522results about How to "Good practical value" patented technology

The invention provides an evolutionary neural architecture searching method and system based on similarity agent assistance, and the method comprises the steps: 1, initializing an architecture population, and selecting an architecture with the best performance as an initial reference architecture; 2, constructing a graph neural network variant as a feature extractor; 3, constructing an agent model, and training the agent model through a joint loss function; 4, the high-potential architecture is reserved according to the fitness value, authenticity performance evaluation is carried out on the high-potential architecture, and an evaluation result is added into the training set; combining the current population with the high-performance architecture predicted and screened by the proxy model, and updating the population through an environment selection strategy; and step 5, repeating the step 3 and the step 4 until the population performance converges, and finally outputting a global optimal architecture. According to the method, the high-performance neural network architecture can be quickly searched under limited computing resources, and an efficient and intelligent novel solution is provided for automatic design and optimization of the neural network.

The invention provides a subject interactive question-answering method based on a large model and a teaching assistant agent. The method comprises the following steps: establishing a subject knowledge base; determining knowledge fragments from the subject knowledge base based on a retrieval enhancement algorithm; taking the knowledge fragments as context information of the large model; on the basis of the subject knowledge base, cue words implying an interaction process are designed; cue words implied in the interaction process comprise an answer background of the large model, a multi-round answer logic structure and content and format requirements of each round of answer; and inputting the knowledge fragments and the prompt words implying the interaction process into a pre-trained large model to obtain a subject interaction question and answer result output by the large model. According to the method, the accuracy and correlation of large model answering can be improved through the subject knowledge base. And through the prompt words implying the interaction process, the large language model performs interactive multi-round questions aiming at a specific subject, so that question answering thoughts are hidden, guided and gradually given in the student question answering process, and the method has better practical value for students.

The invention discloses a power transmission line corona discharge risk assessment method based on bird nest polarization characteristics, and relates to the field of power transmission line safe operation. The method specifically comprises the following steps: establishing a bird nest polarization characteristic mathematical model, considering bird nest material and shape characteristics, and calculating an electric field enhancement coefficient; introducing critical polarization intensity as a criterion of corona discharge; establishing a kinetic equation of polarization intensity changing along with time, and revealing dynamic characteristics of the polarization process; and defining a corona discharge probability function to realize quantitative evaluation of the discharge probability. According to the method, a complete physical model from bird nest polarization to corona discharge is established, key links such as electric field enhancement, polarization dynamics, critical polarization intensity and discharge probability are covered, a systematic theoretical framework is provided for corona discharge caused by bird nests, and the method has remarkable practical value for guiding design, operation and maintenance of power transmission lines and is worthy of popularization and application. And an important contribution is made for ensuring safe and stable operation of a power system.

The invention belongs to the technical field of unmanned aerial vehicle control, and particularly relates to a multi-unmanned aerial vehicle task allocation method based on a multi-strategy fusion improved dung beetle algorithm, and the method comprises the following steps: solving a flight path and path distance from an unmanned aerial vehicle to a task point through an RRT algorithm; a task allocation mathematical model is constructed by adopting a minimum unmanned aerial vehicle total cost and a maximum task total income quantification objective function; setting a multi-strategy fusion improved dung beetle algorithm initialization parameter, initializing a dung beetle population position and differentiating the population; performing loop iteration on the position of the dung beetle population, and ending population updating until an iteration ending condition is met; and decoding the global optimal solution as a final optimal task allocation scheme. According to the method, efficient, stable and real-time task scheduling can be realized in a complex terrain and a large-scale task environment, and the actual requirements of large-scale unmanned aerial vehicle collaborative operation and complex task scheduling are met.

The invention provides a preparation method of a silicon substrate and a heterojunction solar cell, and the preparation method of the silicon substrate comprises the steps: an initial silicon substrate comprises a first surface and a second surface, a first mask layer is formed on the first surface, and a second mask layer is formed on the second surface; performing first dot matrix laser etching treatment on the first mask layer to form a first intermediate mask layer; the second surface is provided with a target area, the target area is used for arranging a grid line, and second dot matrix laser etching processing is carried out on the second mask layer corresponding to the target area to form a second middle mask layer; performing alkali etching treatment on the first surface through the first intermediate mask layer, and performing alkali etching treatment on the target area through the second intermediate mask layer; and removing the first intermediate mask layer to obtain a first suede, and removing the second intermediate mask layer to obtain the silicon substrate with the target area as the second suede. The silicon substrate with excellent adhesiveness between the second surface and the grid lines and ultralow reflectivity on the first surface is prepared by using the method.

The invention discloses a multi-target fusion dairy cow oestrus behavior detection method and system, and belongs to the technical field of target detection of deep learning. The method comprises the following steps: S1, collecting a dairy cow oestrus behavior image, and performing image enhancement on the dairy cow oestrus behavior image; s2, constructing a PBF-YOLO deep learning network model, wherein the PBF-YOLO deep learning network model comprises a lightweight module, a background information enhancement module and a feature aggregation module; and S3, identifying a dairy cow oestrus behavior region of interest through the PBF-YOLO deep learning network model, and detecting the dairy cow oestrus behavior. The invention provides a solution for the situation that the existing lightweight algorithm is low in detection precision and poor in robustness due to the fact that the cow oestrus behavior and other background features are similar and are difficult to distinguish in an actual pasture and cows with different scales appear on the same image.

The invention relates to a eutectic microcapsule wrapping technology and application thereof in cosmetics, and belongs to the technical field of cosmetic production. The wall material of the eutectic microcapsule inclusion comprises at least one of hydrogenated lecithin, caprylic / capric triglyceride, sucrose stearate, PEG-8 caprylic / capric glycerides, PEG-40 hydrogenated castor oil, soya bean lecithin, cyclodextrin, ethyl palmitate and troxerutin. The core material is prepared from at least one of salicylic acid, 4-butylresorcinol, glabridin, a kava extract, cortex magnoliae officinalis total phenols, ginseng oil, ganoderma lucidum spore oil, astaxanthin, coenzyme Q-10, a rosemary extract, ceramide and vitamin E. According to the invention, multiple eutectic coating can be formed, the consumption of the wall material can be reduced, the cost is saved, the wall material and the core material are matched for use, so that the excellent effects of permeation, oxidation resistance, soothing, allergy resistance and skin repair can be achieved, and the utilization rate of active substances is increased.

The invention discloses an in-situ treatment system for enhanced purification of overflow pollution of a river inlet discharge port, and belongs to the technical field of water environment treatment. The in-situ treatment system comprises a magnetic flocculation precipitation unit and an ecological slope protection purification unit. Overflow sewage of a river channel discharge port is firstly intercepted to the magnetic flocculation sedimentation unit by the sewage lifting pump, and in the process, the overflow sewage is firstly preliminarily mixed with a composite flocculant polyferric silicate sulfate in the sewage inlet pipe, and then enters the magnetic seed mixing reaction tank to be further fully and uniformly stirred and mixed with magnetic seeds. According to the invention, an integrated overflow sewage pretreatment device and an ecological slope protection deep purification filter tank are established near a river inlet discharge port side slope by utilizing a slope top and a large-area river side slope. And under the conditions of low investment and low operation cost, the problem that the river inlet discharge port overflow pollution treatment generally lacks a site is solved according to local conditions.

The invention provides a calculation method, device and equipment for a heat transfer coefficient of a railway vehicle and a medium, and the method comprises the steps: for each structure model in a thermal resistance model of the railway vehicle, calculating a plurality of heat transfer layer thermal resistance values corresponding to the structure model, and calculating a structure thermal resistance value of the structure model based on the plurality of heat transfer layer thermal resistance values; for each area model, calculating the area thermal resistance value of the area model based on the structure thermal resistance values of the plurality of structure models corresponding to the area model; and calculating the total thermal resistance value of the rail vehicle based on the regional thermal resistance value of each regional model, and calculating the heat transfer coefficient of the rail vehicle based on the total thermal resistance value. By means of the method and device, the efficiency and accuracy of calculating the heat transfer coefficient of the railway vehicle are improved, and the research and development period can be shortened.

The invention discloses a BiOBr oyster shell composite photocatalyst as well as a preparation method and application thereof. The composite photocatalyst is prepared by taking oyster shell as a carrier, loading a BiOBr material on the oyster shell carrier, converting the oyster shell into an oyster shell functional material, and then adding KI and Bi (NO3) 3 by a solvothermal method to generate the BiOBr / oyster shell composite photocatalyst in situ. The BiOBr / oyster shell composite photocatalyst with visible-light response is constructed, the oyster shell serves as a carrier material, the specific surface area of BiOBr is increased, meanwhile, photon-generated carrier separation can be accelerated, the composite material is endowed with efficient photocatalytic activity and adsorbability under visible light, the efficient degradation effect on dye and antibiotic pollutants in water is achieved, and the visible-light response BiOBr / oyster shell composite photocatalyst has a good application prospect. While resource reutilization of the oyster shells is realized, the method has good practical value and potential application prospect in the fields of water purification and the like.

The invention discloses a method for preparing a carbon-coated high-entropy alloy by a photo-reduction method. The method comprises the following steps: (a) dissolving carbon quantum dots and transition metal salt in absolute ethyl alcohol to prepare a solution A; (b) dispersing a conductive carbon material into the solution A in the step (a) to prepare dispersion liquid B; (c) irradiating the dispersion liquid B obtained in the step (b) under a light source for 0.5-6 hours to obtain dispersion liquid C; and (d) carrying out centrifugal separation on the dispersion liquid C obtained in the step (c) to obtain a solid, and cleaning the solid to obtain the carbon-coated high-entropy alloy. According to the method, amorphous carbon can coat high-entropy alloy nanoparticles or atomic clusters, the catalytic activity and durability of the precious metal are remarkably improved, and the product is particularly suitable for the fields of hydrogen production through water decomposition, fuel cells and the like, can inhibit the reverse water decomposition reaction on the surface of the precious metal and has the advantages of being high in catalytic activity and good in cycling stability.

Along with rapid development of an artificial intelligence technology, calculation-intensive applications generally need a large number of calculation resources, but the calculation resources of mobile terminal equipment (such as vehicles) are often limited, so that the requirements of the applications are difficult to meet. In order to solve the challenge, the invention provides a deep reinforcement learning method based on a Rainbow DQN algorithm, and the deep reinforcement learning method is specially used for solving the problem of calculation unloading in the V2X (Vehicle-to-Everything, V2X). Under the condition of considering task characteristics and system resources, according to the capability differences of various computing units of cloud, edge and end in the aspects of data operation, analysis, decision making and the like, the method integrates multiple improved deep reinforcement learning methods, aims to solve the problem of insufficient computing capability of end vehicles, optimizes the time delay and energy consumption of the whole system, and improves the system performance. The problem of calculation unloading in an Internet of Vehicles scene is solved.

The invention relates to a light source driving method and system capable of continuously dimming, and relates to the technical field of light source driving, and the method comprises the steps: obtaining an initial voltage signal and an initial current signal inputted by a vehicle lamp power supply of a vehicle; filtering the initial voltage signal and the initial current signal to obtain a filtered voltage signal and a filtered current signal; based on the voltage signal and the current signal, predicting and updating the state of a vehicle lamp power supply through an extended Kalman filtering algorithm; according to the brightness required by the vehicle lamp, the output brightness of the vehicle lamp is controlled by adjusting the duty ratio of the PWM signal; the real-time illumination intensity and the environment illumination intensity are obtained through a vehicle lamp sensor, and the duty ratio is adjusted in a self-adaptive mode based on a PID control algorithm; and monitoring in real time, minimizing the total energy consumption of the vehicle lamp based on an energy optimization algorithm to obtain the minimum total energy consumption of the vehicle lamp, and driving and adjusting the vehicle lamp under the minimum total energy consumption. According to the invention, continuous dimming operation with high accuracy, high energy efficiency and higher stability can be provided.

The invention discloses an electric power spot market transaction strategy optimization method and system suitable for a combined distribution and storage main body, and belongs to the field of electrical engineering. According to the method, a quotation strategy optimization model of a joint distribution and storage main body in an electric power spot market is constructed based on a master-slave game architecture. The method comprises the following steps: firstly, respectively establishing transaction strategy optimization models of three subjects including a thermal power generating unit, a new energy power station and industrial and commercial users; secondly, an electric power spot market optimization clearing model is constructed; thirdly, the market optimization clearing model is equivalently converted into a constraint condition of a joint distribution and storage main body optimization problem through a KKT condition, so that a non-convex nonlinear double-layer game model is converted into a nonlinear single-layer optimization problem; finally, nonlinear terms in the optimization problem are linearized, and a mathematical model easy to solve is formed. Therefore, by constructing the transaction strategy model and performing numerical simulation, the optimal transaction strategy of the combined distribution and storage main body is finally output, the optimal transaction strategy can be used for guiding the spot market transaction decision, and the economic benefit is maximized.

The invention discloses a detector nonlinear measurement device and method based on double laser light sources. The detector nonlinear measurement device comprises a first laser, a second laser, a chopping disc, a chopping controller, a first light trap, a second light trap, a gold-plated integrating sphere, a photoelectric detector, a lock-in amplifier and a computer. According to the method, nonlinear characteristic measurement of the detector in a large-value range can be realized, single-time and multi-time accumulative measurement is considered, and a corresponding calculation method is provided. In addition, the luminous flux is adjusted directly through the on-off of the double laser light sources, moving parts are reduced, and the measurement stability is better. By adopting a double-laser measurement method, an optical measurement system is remarkably simplified, and the influence of non-ideal factors such as light blocking of an optical filter is reduced. Meanwhile, the projection laser is reflected for multiple times through the gold-plated integrating sphere and then is uniformly projected to the photoelectric detector to cover the whole photosensitive surface, so that the nonlinear measurement error caused by the difference of the light uniformity of the photosensitive surface is avoided.

The invention discloses a Cesium-based multi-view visual field measuring and calculating method and system, and the method comprises the steps: setting the space coordinates and number of observation points, obtaining a view layout according to the demands of a user, generating a depth map based on the space coordinates and number of the observation points and the view layout, and carrying out the measurement and calculation of the visual field. Carrying out vision field index calculation on the depth map, obtaining a vision field index of each observation point, aggregating the vision field indexes of all the observation points, and completing the vision field calculation operation of the whole scene; according to the method, the function of quickly and accurately analyzing the visual field by utilizing the parallel processing capability of the GPU is realized, the calculation time of the visual field analysis is greatly shortened through the GPU acceleration based on the shadow map, so that the real-time or near-real-time analysis becomes possible, and meanwhile, due to the application of the multi-view analysis and the accumulated visual field index, the visual field analysis efficiency is greatly improved. The accuracy and the reliability of an analysis result are improved, and the method is suitable for wide popularization and application.

The invention provides a campus emotion risk individual assessment method and system based on expression recognition, and the method comprises the steps: collecting monitoring image data for recognition, and obtaining recognized individual expression data and related information data; performing processing according to the individual expression data to obtain group emotion data; correcting the individual expression data according to the group emotion data to obtain individual negative emotion data; and evaluating an emotion risk target individual according to historical statistics of the individual negative emotion data. According to the method, group emotional abnormity can be accurately eliminated in a group scene, the individual emotional risk degree is objectively judged, and the accuracy of identifying the campus emotional risk target is improved.

The invention discloses a method for characterizing three-dimensional morphological characteristics of coarse aggregate, and belongs to the field of road building materials. According to the method, the problem that the three-dimensional morphological characteristics of the coarse aggregate cannot be comprehensively represented by the existing method is solved. The method comprises the following steps: firstly, acquiring a three-dimensional point cloud image of the surface of the coarse aggregate, enveloping and fitting the three-dimensional point cloud image of the coarse aggregate by using an ellipsoid, and then determining the shape of the coarse aggregate according to the lengths of three half axes of the ellipsoid; then, deleting a part with relatively large corner angle change in the coarse aggregate three-dimensional point cloud image, and representing the corner angle of the aggregate by using the calculated volume change rate; and smoothing the three-dimensional point cloud image of the whole coarse aggregate surface, and representing the texture roughness of the aggregate by using the calculated surface area change rate. According to the method, the shape, the corner angle and the texture of the coarse aggregate can be evaluated by quantifying the three-dimensional indexes. The method can be applied to characterization of the three-dimensional morphological characteristics of the coarse aggregate.

The invention discloses a new energy power station spot transaction strategy simulation and market optimization operation method and system based on a master-slave game, and belongs to the field of electrical engineering. According to the method, a master-slave double-layer game architecture is adopted, and a spot market transaction strategy optimization model of each new energy power station is firstly established; secondly, constructing an electric power spot market optimization clearing model; thirdly, the market optimization clearing model is equivalently converted into a constraint condition of a new energy power station optimization problem through a KKT condition, so that a non-convex nonlinear double-layer optimization model is converted into a mathematical programming problem with equilibrium constraint of each new energy power station; and finally, the optimization model of each new energy power station is converted into a constraint condition through the KKT condition, and a ''equalization problem with equalization constraint'' is formed in parallel. Therefore, simulation of transaction strategies of a plurality of new energy power stations can be realized, market optimization clearing is synchronously realized, and the method is used for guiding core businesses such as clearing price prediction and market benefit evaluation of market operators.

The utility model discloses a water quality sampling device, relates to the technical field of water quality sampling, and solves the problems that as different detection items need to detect the water quality of different depths and areas, after the sampling device in the patent samples the water quality, it is difficult to ensure that a sample cannot be mixed with the water quality of water areas in other depths, and the water quality cannot be detected. Therefore, the purity of a detection sample is difficult to guarantee, and the accuracy of a detection result is influenced. A water quality sampling device comprises: a buoy; the buoy adopts a conical block structure; a fixing frame of a frame structure is fixedly arranged at the top of the buoy. A connecting rod of a cylindrical structure is slidably arranged in a circular through groove in the middle of the top of the fixing frame. A penetrating type circular sliding groove structure is formed in the middle of the connecting rod. The bottom of the connecting rod is fixedly connected with a container cylinder of a cylindrical structure; the purity of a sample in the container cylinder is ensured, the accuracy of a detection result is ensured, and the practical value is better.

The invention discloses an axle end threaded hole compliance automatic detection system and method, and the system comprises an axle transmission line which extends in the X-axis direction, a threaded hole compliance detection work station is arranged beside the axle transmission line, the threaded hole compliance detection work station comprises a pedestal, and a lifting platform which can move in the Z-axis direction on the pedestal. The lifting platform is provided with two positioning supporting seats used for supporting and positioning an axle and a portal frame extending in the Y-axis direction, and the portal frame is provided with two positioning clamping jaws capable of moving in the Y-axis direction and used for axially positioning the axle and a power source used for driving the positioning clamping jaws to move. The lifting platform is further provided with a multi-degree-of-freedom displacement unit, and the displacement end of the multi-degree-of-freedom displacement unit is connected with an end effector used for achieving axle end threaded hole detection. The axle detection efficiency is greatly improved, the manual detection cost is reduced, and the product quality consistency is improved.

The present invention provides a stable composition containing mussel mucin, which has the characteristics of good stability, high DOPA content, and not easy to change color. The present invention also provides a preparation method of the preparation, which improves the DOPA stability of mussel mucin by adding specific components and improves the yellowing of color.

The utility model belongs to the technical field of tailor-welding equipment, and particularly discloses a robot welding tool for a hard pipe in a hydraulic pipeline system. The welding tool comprises a bottom plate, a mounting box, a first fixing set and a second fixing set. The bottom plate is horizontally arranged; the mounting box is arranged on the bottom plate and is provided with two parallel vertical plates; the first fixing set and the second fixing set each comprise a limiting piece arranged on the bottom plate, a supporting piece arranged on the mounting box and a pressing piece, two pipe openings of the hard pipe are limited and fixed through the two fixing sets, and tailor-welding machining of the two ends of the hard pipe and the connecting pieces can be achieved. The hydraulic hard pipe welding device can greatly improve the welding efficiency of hard pipes, ensure the welding quality, guarantee the size requirement of products and reduce the production cost, and can be widely used for machining different complex hydraulic hard pipe pipelines.

The invention provides a method and system for evaluating consistency of multiple / secondary blood vessel segmentation results, and the method comprises the steps: reading original data of a blood vessel data set and multiple / secondary blood vessel segmentation result matrixes, and randomly selecting two blood vessel segmentation result matrixes to calculate a union set, an intersection set and a symmetry difference; selecting any point in the symmetric difference, simulating a path diffusion process by using a convolution kernel to perform iterative growth, obtaining a spherical set under the constraint of a union set range and a given geodesic distance, and calculating an intersection of the spherical set and a union set of the blood vessel segmentation result matrixes and an intersection of the intersection and two / time blood vessel segmentation result matrixes; performing calculation and statistical analysis on each point to obtain a statistical result, and mapping the statistical result back to the original data and two / time blood vessel segmentation result matrixes; and thus, a statistical result of any two / time blood vessel segmentation result matrixes is obtained, and a visual processing result is output. The method can be used for measuring the performance of the blood vessel segmentation algorithm or comparing the differences among the blood vessel segmentation algorithms, and accelerates the development and verification process of the blood vessel segmentation algorithm.

The invention provides a method for processing text unstructured information by calling an artificial intelligence API (Application Program Interface) by utilizing Python. The method comprises the following steps: S1, automatically extracting text, table and picture information from a PDF (Portable Document Format) file by utilizing Python and a related library; s2, performing a data analysis and storage technology based on a GPT model by using a function, and outputting a CSV file; and S3, performing table merging and data cleaning on the CSV file. According to the method for processing the unstructured information of the text by calling the artificial intelligence API interface through the python, an automatic process for extracting the text data in batches and a full-automatic process design from PDF document extraction to unstructured data generation are provided, manual intervention is reduced, and the processing efficiency and accuracy are improved. Especially for processing of large-scale documents, the working efficiency is remarkably improved through the process.

The invention discloses a myopia prevention and control effect prediction system and equipment. The system comprises a data acquisition module, a prediction module and a result output module, data of gender (X1), age (X2), corneal eccentricity (X3), relative peripheral retina diopter (X4) and spherical aberration (X5) of a myopia prevention and control object are collected through a data acquisition module, then a prediction module inputs the data acquired by the data acquisition module into a regression equation to predict an eye axis growth amount (Y), and finally a prediction result of the eye axis growth amount (Y) is output through a result output module. According to the system, through mutual cooperation of the data acquisition module, the prediction module and the result output module, the myopia control effect of a myopic child can be predicted before the myopic child fits the orthokeratology lens, and the child with a poor myopia prevention and control effect can be screened out, so that the problem that the ideal myopia prevention and control effect cannot be achieved due to expense is avoided; and auxiliary guidance is provided for estimating the myopia prevention and control effect of myopia children wearing the orthokeratology lens in the early stage.

The invention discloses a group high-voltage current transformer operation metering error evaluation method and a storage medium. The method comprises the steps of S1, data acquisition; the method comprises the following steps: acquiring known measurable electric power data, and extracting various information physical relationships and massive mutual inductor measurement data sets by analyzing three-phase power electrical physical relationships of a power grid; s2, modeling is carried out; determining a direct constraint relation of the current transformer, establishing a plurality of target sub-functions, and converting an error evaluation problem into a multi-target solving problem; s3, optimization solution is carried out; a multi-target multi-constraint problem is solved, and a CT primary side current truth value set # imgabs0 # is obtained; S4, evaluation is carried out; the mutual inductor error is evaluated according to the optimal solution, the error E of the n CTs of the to-be-detected group is calculated, and E is the final evaluation result. The storage medium is used for storing a computer program for executing the method. The method has the advantages of being simple in principle, good in real-time performance, wide in application range, low in evaluation cost and the like.

The utility model discloses a municipal sludge incineration system which comprises a dry sludge bin, an ignition furnace, a pre-combustion chamber, a suspended state incinerator, a collecting device, a heat exchange system and a waste gas treatment device. The pre-combustion chamber is connected with the suspended state incinerator and supplies heated air to the suspended state incinerator; dry sludge enters a suspended state incinerator to participate in combustion with heated air; the collecting device I is used for collecting incineration ash generated by combustion; the heat exchanger I is used for heating combustion-supporting air by using combustion flue gas discharged by the suspended state incinerator, and the combustion-supporting air is introduced into the suspended state incinerator to participate in combustion; the heat exchanger II heats cold air by utilizing excessive waste heat flue gas discharged by the heat exchanger I, so that waste heat utilization is realized; and the collecting device II is connected with a flue gas outlet of the heat exchanger II and is used for collecting fly ash in the flue gas. According to the invention, the incineration heat is in a closed-loop circulation loop, and redundant heat is also recycled, so that the use efficiency of the heat is greatly improved, and the energy consumption is reduced; ash generated by the system is collected in a multi-stage mode, tail gas is effectively treated, and environmental pollution is reduced.

The invention relates to an intelligent detection method and system for a fracture defect of a middle frame bracket of an electronic product, the intelligent detection method and system for the fracture defect of the middle frame bracket of the electronic product are ingenious in design and accurate in positioning, a bracket area is segmented and folded, iteration is continuously carried out based on the width and height of an extracted picture, and the detection accuracy is improved. The folding algorithm is ended until an iteration termination condition is met, so that the area where the fracture defect is located is continuously reduced and refined, and the accuracy and the detection precision of subsequent rapid detection of the fracture defect are further improved; then target detection is carried out on the detection area, each sub-image is independently calculated to obtain a detection result, finally, the detection results obtained by all the sub-images are combined to obtain a total detection result, and finally, back calculation is carried out on the total detection result, so that the accuracy of the detection results is improved; according to the detection method, the health state of the support material and the specific position of the fracture defect are obtained, the detection precision is higher, the detection speed is higher, and therefore the good practical value is achieved.

The invention discloses an inspection system suitable for a steel box girder, which belongs to the technical field of bridge engineering, and is characterized in that a walking track is arranged in the steel box girder, an inspection robot is arranged on the walking track, and a chassis walking mechanism, a space mechanical arm and a shooting module are combined to inspect the steel box girder. The shooting module can be driven by the chassis walking mechanism to penetrate through the partition plate in the box girder to walk when the space mechanical arm is in the initial stacking state, and is driven to the corresponding shooting position through folding / unfolding control of the space mechanical arm after moving in place, and therefore detection of the interior of the steel box girder is completed. The inspection system provided by the invention is simple in structure and convenient to use, can automatically complete the detection of the interior of the steel box girder, ensures the efficiency and accuracy of the detection of the interior of the steel box girder, avoids position omission during manual detection, improves the efficiency of the detection of the interior of the steel box girder, reduces the cost of the detection of the interior of the steel box girder, and improves the detection efficiency of the interior of the steel box girder. The internal detection quality of the steel box girder is ensured, and the practical value is good.