be solved while relatively high detection precision is ensured.

The invention relates to a brain-computer interface system which comprises a flexible electrode array, a receiving device, a flexible connecting wire and external equipment, wherein the flexible electrode array is implanted into a meninx middle artery in a minimally invasive manner through a superficial temporal artery puncture and a vascular intervention procedure; the receiving device is embedded under the skin of a head and is provided with a signal transmitter; the external equipment is provided with a signal receiver. A flexible electrode array is implanted in an anterior branch or posterior branch of a main branch of a middle meninx artery through a superficial temporal artery intravascular minimally invasive technology, the number of electrodes and the array are arranged according to needs, and a receiving device buried in the subcutaneous part of the head is electrically connected with the flexible electrode array implanted in the middle meninx artery. Craniotomy is not needed, operative complications are avoided, minimally invasive implantation reaches the surface of the cerebral cortex, operation of a doctor is easier, the operation time is shortened, and the comfort level of a patient is improved.

The utility model discloses drying equipment for deep processing of red dates, and relates to the technical field of deep processing of the red dates. The drying equipment for deep processing of the red dates comprises a drying box and a transposition mechanism, an opening is formed in one side of the drying box, the transposition mechanism is arranged on the drying box and comprises a rotating plate, a sealing strip and a placing plate, and the rotating plate is rotationally installed between the top of the inner side of the opening and the bottom of the inner side of the opening; the outer surfaces of the two sides of the rotating plate are each fixedly provided with a set of sealing strips, and the two sets of sealing strips are attached to the inner walls of the two sides of the opening correspondingly. According to the drying equipment for deep processing of the red dates, when the red dates need to be replaced, the red dates can be placed on the placing boxes on the outer side of the drying box in advance, during replacement, the placing boxes on the outer side of the drying box and the placing boxes on the inner side of the drying box can be rapidly subjected to position changing, and discharging is conducted on the outer side of the drying box after position changing; feeding and discharging are conducted in the closed state of the drying box, the exposure time of the drying box can be shortened, and heat loss is reduced.

The utility model discloses a ball hitting machine which comprises a main machine shell, and a plurality of ball hitting main machines are arranged in an inner cavity of the main machine shell. The ball hitting main machine comprises a direct-current rotating motor assembly, the direct-current rotating motor assembly is fixedly connected with a ball hitting machine main shaft, the ball hitting machine main shaft is further fixedly connected with a small ball releasing assembly, the small ball releasing assembly is connected with a small ball part, and the small ball part is always connected with the small ball releasing assembly; the direct-current rotating motor assembly is used for recovering the small ball releasing assembly from the released state to the to-be-released state. The small ball releasing assembly is fixedly connected with a spring assembly, and the spring assembly is used for pushing the small ball releasing assembly to rotate. Rotation of the direct-current rotating motor can enable the small ball releasing assembly to enter the releasing state and then restore to the to-be-released state of the small ball part, and the small balls or the small sticks released in the whole process do not need to be manually picked up and recycled. Therefore, the problems that small balls or sticks and other targets need to be manually picked up and recycled, so that recycling is inconvenient, and the targets are lost due to untimely recycling are solved.

This invention discloses a method for obtaining the wide-angle single-station RCS of a medium target based on a compressed sensing algorithm. The steps are as follows: setting a wide-angle uniform incident plane wave illuminating the medium target; defining RWG basis functions and calculating the excitation vector for each incident angle; performing sparsification and compression on the excitation vectors respectively; generating the PMCHWT equations for the target surface and solving to obtain the compressed induced electromagnetic current matrix; recovering the original induced electromagnetic current matrix from each column of the compressed induced electromagnetic current matrix using the OMP algorithm; and calculating the radar cross section of the medium target based on the induced electromagnetic current vectors. This invention can obtain the RCS result at any angle with fewer PMCHWT equation solutions and combined with the OMP recovery algorithm, reducing the solution time for the wide-angle RCS of the medium target while ensuring accuracy.

This invention relates to the field of intelligent control technology for dredged pipelines, specifically disclosing a low-concentration intelligent diversion and interlocking control method for dredged pipelines. The method includes: responding to real-time operational data reported by each vessel at a preset frequency, calling a global fault mapping model to perform topological analysis on the fault information of a single vessel to construct a fault transmission map; based on the fault transmission map and the real-time operational data, calculating the global risk value of the current collaborative vessel group, and determining the risk level to which the global risk value belongs. This invention, by introducing a hydraulic connectivity masking mechanism, enables the system to dynamically correct the topology of the graph neural network in real time based on valve status and flow difference, solving the problem that static models cannot adapt to sudden changes in pipeline physical connections, and ensuring the accuracy of fault impact range calculation.

The present application relates to a generative adversarial network training method, device, computer equipment and storage medium. The method comprises: obtaining a training sample set; the training sample set comprises a plurality of first sample data and corresponding second sample data; the layer thickness of the reconstructed image corresponding to the second sample data is less than the layer thickness of the reconstructed image corresponding to the first sample data; neural network training is performed based on the training sample set to obtain a cyclic generative adversarial network. The cyclic generative adversarial network trained by this method has a high restoration accuracy. The cyclic generative adversarial network screenshot is used to solve the problem of information loss in the layer thickness direction, and an image with good quality is reconstructed to meet the needs of the analysis task.

The application belongs to the technical field of millimeter wave radar anti-collision, and particularly relates to a correlation method for high-voltage line detection of a millimeter wave radar, aiming at the problem of large change of high-voltage line detection results of adjacent antenna rows in a millimeter wave radar anti-collision mode, the application avoids the problem of high-voltage line information loss by setting a survival period and converting to geographic coordinates, and correlates the high-voltage line detection results of adjacent antenna rows, preventing the same high-voltage line from being repeatedly recorded. The method can effectively solve the problem of millimeter wave radar high-voltage line information loss, improve the reliability of high-voltage line warning, and ensure the safety of helicopter flight.

The present application relates to a kind of forest leaf instance segmentation method and system based on deep learning, method includes: obtaining vegetation image, vegetation image is input into leaf instance segmentation model, obtains leaf instance segmentation prediction result;Leaf instance segmentation model is trained using training set;Training set includes: vegetation original image;Feature extraction and enhancement are carried out using backbone module in leaf instance segmentation model, and adaptive spatial fusion mechanism in progressive feature pyramid network is integrated to dynamically adjust feature weight, generate dynamic fusion feature;Through the dynamic asymmetric spatial perception mechanism built-in in dynamic anomaly regression head module, the corresponding multi-source deformation feature layer of dynamic fusion feature is obtained, and the feature fusion strategy of top-down cascaded decoding module is used to optimize multi-scale feature, obtain multi-source fusion feature layer, further using multi-source fusion feature layer, generate leaf instance segmentation prediction result.The present application solves the problems of data scarcity, poor adaptability and low efficiency.

Embodiments of the present application provide a kind of time sequence encoding method of space, three-dimensional reconstruction method and related product.Time sequence encoding method includes: sequentially projecting N line encoding diagrams to the space to be measured;For each pixel position of the space to be measured, based on the pixel gray value of the same pixel position in the N first shooting image, the time sequence value corresponding to the gray scale change peak value in the gray scale change curve is determined as the folded time sequence value;The M pixel interval is encoded by the preset encoding rule;Based on the encoding rule and the encoding value corresponding to the pixel interval of each pixel position, the folded time sequence value corresponding to each pixel position is unfolded, and the unfolded time sequence value corresponding to each pixel position in the space to be measured is obtained.The embodiments of the present application simultaneously utilize space information and time domain information to time sequence encode each pixel position of the space to be measured, effectively solve the problem of data loss in high reflection or high dynamic area reconstruction.

The utility model discloses a building electrical design multifunctional tool table belongs to building electrical technical field, including tool table body, the front side of tool table body top is provided with the storage groove, and the multiple baffle in storage groove divides the space into independent area, can be according to the accurate zoned storage of tool type, avoids all kinds of tools to mix and pile up, and sliding type tool box is used to store commonly used tool specially, effectively solves the tool loss problem of traditional tool table because of disorder, reduces resource waste, makes tool management more rationality, and the push-pull type switch is realized to the cooperation of sliding block and sliding groove of cover plate, and cooperation buckle assembly can complete locking and unlocking quickly, and tool box can be directly along mesa push-pull and use, shortens the tool access time, reduces work interruption, and the flat mesa can be formed after cover plate closes, further expands operation area, solves the mesa disorder problem of traditional tool table tool pile up, reduces the risk of accidental contact tool.

The invention relates to the field of medical equipment, and discloses surgical equipment and a power failure protection method and device thereof. According to the power failure protection method for the surgical equipment, the surgical equipment is provided with standby energy storage equipment, and the method comprises the steps that under the condition that power failure of the surgical equipment is detected, the standby energy storage equipment is called to execute power supply; performing electric quantity information detection on the standby energy storage equipment to obtain electric quantity detection information, determining power supply available duration of the standby energy storage equipment based on the electric quantity detection information, and generating electric quantity prompt information based on the power supply available duration; under the condition that the residual electric quantity of the standby energy storage equipment meets the data protection condition, preset parts of the surgical equipment are controlled to execute a data write-protection instruction, and to-be-protected data of all the preset parts are written into a preset storage space and stored. Normal power supply of the surgical equipment is guaranteed, data to be protected are stored, and follow-up analysis and research work are facilitated.

The invention discloses a data transmission method and device, equipment and a medium, and belongs to the technical field of data processing, and the method comprises the steps: comparing an enqueue information occupied memory value of an original processor core with a residual space corresponding to the original processor core in a cache management unit; executing data physical enqueueing in response to the fact that the residual space corresponding to the starting processor core in the cache management unit is greater than or equal to the occupied memory value of the enqueueing information; responding to the situation that a data dequeue cache queue space corresponding to the target processor core is not full and any piece of dequeue information to be sent to the target processor core exists in a cache region of the starting processor core; and if yes, storing any one piece of dequeue information of the target processor core to a message queue pool through the target arbiter, and executing data physical dequeue. According to the invention, data transmission can be effectively managed, the data flow is controlled during congestion, and blockage and delay of a communication system are avoided.

The utility model provides an organic waste plasma pyrolysis and cracking integrated device, relates to the technical field of garbage resource utilization, the organic waste plasma pyrolysis and cracking integrated device comprises a box shell, a plasma torch, a pyrolysis mechanism and a feed pipe, a cracking chamber is formed in the box shell, a hydrogen outlet is formed in the box shell, and the plasma torch is arranged in the cracking chamber. The hydrogen outlet is communicated with the cracking chamber; the plasma torch is arranged on the box shell, and the output end of the plasma torch is located in the cracking chamber; the pyrolysis mechanism comprises a pyrolysis chamber arranged in the cracking chamber and an exhaust pipe, one end of the exhaust pipe is communicated with the pyrolysis chamber, the other end of the exhaust pipe protrudes out of the box shell, a sealing plug is connected to the exhaust pipe, the exhaust pipe is further communicated with a gas guide pipe, and the other end of the gas guide pipe is communicated to the plasma torch; one end of the feeding pipe is communicated with the pyrolysis chamber, and the other end is connected with a feeding furnace door. The pyrolysis chamber is arranged in the cracking chamber, the high-temperature environment in the cracking chamber is fully utilized, and the high temperature in the cracking chamber can directly act on the pyrolysis chamber, so that the pyrolysis process is more efficient, the heat loss is reduced, and the energy consumption is reduced.

The application discloses a radio data processing method, device and equipment and a storage medium. The method comprises the following steps: acquiring a radio data set to be processed; wherein the radio data set to be processed has corresponding generation time information, the radio data set to be processed comprises radio data to be processed of a to-be-processed task, and the to-be-processed task comprises a plurality of subtasks executed in a preset order; determining a target memory matched with the radio data set to be processed according to the generation time information, a total number of memories and a pre-established memory management table, and storing the radio data set to be processed in the target memory; calling a processing thread corresponding to each subtask in the plurality of subtasks in the target memory according to the preset order, and processing the radio data to be processed by using the processing thread. The technical scheme of the application improves the calculation speed of radio data.

The invention relates to the technical field of rail transit simulation, in particular to a rail transit simulation scene modeling method for enhancing the sense of reality, and solves the technical problem that the rendering effect is poor in an illumination sudden change scene in the prior art. The method comprises the following steps: constructing environment shielding data of a track line; the environment shielding data comprises a sky opening coefficient and an instrument panel geometric shielding coefficient corresponding to each track position of the track line; performing space-time mapping synthesis on the sky opening coefficient based on the real-time running state data of the train, and determining a real-time environment illumination observation value of the rail position where the train is located; performing illumination change analysis on the rail position of the train based on the real-time environment illumination observation value, and determining the illumination sudden change intensity of the rail position of the train; and performing model brightness rendering compensation on the instrument panel area of the train based on the illumination sudden change intensity and the geometric shielding coefficient of the instrument panel.

The invention discloses an event identification method and system for pig house intelligent monitoring, a camera and a medium. The method comprises the following steps: collecting and sampling pig house video streams; performing target identification on the sampled image to obtain an identification result including categories and position coordinates of pigs, personnel, vehicles and the like; performing label matching on the identification result based on a pre-configured label rule, wherein the matching rule at least comprises a target appearance rule, a region intersection rule, a target continuous appearance rule and a target disappearance rule; matching a pre-configured event rule according to a label matching result, and generating hog house monitoring events such as entry of pigs into a corridor and entry of personnel into an effective area; and intercepting an event associated video clip, monitoring a 4G network signal state, and adaptively selecting to locally store or report the event data and the video clip to a cloud platform. According to the invention, accurate and flexible event identification aiming at a pig house scene is realized, and the reliability of monitoring data in an environment without a wired network or a weak network is guaranteed.

The invention discloses a lightweight in-memory computing target recognition network method and system, and the method comprises the steps: obtaining an infrared image data set of a moving target, and labeling an infrared image in the data set; constructing a Mangban efficient lightweight deep neural network, wherein the Mangban efficient lightweight deep neural network comprises a task identification module; taking the category and the position of the marked moving target as gold standards for calculation output of a task identification module, taking an infrared image data set as training data, inputting the training data into a Mangban efficient lightweight deep neural network for deep learning, and obtaining a Mangban efficient lightweight deep neural network model; converting a weight value of the trained Mangbar efficient lightweight deep neural network model into a conductance value of a memristor in a to-be-deployed memory computing chip; and designing a computing hardware circuit in the memristor memory, and deploying the trained Mangban efficient lightweight deep neural network model on the computing hardware circuit in the memristor memory. According to the invention, the demand on the number of memristors in the memory computing chip is reduced, and the computing speed is improved.

This invention discloses a backdoor attack method, system, and medium based on multi-object semantic coexistence in text-to-image diffusion models. This method utilizes the perspective of multi-object semantic coexistence in text-to-image diffusion models to develop MOBA backdoor attack schemes. First, by constructing a trigger alignment dataset and optimizing backdoor implantation and semantic preservation in parallel, the attack success rate and visual concealment are effectively improved, reducing the impact of semantic corruption on attack effectiveness. Second, during model training, an attention-based decoupling backdoor enhancement mechanism is adopted. By decoupling the attention regions of different objects, the semantic integrity of the input prompt is maintained while the backdoor is activated and the attention distribution is reasonably adjusted to reduce the impact of generation bias on attack concealment. The method of this invention ensures both visual concealment and model performance under benign input while achieving efficient backdoor attacks.

The utility model relates to the technical field of building construction, in particular to a construction fence which comprises a bottom plate, a motor is fixedly connected to the interior of the bottom plate, an output shaft is rotatably connected to the left side face of the motor, and two side plates are fixedly connected to the upper surface of the bottom plate; the upper surface of the bottom plate is fixedly connected with seven protection plates arranged at equal intervals, a threaded rod is rotationally inserted into each protection plate, and the outer circumferential face of each threaded rod is in threaded connection with a protection cone. A driving gear can be driven to rotate in the rotating process of an output shaft, a driven gear is promoted to drive a threaded rod to rotate, then a protection cone is driven to ascend and descend, and the effects that the height of the fence can be adjusted, the protection height of the fence is increased, the fence is prevented from being easily climbed over, and the safety of the fence is improved are achieved; the problems that an existing fence for construction is fixed in structure, low in height and prone to being climbed over, and consequently materials on a construction site are lost are solved.