83results about How to "Reduce the risk of miscalculation" patented technology

The embodiment of the invention discloses an anti-cheating method and anti-cheating system of an application program. The anti-cheating method and the anti-cheating system are used for against the cheating behavior targeting the application program with low misjudgment risk. According to the embodiment of the invention, the method comprises the following steps of: obtaining the operation data of the operation program operated by a plurality of users within a preset time quantum; extracting the attribute information adaptive to the preset anti-cheating analysis dimension of each user according to the operation data of each user; dividing the plurality of the users into malicious users and un-malicious users according to the attribute information of each user; and clearing the data of the application program according to the quantity of the divided malicious users.

The invention provides a static analysis method and a static analysis device for an Android application program and relates to the technical field of security detection. The static analysis method includes: S1, unzipping an application program to be detected to obtain a Smali file; S2, traversing the Smali file to acquire a source code information and structuring a control flow diagram and a data flow diagram of the source code information; S3, traversing the control flow diagram and the data flow diagram according to a malicious behavior judging engine to respectively match an API (application program interface) of the application program with another API in a predefined malicious behavior library and marking the successfully matched API of the application program as the malicious behavior API; S4, calculating the malicious measuring value of the application program to be detected; S5, matching the malicious measuring value with a predefined malicious degree index to acquire risk level of the application program to be detected. Through behavior analysis and comprehensive judgment of combined rules, misjudgment rate of virus detection upon the Android application program can be reduced.

The invention provides a method for quickly discovering changed land by adopting unmanned aerial vehicle remote sensing images. The method comprises the steps of firstly performing primary characteristic point retrieval on old orthoimages; secondly performing aerial photography by utilizing an unmanned aerial vehicle to obtain unmanned aerial vehicle images of a landform, and obtaining search comparison domains; thirdly preliminarily determining whether the landform is changed or not through characteristic point matching; and fourthly performing comparative analysis through a spectrum and texture dual-mode comprehensive comparative detection method to accurately determine whether suspicious land appears or not. Compared with a conventional technology, the method has the advantages that a suspicious ground object can be quickly and automatically discovered without arranging a large amount of ground control points and splicing the unmanned aerial vehicle images, so that the precision is high, and the time and labor are saved.

The invention provides a method and device for detecting brake air tightness. The device comprises an air inlet, an exhaust port, a test port, a first electromagnetic valve, a second electromagnetic valve, a third electromagnetic valve, a fourth electromagnetic valve and a direct voltage sensor. The method includes the steps that the first electromagnetic valve, the second electromagnetic valve and the fourth electromagnetic valve are controlled to be connected, the third electromagnetic valve is closed, and then a test piece is inflated from the air inlet; the first electromagnetic valve is closed when the test piece reaches preset pressure after being inflated; the direct voltage sensor is utilized to test air pressure leakage amount in an air path, and if pressure drop in preset time exceeds a preset pressure value, it is determined that the test piece leaks; if pressure drop in the preset time does not exceed the preset pressure value, the second electromagnetic valve is closed, and a differential pressure sensor is utilized to test the pressure difference between the test piece and a standard cavity, if the pressure difference is smaller than the preset pressure value, the test piece does not leak, and if the pressure difference is larger than the preset pressure value, the test piece leaks. According to the method and device, air leakage detection efficiency of the air path is improved, and detection periods are shortened.

The invention discloses terminals and a terminal building system based on an electric load management intelligent recognition system. The electric load management intelligent recognition system comprises a frequency selection circuit; the frequency selection circuit comprises an active band pass filter formed by an integrated operational amplifier and an RC network; the terminals comprise a power interface, a current sensor, a voltage sensor ASSS, an A / D conversion module, a communication module, a relay control module and output jacks; the terminal building system comprises a server, a communication network, multiple terminal electric devices, multiple electric load management intelligent recognition system terminals, and multiple concentrator. Through the above technical scheme, through carriers of the frequency selection circuit, terminals formed by the recognition system can be recognized, and the terminal building system can realize power supply, power control, power limit, meter reading and charging; and electric detection and measurement can be carried out by the multiple terminal electric devices, timely alarming and forced power cutoff can be carried out, records can be exchanged with the server, and detection, control and safety are improved.

The invention discloses a partial discharge fault comprehensive diagnosis method for a GIS device. The method includes the following implementation steps that firstly, three sets of diagnosis results output through an ultrahigh frequency detection method, an ultrasonic detection method and an SF6 componential analysis method of the partial discharge fault comprehensive diagnosis method for the GIS device respectively are respectively output; secondly, a probability matrix is built according to the three sets of diagnosis results; thirdly, a comprehensive diagnosis probability vector is acquired by multiplying the probability matrix by an optimal weight vector; fourthly, the maximum value among the probability value of point discharge, the probability value of suspension discharge and the probability value of particle discharge in the comprehensive diagnosis probability vector is found, the discharge type corresponding to the maximum value serves as a partial discharge fault comprehensive diagnosis result and is output. Early fault hidden danger of the GIS device can be diagnosed, the detection ratio and the accuracy of fault diagnosis are improved, the scientific basis and the guidance are provided for condition overhaul and maintenance of the GIS device on site, and the partial discharge fault comprehensive diagnosis method has the advantages that misjudgment risks are small, and diagnosis accuracy and stability are high.

The invention discloses a dynamic drilling data abnormal trend detection method. The method comprises the following steps: s1, reading and converting real-time data; s2, integrating five different algorithms to calculate trend change indexes of the time series drilling data; s3, obtaining a final risk index through joint voting calculation of the trend detection indexes; s4, automatically updatingan alarm threshold value along with the change of the data by utilizing a dynamic threshold value calculation method; and s5, judging whether the final risk index exceeds a threshold value or not, and deciding whether to trigger an alarm or not. According to the invention, the data driving method is used for detecting the abnormal trend of the real-time data, so that the uncertainty of the real-time data can be solved; the invention can be used for monitoring drilling data exception caused by various reasons and warning risks in a physical application range, and the calculation process is simple and rapid; the invention can be applied to accurately evaluating data exception caused by working condition change in the drilling process so as to effectively realize risk in-time early warning in a drilling management system and reduce the misjudgment probability.

The invention claims a detection device based on fusion classification of lung medical images, the device comprising: an obtaining section obtaining a lung CT scan image of the patient; a preprocessing unit that performs preprocessing to obtain a binary gray-scale image; a morphological etching and expansion part, which performs morphological etching operation to obtain an etched image, and performs expansion operation on the binary gray image to obtain an expanded image; an arithmetic unit that performs an on-off operation on the etched image to obtain an on-off arithmetic diagram, and performs a closed operation on the expanded image to obtain a closed arithmetic diagram; a Fourier transform unit, which obtains a Fourier transform value; a lung medical image fusion classification module,which is used for fusion classification and discrimination of lung CT scanned images; and a focus detection unit configured to determine a suspected lesion area of the patient by combining the Fourier transform value with a fusion classification discrimination result. The detection device can reduce the complexity and improve the identification and diagnosis accuracy of lung focus.

The embodiment of the invention relates to the technical field of communication, and discloses an anti-fraud system and method, a server and a storage medium, and the anti-fraud system comprises a suspicious call analysis module, a network data obtaining module, a fusion analysis module and a warning service module. The suspicious call analysis module is used for identifying the call signaling corresponding to the personal account according to a first preset model and recording suspicious call data; the network data acquisition module is used for discriminating the personal internet surfing data according to a second preset model and acquiring network data for accessing a malicious website; the fusion analysis module is used for performing fraud identification on the suspicious call data and the network data according to a third preset model; and the warning service module is used for intercepting and sending out early warning after judging that the personal account is suspected to becheated. By aggregating the call data and the internet surfing data, the fraud event is more accurately and comprehensively recognized, comprehensive reminding and interception are achieved, and the fraud event prevention and control capacity of family users is effectively improved.

The invention discloses a semi-supervised image classification method based on dictionary deep learning. The method comprises the following steps of: preparing the raw materials; constructing a cost function L1 (Zl, Yl) of the labeled data and a cost function Lu (Zu, P) of the unlabeled data according to a Softmax cost function of the deep neural network; constructing an overall model function according to the cost function L1 (Zl, Yl) of the labeled data and the cost function Lu (Zu, P) of the unlabeled data; And using an alternating optimization algorithm to train the overall model, whereinthe training optimization process comprises joint class estimation based on dictionary learning and Softmax network information, and joint learning of a neural network and unlabeled class estimation.According to the method, dictionary learning is combined with the Softmax classifier of the deep neural network, so that the exploration capability of the deep neural network on unlabeled data is enhanced, and the network feature learning capability and the classifier learning capability are greatly improved. The method is suitable for the field of computer vision or mode recognition.

The embodiment of the invention discloses a control method, a master control device and a zero-crossing point detection circuit. The control method applied to an electronic device includes the following steps that: input voltage signals are monitored; whether a zero-crossing point appears in a preset period is judged based on the input voltage signals, so that a first judgment result is obtained;the current state of the electronic device is determined according to the first judgment result; and when it is judged that the current state of the electronic device is a power-off state, a preset discharge component in the electronic device is controlled to perform discharging. With the control method, the master control device and the zero-crossing point detection circuit provided by the technical schemes provided by the embodiments of the invention adopted, the accuracy of zero-crossing detection can be improved, and the electronic device can consume residual power in time.

The invention discloses a listed company performance comprehensive evaluation method. The listed company performance comprehensive evaluation method comprises steps such as model establishment, model resolution, model selection, condition density prediction, randomness evaluation, and the like. The effects of the listed company performance comprehensive evaluation method are that a neural network quantile regression model of the listed company performance comprehensive evaluation is established, the priorities in the two aspects of the neural network model and the quantile regression model are combined, a variation rule of the performance level of the listed company can be depicted exactly, and powerful functions are presented. As a standard gradient optimization algorithm of the neural network quantile regression model for the performance comprehensive evaluation of the listed company is given, the calculation speed of the model is improved under the prediction of not affecting the model estimation accuracy. As an AIC standard selected by the neural network quantile regression model of the listed company performance comprehensive evaluation is established, a suitable model structure can be selected, and the model is prevented from being too complex and falling in an over-fitting awkward situation efficiently.

The invention discloses a method and a server for restricting malicious information propagation, and belongs to the field of computer. The method comprises: receiving a request of a first social networking services (SNS) community user to comment or to reply for SNS community contents of a second SNS community user; sending the contents of a comment or a reply, or information of the first SNS community user to the second SNS community user or a friend of the second SNS community user to perform verification; and rejecting or approving the request of the first SNS community user to comment or to reply for SNS community contents of the second SNS community user according to a verification result. The server comprises a reception module, a sending module and a processing module. Through the above scheme, a person being visited or a friend of the person being visited can actively verify the comment or the reply before malicious information is published, so that judgment of the malicious information is improved, and misjudgment risk is reduced.

The invention discloses a static target restraining method for a ground surveillance radar system. The static target restraining method for the ground surveillance radar system comprises the steps that processed target detection information of the front end of a radar is obtained; interframe accumulation processing is carried out on the target detection information, and a two-dimensional image isformed according to the target detection information after the accumulation; edge extraction is carried out on the two-dimensional image, and target detection information which belongs to a same target after the extraction is obtained; static target area judgment processing is carried out on the target detection information which belongs to the same target, and a static target area range is markedout; and restraining processing is carried out on the target detection information in the static target area range, and target detection information after the static target restraining is obtained. According to the static target restraining method for the ground surveillance radar system, the problem that the ground surveillance radar system cannot accurately mark off the static target area due to variation of detection scenes and erecting sites is solved, the restraining speed of a static target and the detection precision of a moving target are improved, and a reliable data foundation is provided for subsequent track processing.

The invention discloses an electronic paper detection method and an electronic paper detection device. The electronic paper detection method comprises the steps of: controlling electronic paper to sequentially lighten a display picture according to a pre-written display image set; acquiring a picture display quality parameter of the electronic paper under the current display picture; if all picture display quality parameters do not exceed a corresponding picture display quality parameter range, determining that the electronic paper under the current display picture is qualified; and if the atleast one picture display quality parameter exceeds the corresponding picture display quality parameter range, determining that the electronic paper under the current display picture is unqualified. According to the electronic paper detection method, objective detection of whether the parameter value meets the requireme111385604Ant or not replaces subjective observation of the picture to judge thedisplay quality of the electronic paper, therefore, the probability of misjudgment is reduced, the accuracy and reliability of an electronic paper detection result are improved, the steps of observing the picture by detection personnel are reduced, the time consumed for detecting the electronic paper is shortened, and the production efficiency of the electronic paper is improved.

The invention provides an elevator traction capacity detection method. The method includes the following steps of (1) lifting an empty elevator car to the highest station; (2) setting a lifting torque criminal value tup1-limit of a traction machine, making tmax1-1 equal to 0, and making tmax1-2 equal to 0; (3) setting a lifting height criminal value delta hup1-limit of the elevator car, (4) setting a maintaining torque ttest1-limit which is smaller than or equal to tup1-limit; (5) slowly lifting the elevator car; (6) carrying out the step (9) if the lifting height reaches the lifting height criminal value delta hup1-limit and if the lifting height does not reach the lifting height criminal value delta hup1-limit, maintaining the torque of the traction machine and detecting whether a traction rope slides relative to a traction wheel when the drive torque ttest1 is equal to ttest1-limit; (7) if the traction rope slides relative to the traction wheel, recording the torque, making tmax-1 equal to ttest1-limit and carrying out the step (9) and if the traction rope does not slide relative to the traction wheel, making tmax1-2 equal to ttest1-limit and carrying out the step (8); (8) if ttest1-limit is smaller than tup1-limit, increasing the lifting torque upper limit of a single test by a fixed value delta t1 and repeating the steps from the step (4) to the step (7), and if ttest1-limit is equal to tup1-limit, carrying out the step (9); (9) driving the elevator car to descend and reach the highest station again and braking to finish the test; (10) processing data obtained through detection.

The invention discloses a method for selecting an alternate airport on a flight route segment. The method comprises the following steps: screening a plurality of candidate alternate airports based ona flight plan of a current route segment; dividing the plurality of candidate alternate airports into a return voyage area alternate airport and a to-be-flied area alternate airport; according to theflight parameters, the current position and the wind power parameters of the aircraft, and the position information and the wind power parameters of the candidate alternate airports, calculating the distances from the current position to the candidate alternate airports, the required flight duration and the fuel consumption; selecting an alternate airport with the shortest flight duration from thealternate airports in the return voyage area and the to-be-flied area; and providing related alternate airport information including flight duration and fuel consumption. According to the invention,the optimal alternate airport selection can be accurately and automatically selected, and enough visual and comprehensive prompts are provided for pilots, so that the pilots can easily make the optimal standby landing selection in a convenient and simple operation mode.

The invention relates to an optical lens. The optical lens comprises a first lens, a second lens, a third lens, a fourth lens and a fifth lens, wherein the first lens has negative focal power, the object side surface of the first lens is a concave surface, and the image side surface of the first lens is a concave surface; and the first lens, the second lens, the third lens, the fourth lens and thefifth lens are sequentially arranged from the object side to the image side so as to form an optical lens with high resolution, small aperture and large camera shooting range. Meanwhile, the opticallens can meet a requirement of miniaturization.

The invention discloses a main pump shaft seal water leakage monitoring device. The device comprises a valve cooling control system, a monitoring sensor and an alarm loop. A monitoring loop is arranged in the valve cooling control system, the monitoring sensor is provided with a temperature sensor and / or a humidity sensor, and the signal output end of the monitoring sensor is connected with the signal input end of the monitoring loop; the monitoring sensor is used for collecting temperature signals and / or humidity signals at a main shaft seal and transmits the signals to the monitoring loop; and the signal input end of the alarm loop is connected with the signal output end of the monitoring loop, and the alarm loop is used for giving out an alarm motion according to the received alarm signals transmitted by the monitoring loop. According to the device, the main pump shaft seal water leakage condition can be rapidly, efficiently and accurately monitored, and stability and reliability of a direct-current valve cooling system and even a converter valve are ensured.

The invention discloses a method and equipment for identifying the orientation of an intracerebral fragmentation electrode in a craniocerebral medical image. The method comprises the following steps: firstly analyzing the craniocerebral medical image, obtaining the orientation coordinates of the intracerebral fragmentation electrode according to an analysis result, and building an individual craniocerebral model according to the analysis result; calibrating the obtained azimuth coordinates of the intracerebral fragmentation electrodes on an individual craniocerebral model to form individualized intracerebral electrode positioning data; and encoding the individualized intracerebral electrode positioning data to form an encoded data packet, and directly displaying the orientation of the intracerebral fragmentation electrode on a display screen of mobile equipment by the mobile equipment according to the encoded data packet. The invention further discloses the equipment corresponding to the method. According to the invention, the spatial orientation of the intracerebral fragmentation electrode in the brain can be visually displayed on the mobile device, a doctor does not need to compare an image to carry out manual judgment, the judgment accuracy is improved, and the manual judgment time is shortened; and remote inquiry can be carried out conveniently.

The invention provides a communication security maintenance device based on power edge gateway equipment, and relates to the technical field of electrical equipment. The device comprises a hardware part and a software part, wherein the hardware part comprises an encryption identification unit and communication interface equipment thereof, and is used for carrying out information reading, white list qualification comparison and pre-classification on communication data transmitted by lower-layer equipment, and is in data connection with a software port; and the software part comprises a basic discrimination module, a periodic white list equipment inspection module and an information detection and fault discrimination module. According to the invention, on the basis of a machine learning algorithm and hardware encryption feature coding information, unique identification and security evaluation of the bottom sensing device are realized, periodic white list updating is arranged, and the security requirement of the whole device data transmission process is met. And on the basis, a function of realizing multi-gateway matrix type auxiliary security evaluation in the edge gateway at the same level is designed, so that the purpose of improving the judgment accuracy is achieved.

The invention discloses an optimal driving speed control system for a valuable transport fleet and belongs to the technical field of transport fleet speed control. The system comprises a positioning navigation satellite, a plurality of vehicle-mounted satellite positioning modules, a plurality of inter-vehicle data communication modules, a pilot vehicle operation and storage module, a pilot vehicle vibration measurement module, a pilot vehicle image display and voice broadcast unit, a following vehicle operation and storage module, a following vehicle vibration measurement module and a following vehicle image display and voice broadcast unit. According to the invention, a driver of the following vehicle can display the current horizontal driving speed of the following vehicle through the image display and voice broadcast unit; when the following vehicle passes through the concave-convex part of the vertical height level value of the known convex-concave point in front, the allowed maximum running speed and the recommended following speed value of the current following vehicle quickly judge the optimal running speed of the following vehicle where the following vehicle is located, acorresponding voice broadcast prompt is obtained, and the overall average running efficiency of the motorcade is improved.

The invention discloses a manufacturing method for manufacturing a 3D printing simulation board of a funnel chest correction board. The manufacturing method comprises the following steps: (a) a 3D thoracic cavity model for a funnel chest patient is established; and (b) the shape of the contour of a corrected thoracic cavity of the funnel chest patient is predicted so as to establish a 3D printingmodel of the corrected predicted thoracic cavity contour; (c) a postoperative thoracic cavity contour model is established according to the two models; (d) the implantation position of the funnel chest correction board is found through the postoperative thoracic cavity contour model, and a 3D printing model of a simulation board is established; and (e) the 3D printing model of the simulation boardis printed through a 3D printing device to obtain the 3D printing simulation board customized according to the shape of the thoracic cavity of an individual patient and the radian of the two sides ofthe thoracic cavity, thus a physician can manufacture the funnel chest correction board customized according to the shape of the thoracic cavity of the individual patient and the radian of the two sides of the thoracic cavity by taking the 3D printing simulation board as a template, the funnel chest correction board is precisely implanted in the correct position, turning over is conducted in place once, and the operation time is dramatically shortened.

The embodiment of the invention discloses a steel wire rope defect identification method, which comprises the steps of carrying out image preprocessing on a target steel wire rope image to obtain a preprocessed image; downloading a newest target defect model from a knowledge base, identifying the defect of the preprocessed image by using the target defect model in combination with a deep learning technology, and uploading a detection result to a data platform once the defect is found, thereby achieving the effects of improving the detection efficiency of the steel wire rope defect, reducing the misjudgment risk and saving the labor cost. In addition, the invention further provides a steel wire rope defect recognition device, computer equipment and a storage medium.

The invention relates to the technical field of vehicles, in particular to an off-hand driving monitoring method and device as well as a vehicle and a storage medium. The method comprises the following steps: obtaining temperature parameters of the surface and / or the nearby position of a steering mechanism; and judging whether or not the vehicle is currently in an off-hand driving state accordingto the temperature parameters. According to the off-hand driving monitoring method and device as well as the vehicle and the storage medium provided by the invention, the temperature parameters replace pressure or torsion parameters to serve as judgment basis of off-hand driving, and whether or not the driver adopts off-hand driving can be detected without applying force to the steering mechanismby the driver, thereby achieving a low misjudgment risk, small data calculation amount and low manufacturing cost.

The embodiment of the invention provides a JSON file differentiation comparison method and device, and relates to the technical field of data processing. Keys of a first key value pair in the first test sample file are traversed, a second key value pair which is the same as the keys of the first key value is found from the second test sample file, and the value of the first key value pair is compared with the value of the second key value pair, so that a differential comparison result of the first test sample file and the second test sample file is obtained. According to the scheme, the sorting pre-operation on the JSON test sample file can be reduced, and the technical problems of low accuracy of disordered comparison and misjudgment are solved; meanwhile, the identification and statistics of the comparison result do not need to be manually participated for the second time, so that the misjudgment risk of manual operation can be reduced and the working efficiency can be improved; the comparison result is output, so that tracking and backtracking of comparison differences of JSON test sample files can be facilitated; in addition, compared with the prior art, only one-time traversal is needed, the occupation of calculation resources in the comparison process can be reduced, and the text comparison speed is increased.

The invention relates to a technology for detecting chlorobenzene compounds and particularly relates to a method for determining the chlorobenzene compounds in a plastic product. The method comprises the following steps: (1) sample preparation: selecting samples and crushing the samples into uniform samples with a high polymer material sample crusher; (2) extraction: adding the samples into a solvent, tightly plugging, extracting in a microwave extractor, filtering extract liquor, blowing with nitrogen and waiting for purification; (3) purification: activating a PA (polyamide) pillared with a solvent, adding the samples into the solvent for dissolution, eluting with the solvent, blowing to be dry with nitrogen and metering volume with propyl alcohol; (4) gaschromatographic mass spectrometric determination: respectively absorbing a to-be-determined standard solution and a to-be-determined sample solution, injecting a gaschromatographic mass spectrometer for determining and finally calculating the content of chlorobenzene compounds in the samples. The determination method provided by the invention has the advantages that 12 types of chlorobenzene organic compounds in plastics can be simultaneously detected, therefore, the sensitivity is relatively high, the limit of detection is low, the degree of precision is good, the detection period is short, the anti-interference capacity is strong, and the method is fast, stable and accurate.

The invention provides a pump valve fault detection method. The method comprises the following steps: constructing a pump valve fault index system reflecting the characteristics of vibration signals of a pump valve state; establishing an LSTM-AE model by using an LSTM depth model according to an auto-encoder principle, and setting a fault discrimination threshold value; inputting the training data into the LSTM-AE model based on the pump valve fault index system so as to train the LSTM-AE model; and detecting data of a pump valve to be detected are input into the trained LSTM-AE model based on the pump valve fault index system, an output value obtained by the LSTM-AE model is compared with a set fault judgment threshold value, and therefore whether the detected pump valve breaks down or not is judged, and a first judgment result is output. According to the method provided by the invention, the universality and accuracy of practical application of the model are improved.

The invention provides an auxiliary image generation method, a remote control method and a remote control device for a working machine, and the working machine comprises a camera used for collecting a first environment image, a plurality of execution mechanisms capable of moving relatively, and a sensor used for collecting posture features of the execution mechanisms. The auxiliary image generation method comprises the following steps: acquiring attitude features of an execution mechanism; obtaining a first environment image collected by a camera; and based on the attitude features and the first environment image, generating an auxiliary image, wherein the auxiliary image is used for representing the environment information of the working machine and the attitude information of the working machine. According to the invention, the auxiliary image is generated based on the attitude feature and the first environment image and is used for simultaneously representing the environment information of the working machine and the attitude information of the working machine on the same image, so that an operator can be accurately assisted and guided to operate the working machine in a scene of manually operating the working machine, the misjudgment risk of the operator is reduced, moreover, the number of uploaded video paths is small, network resources can be saved, and the image quality is improved.

The invention discloses a roller type dust removing device. The roller type dust removing device comprises an installing frame. The installing frame is rotationally connected with the middle of an installing plate through a horizontal rotating shaft, the length direction of the installing plate is perpendicular to the horizontal rotating shaft, the two ends of the installing plate in the length direction are each rotationally connected with a roller installing base, one ends of the roller installing bases are connected with the ends of the installing plate and rotate in the direction parallel to the length direction of the installing plate as the axis, a rotary roller is arranged between the other ends of the roller installing bases located at the two ends of the installing plate, and sticky coatings are arranged on the surfaces of the rollers. The roller type dust removing device has the beneficial effects that fine particles on the surfaces of liquid crystal modules are automatically removed through the horizontal rotating shaft and the rollers with the rotating radius, self-adaptation adjustment can be carried out on the liquid crystal modules at different positions and different inclination angles, the purposes of the high cleaning speed, high efficiency and high cleanliness are achieved, the running time of a software background is shortened, and the misjudgment risk of detection equipment is reduced. The device can achieve self-adaptation adjustment, manual intervention is not needed, and production efficiency is greatly improved.