2057results about How to "Improve integrity" patented technology

The invention discloses a fiber foamed concrete and a preparation method thereof. The fiber foamed concrete comprises the following raw materials in parts by weight: 5290 parts of cement, 4965 parts of water, 1.361 part of foaming solution, 1.2964 part of fly ash and 10 parts of fiber. The preparation method comprises the following steps: proportionally weighing the components; sequentially adding the cement and fly ash into a stirrer, proportionally weighing the water, uniformly stirring the mixture by the stirrer, and adding the fiber; preparing and adding the foaming solution; gradually adding the foam into the mixture, and continuously stirring until all the components are uniformly mixed; pouring the mixture into a mold on which lubricating oil is brushed, and curing in a curing chamber; after 24 hours, removing the mold; and respectively curing for 3 days, 7 days and 28 days for experimental detection. The fiber foamed concrete has the advantages of favorable thermal insulation property, favorable compression resistance and favorable bending resistance, and is applicable to a railway bed strengthening layer.

The invention relates to a low-loss corner-lifting and movement tearing cooperative type film tearing process of a screen display glass protective film. The process comprises the following steps: S1, tearing a film on an A surface; s2, transferring the screen display glass; and S3, tearing the film on the surface B. On one hand, in single-face film tearing, based on cooperation of the film tearing clamping jaw, the rolling wheel and the air nozzle, angle lifting is completed while the film tearing angle adhesive force is kept larger than the protective film tearing force, the damage and fracture rate of an adhesive tape and a corner is reduced, the high angle lifting success rate is achieved, meanwhile, through movement cooperation of the screen display glass and the film tearing clamping jaw, the film tearing difficulty is low, control is easy, and the film tearing efficiency is improved. The integrity and continuity of film tearing are high; on the other hand, overturning of the screen display glass is not needed in double-sided film tearing, transferring of the screen display glass is achieved in the protective film tearing process, the accidental friction rate in the glass overturning process is reduced, the structure is optimized, the layout space is greatly reduced, the equipment is miniaturized, and the double-sided film tearing efficiency of the equipment for treating the screen display glass is also improved.

The invention relates to a protective film uncovering process based on corner lifting and rolling cooperation type screen display glass. The protective film uncovering process comprises the following steps: S1, corner lifting; s2, tearing off; and S3, collecting the film. On one hand, through cooperation of the film tearing clamping jaw and the rolling wheel, step-by-step peeling with the rolling position and the peeling position aligned is kept, meanwhile, the angle between the connecting section and the screen display glass is gradually increased on the basis that the connecting section is turned inwards under the condition that the adhesion section is not stressed, the success rate of forming an inwards-turned peeling angle is greatly increased, and the production efficiency is improved. The probability that the film-tearing adhesive tape is separated from the protective film or the corner of the protective film is torn is reduced; on the other hand, the complete tearing rate of the protective film is increased based on the relative movement of the glass and the film tearing clamping jaw, the tearing risk of the protective film is reduced, more labor-saving and faster diagonal tearing is achieved, meanwhile, based on pulling formed by the film collecting clamping jaw and the film tearing clamping jaw and airflow blowing cooperation formed by the film collecting clamping jaw and the film tearing clamping jaw, the completeness of film collecting is improved, and the production efficiency is improved. And the problem that the film tearing adhesive tape is adhered to the film tearing clamping jaw is solved, so that the continuity of peeling off the protective film is improved.

The invention provides a new energy construction small target full-link safety management and control method based on cross-view feature collaboration and a dynamic anti-interference closed loop, and relates to the field of new energy construction safety management, and the method comprises the steps: S1, deploying a plurality of heterogeneous sensing devices, and collecting the multi-source data of a small target in a construction region; s2, unifying small target features of different visual angles to a global coordinate system of a construction area through coordinate mapping and feature decomposition fusion according to the multi-source data collected in the step S1, and outputting an aligned small target feature map; and S3, adaptively performing multi-scale feature enhancement and lightweight processing according to the aligned small target feature map output in the step S2 and the scale and texture characteristics of the small target, and outputting an initial recognition result of the small target. Small construction targets such as bolts and buckles can be accurately identified, the cross-view space-time dislocation problem of multi-source equipment is solved, dynamic illumination and equipment vibration interference is inhibited, and full-link management and control from data acquisition, feature enhancement and precision calibration to rectification closed loop is realized.

The invention relates to a gas nano purifier. The gas nano purifier comprises a mounting seat, the gas storage bin is installed on the installation base and used for storing clean gas, the side wall of the upper portion of the gas storage bin extends upwards to form an installation plate, and a gas path through hole is formed in the middle of the installation plate; the gas-water separator is mounted on the mounting plate and is used for carrying out primary purification on to-be-cleaned gas; the adsorption tower is mounted on the mounting plate and is used for carrying out secondary purification on the to-be-cleaned gas; gas to be cleaned sequentially flows through the gas-water separator and the adsorption tower and then enters the gas storage bin; the adsorption tower is communicated with the gas storage bin through an exhaust pipe; and the exhaust pipe penetrates through the gas path through hole. Through the arrangement, the installation operation is greatly simplified, and the integrity and the space utilization rate of equipment are improved.

The invention provides an RNA polymerase variant and application thereof, and relates to the technical field of biology. Wild type T7 RNA polymerase is modified to obtain the T7 RNA polymerase variant, the enzyme specific activity of the T7 RNA polymerase variant is remarkably improved compared with that of the wild type T7 RNA polymerase, the addition amount of the polymerase in the in-vitro transcription reaction process can be reduced, and the cost is saved; in addition, the mutant also effectively reduces the generation of incomplete fragment impurities, and the mRNA integrity is significantly improved to 99%. The invention further provides a method for preparing RNA through in-vitro transcription.

The invention discloses an industrial measurement data intelligent analysis report generation method and system based on a large language model, and relates to the technical field of industrial measurement, and the method comprises a multi-source data and abstraction module which carries out the structural processing of original data DAT output by measurement analysis software; and the large language model service and interface module is used for submitting the PCT to a selected large language model LLM through a multi-model adaptation interface to execute semantic reasoning and output a structured text TXT. Structured analysis and semantic association modeling are carried out on original measurement data through multi-source data and an abstract module, when overall alignment deviation or local feature anomaly exists in the original data, a hierarchical dependency relationship between the data can be established through a semantic graph structure, a deviation transmission path is revealed, and the accuracy of the measurement data is improved. The large language model is analyzed in a unified data context, and the utilization depth of measurement data can be effectively improved, so that the accuracy and integrity of report analysis are improved.

The invention relates to the field of electrical power grid operation and maintenance, in particular to a differentiated operation and maintenance method and system based on real-time weak links of equipment. According to the method, online monitoring data and power supply reliability operation data of power distribution network equipment are collected in real time to form an equipment state multi-source data set; based on the data set, calculating an equipment health degree evaluation result by adopting a state evaluation model, and obtaining an importance degree evaluation result through an analytic hierarchy process in combination with the position and load importance of the equipment in the network topology; and inputting the two results into a preset operation and maintenance strategy matrix to match a corresponding operation and maintenance strategy type, and finally generating a differentiated operation and maintenance strategy instruction containing specific operation and maintenance measures, an execution time window and a resource allocation scheme. According to the method and the device, the problems of lack of equipment-level pertinence and unreasonable resource allocation of an operation and maintenance strategy in the prior art are effectively solved, and the operation and maintenance accuracy and the resource utilization efficiency are improved.

The invention belongs to the field of thin films, and particularly discloses an atomic layer deposition method and a thin film prepared through the atomic layer deposition method. According to the deposition method, the reaction cavity, the two input pipeline systems, the outlet valve and the vacuum pump in the atomic layer deposition system are used for controlling precursor pulse deposition and inert gas purging conditions in the atomic deposition process, and the film deposition state is improved; the efficiency of purging and removing substances and impurity residues such as a precursor and a reaction product thereof is improved, the problem that the quality of a deposited film is affected by the impurity residues in a reaction system or on the surface of a sample in the film deposition process is avoided, the uniformity and compactness of film deposition and the coverage rate and conformal rate of a film on the surface of a base material hole are improved, the defects of the film are reduced, and the yield is improved. And the method can be applied to a substrate with a high aspect ratio structure to deposit the film, so that the purpose of manufacturing a large-size complicated special-shaped structure device is achieved.

The invention relates to a mobile robot dynamic obstacle real-time detection method and system based on embedded vision. The method comprises the following steps: performing motion compensation on a historical frame image, generating a motion compensation image, performing frame difference analysis on the motion compensation image and a current frame image, performing adaptive threshold segmentation in combination with local texture features, and generating a motion candidate mask; and extracting feature points in the mask region, calculating an optical flow vector and generating RGB image blocks, and performing dynamic attribute verification after fusion to obtain a verified dynamic point set. And further calculating an optical flow direction angle and a fusion speed value, constructing a motion feature vector and clustering, and generating a central position coordinate of the obstacle instance. And through rotation matrix conversion and Kalman filtering processing, a smooth position estimation value and a smooth speed estimation value are obtained, and the relative distance and the dynamic threat degree of the obstacle are calculated. According to the method, through motion compensation, filtering optimization and the like, the real-time performance, precision and motion estimation accuracy of dynamic obstacle detection in an embedded scene are improved.

The invention provides a laser stripe center extraction method based on gray coefficient binarization, and aims to solve the problems of non-uniform brightness distribution of laser stripes, difficulty in extraction of weak stripes, breakage of center lines and the like. The method comprises the following steps: enhancing an image; realizing self-adaptive threshold segmentation through a local gray scale statistical model and a peak variable coefficient by using a gray scale coefficient binarization method; extracting a sub-pixel-level center point by combining a secondary positioning method of skeleton constraint; and carrying out fitting smoothing by adopting bicubic interpolation and a smooth spline method. According to the method, the problem of extraction of strong and weak stripes under complex working conditions is effectively solved, and the precision, connectivity and robustness of stripe center positioning are remarkably improved while background noise is suppressed.

The invention relates to the technical field of image processing, in particular to a lens edge detection method and system based on image generation, and the method comprises the following steps: collecting a lens image through an industrial camera, carrying out the weighted graying, extracting an edge coordinate through a Laplace operator, constructing an edge pixel coordinate set, and carrying out the image processing; screening an edge validity marking result by combining a gray difference threshold, calculating a phase difference between a gradient direction and light source incidence based on a Sobel operator, extracting pixels in a consistent direction, carrying out spatial clustering, generating a lens edge pixel cluster, and fitting a continuous curve to construct a complete lens edge contour. Noise and reflection interference are restrained through pixel neighborhood gray difference, effective pixels are screened in combination with a gradient direction and light source incident phase relation, edge direction consistency is enhanced, pixels are clustered according to spatial continuity and gradient intensity, contours are continuously fitted based on pixel cluster distribution, and lens edge integrity and geometric consistency under complex illumination are improved.

The invention discloses a time sequence generation method and system based on uncertainty quantization and frequency domain constraint, and belongs to the technical field of artificial intelligence and data mining. Outputting a mean value of the completion value and a variance representing the uncertainty of the completion result; an entropy weight adaptive mask generation mechanism is constructed, a soft mask matrix of continuous values is generated according to variance and local observation density, and high uncertainty corresponds to low mask weight; constructing a mixed loss function containing time domain weighted loss and frequency domain consistency loss; and training a diffusion model based on the soft mask matrix and a mixed loss function to generate a regularized high-fidelity time sequence. According to the method, the Bayesian uncertainty quantization and soft mask mechanism is introduced, so that the problems that a binary mask mechanism in the prior art is rigid and cannot sense completion errors are solved, and meanwhile, the periodic characteristics of generated data are ensured in combination with frequency domain constraints.

The invention relates to a road infrastructure full-automatic inspection method, device and equipment based on an unmanned aerial vehicle group, and a medium thereof. The method aims at the problem that disease characteristics collected by unmanned aerial vehicles in traffic flow and vegetation shielding scenes are broken, and the method comprises the following steps of: decoupling texture and motion characteristics of a dynamic shielding area; feature maps of the effective area and the shielding area are separated; performing cross-frame splicing and continuous repair on the fracture features by using an optical flow constraint generation network, and reconstructing the geometric integrity of a shielded region; infrared data are fused to correct thermal deformation errors, and a high-precision continuous curved surface model is generated; and finally, based on a disease quantitative feature vector matching maintenance strategy, realizing full-process automation from data acquisition to decision output. According to the method, the continuous modeling precision and the maintenance decision reliability of pavement disease detection in the shielded environment are remarkably improved.

The invention belongs to the technical field of semiconductor testing, and discloses a thermistor performance monitoring method and system based on a power module reliability test. The method comprises the following steps: configuring a pin connection structure electrically connected with a thermistor lead on a test burn-in board for a high-temperature, high-humidity and high-voltage reverse bias test; in the process of executing a high-temperature, high-humidity and high-voltage reverse bias test on the power module comprising the thermistor, a lead of the thermistor is electrically connected with a preset current detection loop through the pin connection structure, so that monitoring current flowing through the thermistor is received by the current detection loop, and thermistor performance monitoring is executed. According to the application, the problem that the self degradation risk of the thermistor cannot be monitored in real time in the H3TRB test process to cause a monitoring blind area can be solved, the whole-process monitoring of the performance fluctuation of the thermistor in the strong electric field and high humidity coupling environment is realized, and the integrity of the reliability evaluation of the power module and the safety of the test system are remarkably improved.

The invention relates to the technical field of ship safety, and discloses a ship safety management monitoring system based on the Internet and a Beidou satellite, the system comprises a coordinate analysis module, a ship state module, a data fusion module, an abnormity identification module and an early warning message sending module, a satellite positioning signal is analyzed, and real-time latitude and longitude coordinates of a ship are obtained; generating comprehensive operation parameters of the ship according to the attitude, the navigational speed, the load and the main engine operation parameters of the ship; performing data fusion on the longitude and latitude coordinates after timestamp alignment and the comprehensive operation parameters to obtain an operation feature vector of the ship; performing abnormal mode identification on the operation feature vector based on historical navigation data and a preset safety threshold to obtain a risk category of the ship; when a risk category occurs, generating a treatment measure message of the risk category, and sending the treatment measure message to a ship-end display terminal and a shore-based management terminal through communication between the Internet and a Beidou short message; the ship safety management efficiency can be improved.

The invention discloses an underground engineering structure total space deformation analysis method based on a multi-depth camera array, and relates to the field of space deformation, and the method comprises the steps: employing the multi-depth camera array to collect a depth image and a color image of an underground engineering structure, and generating an alignment depth image of each depth camera; performing internal reference calibration on the depth camera, and converting an aligned depth image of the depth camera into three-dimensional color point cloud data of the depth camera; performing external parameter calibration on each depth camera to obtain an optimized transformation matrix from each slave camera to the master camera; transforming the three-dimensional color point cloud data of each subordinate camera to a coordinate system of the main camera by adopting an optimized transformation matrix to obtain a three-dimensional model of the underground engineering structure; and determining corresponding point pairs in the three-dimensional model obtained in different periods, and calculating the Euclidean distance between the corresponding point pairs to realize the full-space deformation analysis of the underground engineering structure. According to the invention, low-cost and high-precision underground engineering deformation monitoring is realized.

An image debugging system relates to the technical field of image processing. The image debugging system comprises a test platform, a proxy server and an image processing unit. The image processing unit comprises an image processing assembly line, the image processing assembly line comprises a plurality of image processing modules, the image processing unit is connected with at least one image sensor, and the image sensor is used for collecting optical signals and outputting image data to the image processing assembly line. Image processing is carried out by the corresponding image processing module; the test platform is used for generating an image interface and generating a control instruction according to a set protocol data packet, the proxy server is used for receiving, analyzing and forwarding the control instruction to the image processing unit, and the image processing unit responds to the control instruction to configure a corresponding image processing module in the image processing assembly line, or sends the control instruction to the image processing assembly line. And the image processing unit responds to the control instruction to obtain image data processed by the corresponding image processing module in the image processing assembly line.

The utility model discloses a permanent and temporary combined assembly type underground structure plate wall connecting node which comprises an assembly type underground diaphragm wall, a full-length steel plate is installed on the side wall of the assembly type underground diaphragm wall, and a prefabricated structure plate is hoisted on the full-length steel plate. The gap between the prefabricated structural slab and the fabricated underground diaphragm wall is filled with ultra-high-performance concrete, and a steel bar structure is arranged in the ultra-high-performance concrete. According to the utility model, the rigid connection between the two prefabricated parts is realized, the integrity of the structure is improved, and the safety and durability of the underground structure are fully ensured; an underground diaphragm wall which is usually used as a temporary structure is changed into a permanent structure, so that the construction period is shortened, and the investment is saved; the ultra-high performance concrete poured in the joint can greatly shorten the anchoring length of the steel bars, enlarge the arrangement distance of the steel bars, obviously reduce the using amount of the steel bars on site and the binding workload, is convenient to install and simple in process, obviously improves the construction efficiency, and has high popularization and application value.

The invention discloses an enhanced generation query method and system based on multi-granularity retrieval and reinforcement learning strategies. According to the method, semantic analysis is carried out on an original query by utilizing a large language model, the original query is decomposed into a plurality of sub-problems, and each sub-problem generates a query vector; in the multi-granularity vector knowledge base, respectively executing vector similarity retrieval on the query vector of each sub-problem to obtain a related document fragment vector of each sub-problem under each granularity, and further generating a multi-granularity vector set of each sub-problem; weighting the multi-granularity vector set of each sub-problem to obtain a comprehensive context information block, determining the weight of the multi-granularity vector set of each sub-problem through a multi-arm reinforcement learning strategy, and inputting the comprehensive context information block into a large language model to output a query result. According to the method, the retrieval accuracy is improved through semantic decomposition and multi-granularity knowledge collaborative retrieval, results are intelligently integrated through a self-learning mechanism, and lightweight self-adaptive optimization is achieved.

The invention discloses an AI substitute inspection method and system based on multi-modal perception, and relates to the technical field of intelligent detection. According to the method, images, audios, environmental parameters and equipment state data of an inspection area are synchronously collected through a multi-modal sensor array, and after space-time alignment preprocessing, multi-modal features are extracted and input into a pre-trained fusion model to generate unified feature representation. The fusion model adopts a Transform architecture to realize cross-modal feature interaction, an anomaly detection algorithm is used to identify and classify anomalies, a confidence coefficient is calculated in combination with an evidence theory, and finally an inspection report containing anomaly positions, types and processing suggestions is generated. According to the invention, the information limitation of single-mode inspection is solved, the detection accuracy and robustness in a complex environment are improved, unmanned inspection in a high-risk scene is realized, the labor cost is reduced, the safety is guaranteed, and the method is suitable for automatic inspection in the fields of industrial equipment, public facilities and the like.

The invention discloses a method for measuring and planning deformation in additive and subtractive alternate mixed manufacturing of metal cavity parts. According to the additive and subtractive mixed manufacturing technology, by combining the design freedom degree of additive manufacturing and the high precision advantage of subtractive manufacturing, complex cavity parts can be effectively machined, but accumulative stepped deformation can be caused by the repeated thermodynamic effect and alternate procedures; according to a traditional optical measurement method, measurement data are difficult to completely capture due to the problems of internal shielding and surface reflection of a cavity, and point cloud registration errors exist. According to the method, multi-axis on-machine measurement is embedded into an additive and subtractive alternate process sequence, a measurement point directly comes from a cutter contact of a cutting path, the multi-axis on-machine measurement precision and the measurement accessibility of a deep cavity structure are ensured, and a measurement path is optimized by combining probe attitude planning and a customized traveling salesman algorithm; the probe and the cutter share the same coordinate system, so that the point cloud registration link required by optical measurement is eliminated, and the data precision is remarkably improved.

The invention discloses a pavement crack accurate extraction method and system based on multi-scale image segmentation, and belongs to the technical field of pavement detection, and the method comprises the steps: obtaining a to-be-detected pavement image, carrying out the brightness normalization and geometric correction of the pavement image, and constructing a multi-scale image set with different resolutions; generating a spatial texture channel and a spectral domain response channel for each scale image in the multi-scale image set, and fusing the spatial texture channel and the spectral domain response channel to obtain a spectral-space coupling input tensor, wherein the spectral domain response channel is obtained by extracting energy characteristics of a plurality of frequency bands after performing spectrum transformation on the scale image; a multi-scale image set is constructed for road surface images, and coupling expression of spatial texture information and spectral domain response information is introduced, so that feature consistency of fine cracks under different resolutions is enhanced, and scale drift and response distortion caused by shadows, light reflection and seam textures are inhibited at the same time; therefore, the separability and the stability of the crack in a complex scene are improved.

The invention relates to a self-supervised multi-modal fusion and collaborative optimization method suitable for a curve and ramp scene. The method comprises the following steps: acquiring an original perception image; generating four paths of geometrically consistent recovery images; constructing a multi-modal fusion model, and generating a depth confidence map aligned with the restored image; uniform ground plane parameters are obtained; collecting all coordinate points to construct an initial 3D lane candidate point set; generating a high-density locally enhanced BEV lane representation; and outputting the fused global consistency 3D lane map. And outputting the structured 3D lane line marking data. Non-planar structures such as curve superelevation and longitudinal ramps can be accurately depicted, and geometric consistency in a complex road scene is greatly improved; the problems of single vehicle shielding, sensor noise and the like are effectively relieved, the precision, the integrity and the system-level reliability of 3D lane reconstruction are remarkably improved, meanwhile, dependence on a high-precision map or 3D true value marking is avoided, and the method has high engineering landing value.

The invention discloses an image fusion method based on a double-branch feature decoupling auto-encoder, and belongs to the field of computer vision and image processing. According to the method, for the problems of modal pollution and structural distortion in infrared and visible light image fusion, structural semantic information and high-frequency texture details of a source image are extracted respectively by constructing a content feature coding module and a detail feature coding module, and feature decoupling is achieved. And performing deep fusion on the decoupled features by using an adaptive feature weighting mechanism, and reconstructing a fused image with infrared target saliency and visible light detail definition through a shared decoder. According to the method, an end-to-end two-stage training strategy is adopted for optimization, complex prior or post-processing is not needed, the effects of improving the fused image contrast, edge preservation and target recognition performance are remarkable, and the method is suitable for the fields of weak light monitoring, intelligent perception, unmanned system navigation and the like.

The invention provides a pipeline welding real-time feedback control method and system based on machine vision. According to the method, multi-modal data such as a three-dimensional point cloud, a temperature field and a welding seam image are synchronously collected, and key characteristics such as a pose deviation gradient, a cooling rate curve and a welding seam size sequence are extracted after time sequence alignment, filtering and correction fusion. And a CNN and LSTM mixed adaptive deep learning model is utilized to analyze the welding quality state in real time, and sub-optimal parameters are dynamically optimized based on a reinforcement learning strategy, or a three-level response mechanism including heat reduction input, expert knowledge base matching and precise repair is triggered to process defect early warning. The system has an incremental learning capability and can be self-optimized along with a control period. The full-flow intelligent monitoring, real-time feedback and self-adaptive accurate control of the welding process are achieved, and the welding quality and the automation level are remarkably improved.

The invention relates to the technical field of antenna detection, in particular to a laser cutting antenna line type defective product detection method and system, and the method comprises the steps: obtaining processing batch information, building a coiled material motion coordinate system, and storing target antenna parameters and reference information of a dispensing region; acquiring image data of the antenna, and aligning the image data with the reference information in the coiled material motion coordinate system; extracting an antenna boundary according to the image data, and calculating a geometric deviation based on the antenna boundary and the reference information; obtaining a defect coordinate position according to the geometric deviation, and comparing the dispensing area with the geometric deviation to judge the severity of the defect; mapping the coordinate position to a coiled material motion coordinate system to obtain bad position data, and executing differentiation processing according to the severity; and storing the bad position data and the processing batch information corresponding to the differentiation processing, and generating a detection result. According to the invention, the antenna boundary and geometric deviation are analyzed through the image data and the reference information, defects in the antenna cutting process are identified, and the detection accuracy is improved.

The invention discloses a low-sodium myofibrillar protein flavor prediction method and a low-sodium myofibrillar protein flavor prediction device. The method comprises the following steps: firstly, constructing a protein-flavor interaction system containing myofibrillar protein, potassium chloride, kappa-carrageenan and representative aldehyde flavor compounds; then, by means of spectroscopy, rheometer combination, an electronic tongue and the like, protein conformational change, typical aldehyde molecule binding rate and taste features are represented in a multi-dimensional mode. And establishing a random forest prediction model, taking protein conformation parameters as input, and taking a flavor binding rate and effective taste intensity as output. The model can reveal cross correlation between protein conformations and functions and identify key conformation parameters that affect flavor. And in combination with quantification and prediction capabilities of the model, a conformation-flavor relationship can be systematically analyzed, a basis is provided for optimization of a low-sodium formula, flavor is enhanced while salty taste is increased and bitter taste is inhibited, and healthy and sustainable development of low-sodium meat products is promoted.

The invention discloses a road crack segmentation method and device based on deep learning, electronic equipment and a program product. According to the method, two collaborative and functional complementary processing paths are adopted to perform parallel processing on a to-be-segmented image so as to obtain corresponding feature information: on one hand, the to-be-segmented image is converted into a feature sequence containing spatial position information, and global semantic modeling is performed on the feature sequence; global semantic features representing the overall shape of the crack and long-range semantic dependence are obtained; and on the other hand, local texture and edge detail information of the crack is directly extracted from the to-be-segmented image, and detail enhancement features are obtained. Thirdly, fusing the global semantic features and the detail enhancement features to obtain target fusion features considering crack positioning, connectivity and boundary and texture expression; according to the segmentation result output based on the target fusion features, even tiny cracks can be effectively depicted, and then the continuity, boundary definition and robustness of tiny crack segmentation under the complex road background can be improved.

The embodiment of the invention provides a video subtitle recognition method and device, equipment, a medium and a program product. The video subtitle recognition method comprises the steps that text content and text observation features of all video frames in a target video are acquired; subtitle recognition prompt information is constructed based on the text content and the text observation features, and the subtitle recognition prompt information is used for indicating a subtitle recognition model to determine subtitles in the text content according to the text observation features; based on the subtitle recognition prompt information, the subtitle recognition model is called to generate a subtitle recognition result, and the subtitle recognition result comprises a subtitle spatial-temporal feature and a subtitle text. The subtitle recognition prompt information is constructed based on the text content and the text observation characteristics of each video frame, so that the subtitle recognition model can perform context correlation analysis on the text content in combination with the text observation characteristics, and the text segments with correlation are recognized as subtitles in the continuous frames, so that the subtitle recognition accuracy is improved.