2121results about How to "Reduce work stress" patented technology

The invention discloses an auxiliary film reading system for a capsule endoscope image. A data acquisition module is used for acquiring capsule endoscopic image data of an examined patient; an image position classification module is used for classifying capsule endoscopic images according to different shot parts by using a first convolution neural network (CNN) to obtain image sequences of different shot parts; an image sequence description module is used for carrying out image feature extraction on the image sequences of different shot parts by using a second convolution neural network to obtain feature vector sequences of different alimentary canal part image sequences and for transforming image features in the feature vector sequences into descriptive texts by using a recursive neural network (RNN) to generating an auxiliary diagnosis report. Therefore, the work load of observing an alimentary canal image by a doctor can be reduced and thus the diagnosis efficiency of the doctor can be improved.

The invention provides a production method of fructus monordicae extract with over 60% of mogroside V content. In the method, the fresh momordica grosvenori is subjected to the following technique processes: smashing, saccharification, extraction by water, concentration, sedimentation and centrifugation, ion exchange resin refining, macroporous resin refining, concentration, aluminum oxide refining, concentration, spray drying and finishing product. The method can produce high-quality fructus monordicae extract products, reduce material loss, shorten period of production, increase product yield, lower production cost, and is suitable for industrial production. The mogroside V content in the fructus monordicae extract produced by the method is more than or equal to 60%; the fructus monordicae extract has high product purity, single composition, pure white color, good color and taste, fresh and cool sweet taste like that of sugar, simultaneously overcomes the technical problem for industrial production of the high-content and high-quality fructus monordicae extract, and successfully realizes large-scale production.

Disclosed is an abrasive tool made of epoxy resin-bonded compound abrasive material, which comprises epoxy resin, abrasive material and curing agent. The weight percentages are as follow: the epoxy resin is 15 to 78 percent, the abrasive material is 2 to 44 percent and the curing agent is 4 to 18 percent. Wherein, the abrasive material comprises one of diamond, silicon carbide or corundum or the combination of more than one material. The curing agent comprises one of 4.4`-methylenedianil, solid diaminodiphenyl methane or liquid modified diaminodiphenyl methane or the combination of more than one material. The component of the abrasive tool also comprises the other auxiliary material, which comprises one of alumina, barium chloride, sodium chloride, quartz sand, organobentonite, wetting agent, zirconium silicate, calcium carbonate, graphite powder, garnet, magnesium sulfate, phenolic resin, zinc stearate and magnesium stearate, or the combination of more than one material. The product can be made into any shape the same with the present product, which can be applied in various grinding and polishing equipment with wide application scope.

The invention discloses a compute task distributed dispatching device in a server cluster. Compute tasks from front-end devices are dispatched to processing servers in the server cluster for carrying out compute processing. The compute task distributed dispatching device in the server cluster comprises a register management module, a distributing module and a coordinating module, wherein the register management module is used for receiving register of the processing servers and receiving computing resource information and load information reported by the processing servers, the distributing module is used for distributing the processing servers bound to the front-end devices and issuing the bound processing servers to the front-end devices, and the coordinating module is used for receiving assistance requests sent by the registered processing servers when the processing servers are overloaded and coordinating other processing servers with residual processing capacity to assist the corresponding processing servers to complete the compute tasks contained in the assistance requests according to the processing capacity and the load information of the other processing servers. According to the compute task distributed dispatching device in the server cluster, the work pressure of a management server is reduced, automatic adjustment of overloaded processing servers is achieved automatically, and the load balance of the server cluster is guaranteed.

The invention discloses a method for processing head structure image data based on a deep learning algorithm. The method comprises the following steps of acquiring head magnetic resonance structure image data and storing the head magnetic resonance structure image data in a computer; first extracting corresponding features of multiple frames of pictures through convolutional neural network models respectively, then taking a feature of a sequence picture as a sequence input of an LSTM (Long Short-Term Memory) model, and taking combination of the feature of the sequence picture and a tag corresponding to a whole sequence as a training sample of the LSTM model; sequentially inputting 100 layers of pictures in the middle of an MRI (Magnetic Resonance Imaging) image by adopting 3 layers of LSTM models, wherein each layer of picture corresponds to a 1024-dimensional feature, and the end of the model is a full connection layer of two classifiers; importing to-be-analyzed medical data into a deep learning model to perform matched medical analysis; and when the model training is finished, arranging test samples in the same format as an MRI sequence layer, and then performing classification by using the trained deep learning model to judge whether the MRI image belongs to a smoker, a drug addict or a disease patient.

The invention discloses a mid-term air traffic conflict detection method based on 4D flight path and radar data, which relates to air traffic control automation. In the method, accurate 4D flight path data and real-time radar data are projected onto a display panel of an air traffic control system, each aircraft pair is filtered according to the accurate 4D flight path data and the real-time radar data, and aircrafts which are unlike to come into conflict with other aircrafts in 30 minutes are determined; and the detection is focused on the aircraft pairs which have potential to run into conflicts, the covariance between the relative speed and relative position of the aircraft pairs is calculated by using coordinate transformation according to the vertical interval and time interval of two aircrafts, and then the probability that a predicted position interval is smaller than a safe interval at time t is calculated and a detection report is made. In the invention, the drawbacks of long-term and short-term conflict detection are overcome, and the long-term conflict detection, mid-term conflict detection and short-term conflict detection are combined to ensure flight safety. The average conflict detection time is about 0.028 seconds, and the method has high real-time performance.

The invention relates to a process for purifying and recycling hot water by a membrane method as well as used equipment and process flow thereof. The process is a technology for treating, purifying and recycling hot and industrial high-temperature wastewater on the basis of a membrane method. The process has the characteristics that the combined process of an ultrafiltration and microfiltration membrane system, reverse osmosis, a nanofiltration membrane system, pretreatment, active carbon adsorption, ion exchange, and the like can be operated at high temperature and purify different hot water and wastewater, has dual effects of recycling water resources and heat energy and can adopt different process combinations and process flows according to components of water quality so as to achieve the aims of simplifying the process flows and reducing the system and running cost.

The invention discloses an image retrieval method and apparatus, an image information acquisition method and apparatus, an image identification method and apparatus, an image identification system, a method and apparatus used for calculating an image feature value, and an electronic device. The image retrieval method comprises the steps of extracting a local feature of a to-be-retrieved image; calculating a feature value of the local feature by adopting a pre-trained depth autoencoder network model; performing matching on the feature value and feature values of registered images in an image retrieval database; and selecting a registered image, which meets a preset condition, as a retrieval result of the to-be-retrieved image according to a matching result. By adopting the method, a depth autoencoder network can effectively keep distance information and a distinguishing capability between the feature values in a process of performing compressed representation on the local feature, so that the image retrieval precision can be effectively improved, the workload of rearrangement and filtration can be reduced, and the retrieval efficiency can be enhanced.

The invention relates to the field of identification, and particularly relates to a method for carrying out electronic identity recognition by utilizing two-dimensional codes. The method is characterized in that at least one client terminal, a ticket reservation system, a ticket business server, a ticket business database, a ticket selling workstation, an electronic tag printer, an identity card reader, a ticket checking entrance, wireless radio-frequency equipment and an identity card database are adopted in the method, wherein the client terminal refers to a passenger, namely a person who needs to purchase a ticket for taking a bus and acquires a two-dimensional code electronic tag through intelligent terminal equipment during ticket purchasing; the ticket reservation system refers to a network ticket reservation system; and the ticket selling workstation refers to the existing station ticket window and various agent ticket offices. The method for fast and conveniently carrying out the electronic identity recognition, which is provided by the invention, mainly aims at overcoming the defects of the existing paper ticket and solving the problem of difficult queuing during the ticket purchasing and can reduce the labor intensity of ticket checking personnel, improve the incoming efficiency of the passengers and realize the high security and the high antifalsification through the uniqueness of a second-generation identity card.

The invention discloses an ultrasonic-assisted tea oil extraction method using subcritical CO2, comprising the steps as follows: (1) taking dry tea seed, hulling, crushing, and placing in an extraction kettle; (2) injecting the subcritical CO2 into the extraction kettle to mix with the tea seed; (3) extracting the raw material for 0.5-5 hours; (4) decompressing and evaporating so as to completelygasify the subcritical CO2 liquid, wherein the temperature in evaporation process is controlled at 30-70 DEG C, and finally discharging the extracted tea oil from the bottom of a separation kettle. The method has the advantages that (1), the advantages of the subcritical CO2 extraction technology and the ultrasonic extraction technology are sufficiently utilized to shorten the extraction time andimprove the extraction yield; (2), the work pressure is low, and the natural activity of heat-sensitive material cannot be influenced since the extraction temperature is less than 30 DEG C, and the nutritional value and quality of the tea oil are improved; (3), the device cost is low, the refining process is less, the tea oil loss is less and the product oil yield is high; (4) a natural green food is obtained since the CO2 solve is residue-free.

A method for analysing medical treatment data based on deep learning, and an intelligent analyser thereof, which can effectively alleviate work stress for doctors in a hospital or medical researchers and can perform scientific analysis on a large amount of medical treatment data or medical data and obtain an analysis result matched therewith. The core content of the method is the establishment of a model in a computer using a deep convolution neuron algorithm in deep learning. The model assists the doctors to make correct judgments and effective decisions for the large amount of medical treatment data using mass medical treatment data selection and optimization model parameters by “training” the model to automatically learn a pathology analysis process of the doctors or the medical researchers and then helping them in processing the large amount of medical treatment data or medical data. The present invention can greatly reduce the work stress for the doctors or the medical researchers and improve the work efficiency thereof. The present invention can free the doctors or the medical researchers from heavy analysis work on the medical treatment data or medical data, so that they can put more effort into other more important work.

The invention relates to a well drilling method of a through tubing of a gas lift under-balanced coiled tubing, which is applied in the technical field of petroleum well drilling. The through tubing operation way of the coiled tubing is adopted for carrying out under-balanced well drilling or low-pressure well drilling on target well sections at the lower part in a shaft with a drilled upper borehole, wherein the target well sections are a vertical well, a directional well or a horizontal well. The method has the following effects: the adoption of the through tubing operation of the coiled tubing can form three circulation channels by utilizing the coiled tubing, an oil tube and a sleeve, and a matching pipe column has good universality, does not need a special drilling column or an inner tube and does not need to use a rotary blowout preventer and other special equipment; a gas injection channel and a slurry injection channel are independent mutually, thereby being capable of more conveniently controlling the well bottom pressure and the under-pressure value and using an underground power drilling tool and a slurry pulse measurement system which are applicable to slurry well drilling to carry out well drilling operation and trajectory control; and the method does not need to be connected with a single and can keep the continuous circulation during the tripping process and the drilling process, and realize the full-process controllable under-balanced well drilling operation and the full-process controllable low-pressure well drilling operation.

A thermal energy storage and recovery arrangement is provided. The arrangement has a first thermal energy storage, a compression and expansion unit coupled thereto, and a second thermal energy storage coupled to the compression and expansion unit. The first and second thermal energy storage are adapted to work at a temperature equal to or higher than ambient temperature. In a charging mode, the first thermal energy storage is adapted for receiving and releasing thermal energy to a received working medium. The compression and expansion unit is adapted for receiving and compressing the working medium from the first thermal energy storage. The second thermal energy storage is adapted for receiving the compressed working medium and thermal energy therefrom and storing at least part of this energy. The compression and expansion unit is adapted for receiving the compressed working medium from the second thermal energy storage and expanding the compressed working medium.

The invention discloses a satellite communication network abnormity detection method based on network control. A fuzzy C-means clustering algorithm is utilized to study normal audit data and an abnormity detection mechanism is established, so that a post analysis on an abnormal event is realized; a support vector data description single classifier is utilized to summarize a normal short sequence and an abnormal short sequence as well as a hidden Markov model is used to carry out abnormity detection on communication signalings of all earth stations in a whole time slot, so that real-time detection based on the communication signalings can be realized; and by utilizing the above-mentioned methods of the post analysis and the real-time detection, a satellite communication network abnormity detection prototype system based on network control is realized. According to the invention, the post analysis method is suitable for large-scale data as well as a declaration omission rate and a misdeclaration rate are low; and the real-time detection method has a good detection performance as well as signaling detection time is short and the reaction speed is fast; therefore, an on-line detection object can be realized and safety of a signaling system can be enhanced.

The invention discloses a FOTA (Firmware Over-The-Air) upgrade method, which comprises the following steps of: establishing a partition upgrade information identification table, and setting a backup buffer zone, wherein the partition upgrade information identification table is used for storing the upgrade state of each partition; before partition upgrade is carried out, storing a system file in the partition which needs to be upgraded into the backup buffer zone; according to a partition upgrade file in a FOTA upgrade package, upgrading each partition which needs to be upgraded in sequence; and when the partition upgrade fails, obtaining the partition information of the partition which fails in being upgraded from the partition upgrade information identification table, obtaining an original partition system file stored in the backup buffer zone according to the obtained partition information of the partition which fails in being upgraded, and recovering the partition which fails in being upgraded. The invention simultaneously discloses a FOTA upgrade system.

The invention relates to a broadcasting PGM (Probabilistic Graphical Model) master-slave output one-click emergency system and a broadcasting PGM master-slave output one-click emergency method. The system is of a structure that master and slave program signal links are respectively formed by master and slave broadcast control workstations, master and slave inverters, master and slave video servers, a plurality of video-audio distributors and master and slave switching equipment; and the master and slave program signal links are respectively connected with two input ends of a 2*1 switch arranged on a control panel at the output end of the master switching equipment. According to the invention, a control signal link and the program signal links are adopted; a last-level signal inversion control mode adopted by a conventional broadcasting system is changed; once the abnormal condition of a current link broadcast signal is monitored, the 2*1 switch can be directly inverted to a backup link by the universal control panel; the complex fault judgment process is saved; the working pressure of a broadcasting operator on duty in a television station is reduced; the fault judgment process is simplified; the timely observation and the rapid processing on a fault are implemented; and a broadcasting accident is effectively avoided.

The invention provides a mobile commerce identity authentication method. In the method, key negotiation is realized by the most efficient elliptic curve-based public key cryptosystem in public key cryptosystems. An elliptic curve algorithm has the characteristics of low calculated amount and high safety, so the elliptic curve algorithm is used for generating a public key and a private key, and a public key cryptosystem without a third-party is realized and applied to the two-way authentication service of a client and a server.

The invention discloses a spinning printing and dyeing waste water reclamation method, a device and the application thereof. The method comprises the steps of: filtering spinning printing and dyeing waste water; sequentially carrying out homogenizing, equalizing quantity and primary sedimentation on filtrate; adding sodium chloride into waste water supernate, and evenly mixing for indirect electro-catalysis oxidation; sequentially evenly mixing the waste water treated by indirect electro-catalysis oxidation with basic aluminium chloride and polyacrylamide; removing the precipitated supernate, and sequentially treating by activated carbon adsorption and reverse osmosis; removing trace organic chloride generated in the electrolytic process and the rest COD and color; and obtaining recyclable water resource. The device for realizing the method comprises a grate, a regulating reservoir, a sealed electrolytic cell, a flocculation tank, an inclined tube sedimentation tank, a sand filter tank, an active carbon filter, an intermediate water tank, a reverse osmosis unit, a clear water reservoir, a brine tank, a chlorine absorbing tank, a sludge concentration tank and a plate-and-frame filter press. The invention can effectively treat spinning printing and dyeing waste water, and has the advantages of high treatment efficiency, stable effluent quality, simple technological process and the like.

The invention discloses an auxiliary diagnosis method based on depth learning. The method comprises steps that S1, segmentation for raw corpus data is carried out to establish a word embedding query table; S2, a training set is generated based on key characteristic fields of electronic case data, the word embedding query table is utilized for digitalization, the convolutional neural network is utilized to generate an auxiliary diagnosis model; and S3, key characteristic fields of a newly-inputted electronic case are extracted, digital conversion is carried out through the word embedding querytable, the auxiliary diagnosis model is utilized for matching, and the matched diagnosis result is outputted. The invention further provides an auxiliary diagnosis system based on depth learning. Thesystem comprises a corpus data extraction module, a word embedding query table construction module, a historical electronic case data extraction module, a new electronic case data extraction module, asegmentation module, an electronic case digital module and an auxiliary diagnosis module. The method is advantaged in that the diagnosis result is timely and accurate, and doctors are effectively facilitated to carry out rapid condition diagnosis.

The invention relates to an automatic broadcast four-stage automatic switching device and an automatic broadcast four-stage automatic switching method. The device comprises a master broadcast control workstation, a standby broadcast control workstation, a master video server, a standby video server, a following camera and switching equipment, wherein a converter is arranged between the master andstandby broadcast control workstations and broadcast equipment, such as the master video server, the standby video server and the like, and controls the conversion of control links of the master and standby broadcast control workstations through GPI signals of the standby broadcast control workstation by means of an RS422 interface circuit; and the switching equipment is controlled by the broadcast control workstations. The automatic broadcast four-stage automatic switching device realizes the four-stage automatic switching of broadcast signals according to the order of the master and standbyvideo servers, the following camera and an emergency signal source, changes a conventional operation mode that broadcasters perform manual emergency switching according to the signal output conditionof the switching equipment, greatly reduces the working strength and mental stress of the broadcasters in television stations, realizes the automatic detection and quick processing of faults, effectively avoids the occurrence of broadcast accidents and comprehensively improves the security and reliability of a broadcast system.

The present invention discloses a method of optimizing the static safe and stable operation limit of an electric power system. The method comprises the steps of using a BPA electric power system calculating and analyzing software to analyze an original data report containing the network structure and operation data of an evaluated electric power system; then carrying out the BPA power flow calculation on the data in the original data report; if the BPA power flow calculation is not convergent, correcting a control variable or a constrained condition until the BPA load flow calculation is convergent; then carrying out the N-1 power flow calculation; if the N-1 power flow calculation is not convergent, carrying out the N-1 optimal power flow calculation until the N-1 optimal power flow calculation is convergent, and outputting an N-1 power flow calculation report; and then using the data in the N-1 power flow calculation report to substitute for the data in the original data report, so that the situation that the evaluated electric power system is located in a power grid operation manner of the static safe and stable operation limit is guaranteed.

The invention discloses a dynamic modeling method of a conventional substation based on the IEC standard, which includes the steps as follows: firstly, building an object-oriented model; secondly, obtaining an entity model through mapping and supplementation; thirdly, building a UCIM model of a master station system as per the entity model; and fourthly, updating the UCIM model. The method effectively solves the problems of communication collision and efficiency between communication of an information point oriented dispatching master station and communication of an object-oriented substation, improves the efficiency of technology upgrade and innovation of a great amount of conventional substations, and reduces working pressure of substations and management maintenance staff for dispatching automatization.

Provided is a vegetation maintenance machine. The maintenance machine comprises a base and a supporting piece fixed with the base. The supporting piece is of a hollow structure, and a water storage device, a liquid storage device, an atomizer and a top cover are arranged in the supporting piece, water is reserved in the water storage device and used for watering vegetation, and nutrient solution for supplying nutrient to the vegetation is reserved in the liquid storage device. The atomizer provides an atomization function for the vegetation. The top cover is arranged at the top of the supporting piece, and a plant growth lamp is arranged on the top cover and illuminates the vegetation. A controller is arranged on the base or the supporting piece and connected with an image collector and the like through a wire. The vegetation maintenance machine can achieve remote nursing of the vegetation and improves survival rate of the vegetation.

The embodiment of the invention relates to an incremental data extracting method, an incremental data extracting device and an incremental data extracting system. The incremental data extracting method comprises the following steps of: acquiring primary key information of an incremental data from a standby database; according to the primary key information, inquiring the entire incremental data in a main database in data synchronism with the standby database; and inserting the inquired entire incremental data into a target database. After adoption of the method, the device and the system for incremental data extraction, a lot of time and system resources can be saved and the incremental data extracting efficiency is greatly improved.

An embodiment of the invention provides a device and a method for treating refractory wastewater by coupling an electrocatalytic membrane with a three-dimensional electrode. The device mainly comprises a Ti4O7 / Ti filter element anode, a water-permeable separation mesh, a graphite felt cathode, C@Fe3O4 particles, an aeration pipe and a current-stabilized and voltage-stabilized power supply; and the device is in the shape of a cylinder on the whole, the Ti4O7 / Ti filter element anode is fixed at the center of the bottom of the device and a layer of water-permeable separation mesh wraps the surface of the anode, the graphite felt cathode clings to the inner wall of the cylinder, the aeration pipe is placed at the bottom of the device and beside the graphite felt cathode, the C@Fe3O4 particles fill the space between the water-permeable separation mesh and the graphite felt cathode to form the three-dimensional electrode, the graphite felt cathode and the Ti4O7 / Ti filter element anode are connected with a negative electrode and a positive electrode of the current-stabilized and voltage-stabilized power supply, respectively, and organic matter in the wastewater is degraded by a constant-current operation mode. In the device and the method provided by the invention, the Ti4O7 / Ti filter element anode is coupled with the C@Fe3O4 three-dimensional electrode system to enhance the treatment effect on refractory pollutants, and through the synergistic effect of the cathode and the anode, the efficiency of the system for oxidizing pollutants is improved, and the energy consumption for wastewater treatment is reduced.

The invention discloses a fire protection and security integrated intelligent video monitoring system suitable for open space and relates to the technical field of the fire protection and security. According to the system, international leading artificial intelligence and an image processing technique are combined, flames and smoke can be detected simultaneously, not only is a fire alarm function for the fire protection achieved, but also functions such as invasion alarm for the security are integrated, a plurality of technical indexes reach a domestic leading level, the alarm is timely, response time for identifying the flame and the smoke only needs 2-5 seconds, the fire alarm in a quite early stage can be generated timely and the monitoring system is stable and reliable, 24-hour operation in the smallest detection area of the flames and smoke detection in large space indoors can be achieved, and interference caused by various light sources, highlight backgrounds and changes of light rays can be filtered. Complex environment conditions such as spatial height, thermal barrier, explosion or poison have little limit to non-contact detection of targets, and an effective solution is provided for the fire detection and the invasion alarm for the security in the large space indoors, the open space and special places where the existing detecting methods are not effective.

The invention provides a method for improving load balance efficiency of an Nginx server. According to the method, a real-time load state of each rear end server is further considered on the basis of a traditional weighting polling load balance scheduling policy, and a task amount of each server is distributed according to the information processing capability of each rear end server and a real-time load state, so that the distribution of the task amount is more balanced and the load balance efficiency of the Nginx server is improved.

The invention discloses an automatic power grid scheduling system of intelligent unattended substations, comprising a scheduling center of the automatic power grid scheduling system and a plurality of unattended substations; an SDH (Synchronous Digital Hierarchy) self-healing type optical fiber ring network is paved between the scheduling center of the automatic power grid scheduling system and each unattended substation and is uniformly scheduled by the automatic power grid scheduling system; each unattended substation utilizes a cooperative work mode of upper and lower computers; a video monitoring module of each substation is linked with a guard environmental safety module, so as to make up inherent defects that a detection region of a traditional safety system is limited and the like; meanwhile, the automatic power grid scheduling system is used for assisting a supervisor of the automatic power grid scheduling system in monitoring and controlling, and the secure and intelligent levels of the substations are improved; the video monitoring module, the guard environmental safety module and a substation system are integrated; when meeting illegal invasion events, the scheduling system has the capability of giving real-time responses, a power grid is recombined and emergency and optimal scheduling is supplied, so that the whole intelligent level and the service quality of the substations are improved.

The invention discloses a high-yield reactor, and a production method and an application thereof. The high-yield reactor comprises a bioreactor and an ATF perfusion apparatus connected with the bioreactor. The protein production method using the high-yield reactor comprises the following steps: 1, connecting the ATF perfusion apparatus with the bioreactor, and completing an offline disinfection or online disinfection program; 2, inoculating seed cells to the bioreactor to a work volume according to the density, and sampling and detecting the number and biochemical indexes of the cells each 24h; 3, starting the ATF apparatus program for perfusion culture when the density of living cells increases to a certain density; and 4, collecting a culture solution in the bioreactor or filtered through hollow fibers in the ATF perfusion apparatus, and purifying to obtain required protein products. The high-yield reactor has the advantages of great shortening of the technological exploitation time of monoclonal antibody or fusion protein expression, production cost reduction, benefiting for the project registering and reporting and the project period shortening, and acceleration of the accomplishment application speeds of the biopharmacy industry.

The invention belongs to the technical field of computer vision, and particularly relates to a labor protection article wearing condition detection and identity recognition method based on deep learning. The method comprises the following steps of collecting data to make a special data set for worker labor protection article wearing condition detection and identity recognition; preprocessing, labeling and dividing the data set; designing and training each depth model based on multiple target detection algorithm principles; carrying out model integration on a plurality of depth models by adopting a Stacking Enable method; and training a face recognition model and a pedestrian re-recognition model based on deep learning by using the public data set, constructing a face recognition and pedestrian re-recognition comparison database of a work site, performing adaptive optimization on the models in combination with the image database, and performing detection and monitoring. The safety of workers in a construction site is improved, safety accidents are reduced, good social value and commercial value are achieved, and the application prospect is wide.