175results about How to "Improve discrimination accuracy" patented technology

The auxiliary safety operation equipment based on omnibearing computer vision includes an omnibearing vision sensor for acquiring omnibearing video information outside and inside the vehicle, and an auxiliary safety operation controller for detecting fatigue operation and warning on fatigue operation. The omnibearing vision sensor installed in the right of the driver's seat and connected to the auxiliary safety operation controller detects the facial state, eye state and mouth state of the driver and the steering wheel state, monitors the vehicle direction, speed, etc, and warns the fatigue operation in case of detecting fatigue operation state. The present invention detects the characteristic parameter of fatigue operation state, and has high judgment precision and high measurement precision.

The invention discloses a malicious traffic detection method. The method comprises the following steps: correspondingly establishing malicious and normal data sample libraries by using obtained malicious and normal data traffic samples; executing a data cleaning operation and a preprocessing operation on the data sample libraries in sequence to obtain training data, and constructing a traffic detection model by using the training data and a deep learning algorithm; judging whether to-be-measured data traffic contains malicious data by using the traffic detection model; and if so, sending alarminformation carrying the to-be-measured data traffic belonging to malicious data via a preset oath. Feature learning and training are performed by using the malicious and normal data traffic samplesvia the automatic learning property of the deep learning algorithm, the feature information extraction operation is completed without consuming precious human resources, thereby improving the improving the work efficiency and improving the discrimination of the malicious traffic. Precision. The invention further discloses a malicious traffic detection system and apparatus and a computer readable storage medium, which have the above beneficial effects.

The invention relates to the field of computer vision and discloses a computer vision-based vehicle damage distinguishing method so as to solve a problem that a manual distinguishing approach adopted in conventional technologies is incomplete, inaccurate and inefficient in distinguishing operation. The computer vision-based vehicle damage distinguishing method comprises the following steps: in step a, a binocular image collection system is calibrated; in step b, the binocular image collection system is used for collecting images in a monitored area, and a depth map for the collected images is obtained; in step c, a convolutional neutral network is used for subjecting the depth map to characteristic extraction training operation, and a vehicle damage condition discrimination model is trained; in step d, the vehicle damage condition discrimination model is used for determining a damage degree based on collected vehicle images. The computer vision-based vehicle damage distinguishing method is suitable for distinguishing vehicle damage.

The invention discloses a real-time dynamic judgment method and device for a road traffic state. The method includes the following steps: during a set period, GPS information transmitted by a set number of floating vehicles in a selected geographic area is acquired in a real-time manner in a data acquisition period; unreasonable GPS information is screened in the acquired GPS information; geographic position points, corresponding to geographic position information included in each piece of GPS information which is reserved after the screening, are projected respectively to road segments of an electronic map corresponding to a selected geographic area; and according to acquisition times and floating-vehicle identification information, which are corresponding to the geographic position points projected to each road segment, and the length and limit speed of a corresponding road segment, the traffic state of the corresponding road segment in the set period is determined. The real-time dynamic judgment method and device are capable of covering all road segments of a selected geographic area, large in applicable scope and high in judgment precision.

The invention relates to a face-identification-based living body determination method and equipment. The method comprises: obtaining a face image of a sample, carrying out first living body detection and second living body detection on the face image, and calculating a first living body detection score and a second living body detection score of the sample; determining a score fusion strategy of fusion of the first living body detection score and the second living body detection score; obtaining a face image of a to-be-identified object, carrying out first living body detection and second living body detection on the face image, and calculating a first living body detection score and a second living body detection score of the to-be-identified object; according to the score fusion strategy, calculation a fusion score based on the first living body detection score and the second living body detection score of the to-be-identified object; and carrying out living body determination on the to-be-identified object based on the fusion score. Besides, a first face feature is extracted according to a mirror face reflection characteristic in the first living body detection; and a second face feature is extracted according to face key point changing in the second living body detection.

The invention discloses a document image classification method and device, and belongs to the field of computer vision. The method comprises: training a text feature vector extraction model and an image feature vector extraction model respectively, extracting a fusion feature vector of a document image in an embedded feature mode that text feature vectors and image feature vectors are fused, and classifying the document image based on the similarity of the fusion feature vector. According to the method, various document images can be quickly registered and classified, the service process can be greatly simplified, the OCR API can be simplified, all documents can be recognized through one API, and permanent use of one-time access is truly achieved.

The invention relates to a seafloor instability evaluation method. The method comprises: firstly, selecting out critical geological factors influencing seafloor slope stability according to geological environment conditions of seafloor slope occurrence and basic conditions of seafloor slope instability; secondly, selecting out critical induction factors influencing seafloor slope stability through a decomposition mode and a decomposition effect of natural gas hydrates, and constructing a seafloor slope stability influence factor index system; thirdly, constructing a seafloor slope geomechanics model by generalizing the seafloor slope stability influence factor index system; and finally, based on a Mohr-Coulomb failure criterion, proposing a mathematic method for evaluating seafloor slope stability, i.e., an improved finite element strength reduction method, and analyzing an evolutional process of seafloor slope stability and seafloor slope instability induced by decomposition of natural gas hydrates. The method is scientific in principle, high in calculation precision and short in calculation time, considers the induction factors of seafloor slope instability caused by decomposition of natural gas hydrates for the first time, gives out a deformation process of slope instability in combination with the geomechanics model and the finite element numerical analysis method, and has important practical application values.

The present invention provides an urban vegetation classification method based on unmanned aerial vehicle images and reconstructed point cloud. The method comprises the steps of: performing point cloud reconstruction of original unmanned aerial vehicle images; generating nDSM (normalized digital surface model) information of a research area; performing vegetation index calculation based on visiblelight; and performing classification discrimination of image objects. The method provided by the invention reconstructs point cloud of the research area based on a structure from motion (SFM) and cluster multi-view stereo (CMVS) and based on a patch-based multi-view stereo (PMVS) algorithm, performs filtering and interpolation to generate a digital elevation model (DEM) of the research area and the nDSM, and combines image spectral information to perform classification extraction of urban vegetations with different heights; an image analysis method facing the objects is employed to achieve differentiation of the categories of vegetations with different heights according to spectral information such as the nDSM information, normalized green-red difference indexes (NGRDI) and visible lightwave band difference vegetation indexes (VDVI) so as to greatly improve the differentiation precision.

The invention discloses a combined type intrusion detecting method on the basis of data fusion of an improved core vector machine, which belongs to the technical field of network safety. The method comprises the following steps: extracting characteristic data of all time monitoring points required by misuse intrusion detection and abnormal intrusion detection from a network safety equipment log of a target network; respectively constructing the characteristic data into black and white list data sample subsets; training the black and white list data sample subsets, thereby respectively obtaining a primary misuse intrusion detection model and a primary abnormal intrusion detection model; realizing the data fusion of the primary misuse intrusion detection model and the primary abnormal intrusion detection model through a D-S evidence theory, thereby obtaining a combined intrusion detection model and a combined confidence interval of the combined intrusion detection model; and obtaining a final detecting result. The method provided by the invention has excellent properties at the aspects of increasing the instantaneity of an intrusion detecting system, reducing the missing report rate and false report rate of the intrusion detecting system and increasing the generalization of the intrusion detecting system.

The invention discloses a spam mail filtering system and method based on clusters. The system comprises a cluster module, a mail training module, a characteristic base storage module and a mail filtering module. The method includes that S1.1 training mails and unmarked mails are obtained from a mail backup system; S1.2 a cluster preprocessing module preprocesses the mails; S1.3 a cluster analyzing module divides the preprocessed mails into different clusters; S1.4 a cluster central computing module computes vector expression of the clusters; S1.5 the training module studies marked mails in the clusters and updates characteristic bases corresponding to the clusters; S2.1 mails required to be filtered are obtained from a mail system; S2.2 a mail class attribute distinguishing module calculates the cluster most similar to mail content; S2.3 a mail characteristic extraction module conducts characteristic extraction on the mails to be distinguished; S2.4 a mail distinguishing module shows a distinguishing result according to the mail characteristics and the corresponding characteristic bases. The system and method has the advantages of being high in extraction speed, high in accuracy and good in effect.

The invention provides a multi-information online detection system for a machine tool multi-source energy consumption system, which is composed of a hardware part and a software system, wherein the hardware part comprises a related power sensor, a transmission and conversion interface and an upper computer terminal used for receiving and processing information; and the software system comprises an energy efficiency real-time detection module, a system parameter configuration module, an energy efficiency analysis module and an energy efficiency analysis result history query module. The system provided by the invention can detect and monitor energy consumption dynamic state information of a main power system and all auxiliary systems of a machine tool on line, and can detect and monitor input energy, effective energy, process power, transient power, the workpiece machining process energy utilization ratio, the specific energy efficiency and the equipment effective utilization ratio in the whole workpiece machining process on line. The detection system is of an integrated design, convenient and practical, a detection and analysis system is provided for improving various parameters of the energy efficiency of a mechanical machining system with high volume and wide scale, and a support tool is provided for enhancing state analysis, management control and energy conservation optimization for energy consumption and energy efficiency in the mechanical machining process.

The invention belongs to the bridge monitoring technical field and provides a bridge structure modal parameter authenticity and falsity discrimination method and a terminal device. The method includesthe following steps that: the monitoring data of the structure of a bridge under an environmental excitation condition are obtained through data acquisition devices disposed in a plurality of measuring points of the bridge; the NExT (natural excitation technique) is adopted, at first, the cross power spectrum density function of the monitoring data is calculated, and Fourier inverse transformation is performed on the cross power spectrum density function, so that the cross-correlation function of the monitoring data is obtained, the cross-correlation function is adopted as the input data of amodal parameter identification algorithm, so that modal parameter identification can be carried out, and a stability diagram can be drawn; the auto-power spectrum density function and cross-power spectrum density function of the monitoring data of the measuring points are calculated, the auto-power spectra and cross-power spectra of all the measuring points are summated, and the auto-spectra andthe cross spectra are summated in the same order of magnitude, so that the sum function of the power spectra can be obtained; and a diagram of the sum function, and true and false modal parameters canbe discriminated according to the stability diagram of the sum function. According to the method, the stability diagram method and a power spectrum summation peak method are combined to perform bridge structure modal parameter authenticity and falsity discrimination, so that discrimination accuracy can be improved.

The invention discloses a highway congestion level judgment method based on multi-model fusion. The method comprises the steps of (S1) acquiring highway traffic history data and performing normalization processing, (S2) presetting a number K of congestion levels according to normalized data, carrying out K-means clustering analysis on the data to obtain a congestion level of the historical data, which is a judgment model, (S3) classifying and analyzing normalized highway traffic real-time data by using different classification algorithms according to the congestion level of the historical data, and obtaining multiple congestion levels of the real-time data, which is a classification model, and (S4) fusing the classification model by the judgment model and obtaining the real-time congestionlevel of a highway. According to the method, the historical data is used as a model, a plurality of classification algorithms is used to simultaneously train the model, results obtained by the modelsare fused, and the discrimination accuracy is effectively improved.

Disclosed is a road traffic state estimation method based on microwave radar data. The method includes the steps that firstly, the maximum traffic capacity of road segments is acquired through statistics, and saturability is estimated; secondly, the mean value and standard difference of cross section speeds of a traffic flow lasting five minutes on each road segment are calculated through statistics; thirdly, an individualized subordinating degree function is automatically generated for each road segment; fourthly, a smoothness constant, a slow moving constant and a congestion constant are given respectively based on fuzzy rules, the strength of each rule is calculated for each set of input data (speed and saturability), three output maximum rule strengths are sequentially recorded as w1, w2 and w3, and then an output value Z is worked out through the weighted average method; fifthly, it is judged that traffic is smooth if the final value Z is the most approximate to the smoothness constant, it is judged that cars need to run slowly if the final value Z is the most approximate to the slow moving constant, and it is judged that the traffic is congested if the final value Z is the most approximate to the congestion constant. A model of the method has good continuity, high adaptive ability and good practicability.

The invention discloses a cloud-computing oriented network intrusion detect method. The method comprises the steps of generating a pseudo-gradient promotion decision-making tree set for judging network intrusion behaviors; determining weight information of pseudo-gradient promotion decision-making trees according to classification features of non-leaf nodes corresponding to the pseudo-gradient promotion decision-making trees; carrying out parallel judgment on received network behavior records through adoption of the pseudo-gradient promotion decision-making trees, thereby obtaining independent judgment results; and multiplying the independent judgment results by the corresponding weight information, thereby generating final result information for judging whether the network behavior records are network attacks or not. The method and the system are characterized in simplicity, legibility, high distinguishing precision and high smart comprehensive processing capability. The pseudo-gradient promotion decision-making tree can be generated in a distributed parallel mode under a cloud computing environment. The dynamic update operation speed of the decision-making threes is improved, and the timeliness and accuracy of detecting new-type intrusion events by an IDS are improved.

The invention relates to the field of power system transmission line relaying protection, in particular to a permanent fault identifying method for the transmission line with collateral reactor. The purpose of this invention is to overcome defects of not effectively using the transmission line with collateral reactor in the present technology. In order to overcome problems in the present technology, the embodiment of this invention is that: a permanent fault identifying method for the transmission line with collateral reactor orderly comprises the following steps: after a single-phase fault is found, the protective device phase-selecting unit confirms the fault phase; configuring the criterion and calculating the action value by using current quantity in the collateral reactor and neutral point small reactor; when the calculated action value is less than the setting value, the fault is identified as instant fault and a superposition valve processes superposition action; when the calculated action value is more than the setting value, the fault is identified as permanent fault and the superposition valve does not process superposition action. This invention has advantages of high identifying precision and reliability.

The invention provides a new karst stage judgment method. The karst reservoir is a new field of recent oil and gas exploration and development, and the judgment of karst development stage is the difficulty in a karst reservoir modeling technology at home and abroad. The existing karst stage judgment method mainly takes an exposure marker as basis of judgment, but various karst process forms exist, so that the karst processes caused by factors such as hydrothermal solution are difficult to judge due to the lack of an exposure marker. The invention provides a new method for objectively judging the karst stage by combining a method for judging the filling stage by carbon and oxygen isotopes of karst fillings and a method for judging the karst fluid active stage by inclusion homogenization temperature aiming at the problem of karst stage judgment. A carbon and oxygen isotope discrimination chart and an inclusion homogenization temperature comparison diagram are established by the new karst stage judgment method, so that the karst development stage can be accurately and rapidly judged, the limitation that the traditional karst stage judgment takes the exposure marker as basis is broken through, and the problem that karst stage judgment is difficult under the condition that the underground deep layer is short of exposure marker can be solved.

The invention relates to a computer identification technology, and specifically relates to a traffic mode discrimination method of semi-supervised SVM based on mobile phone signaling data. The methodcomprises: (1) preparing and preprocessing data; (2) designing label types; (3) extracting travel characteristics; (4) establishing an improved traffic mode manual identification process; (5) trainingan initial classifier; (6) judging the traffic mode of an unmarked sample; (7) judging whether the classifier meets a termination condition or not; (8) updating a high confidence coefficient sample data set; (9) optimizing a semi-supervised SVM classifier based on a Tri-training; (10) judging the traffic mode of an unmarked sample; (11) judging whether the classifier meets a termination condition; (12) updating a data set of a low-confidence sample; and (13) optimizing the shell vector-based semi-supervised SVM classifier. The information acquisition cost is reduced, the data utilization rateis improved, judgment is flexible and comprehensive, the precision is high, and the application scene is wider.

The invention discloses a safety factor calculation method for seabed slope stability. The safety factor calculation method comprises the steps of acquiring and calculating wave characteristic parameters and seabed soil mass strength parameters of a sea area; establishing a numerical model based on the wave power characteristics and the soil body strength parameters; analyzing the wave field dynamic model and the seabed slope numerical model based on numerical wave generation and a finite element strength reduction method to obtain a slope stability safety coefficient; and acquiring stabilitychange analysis and estimation under a certain slope rate based on different relative positions of waves and slopes. The stability coefficient of the underwater side slope under the wave load can be estimated, and convenience is provided for selection of the construction slope rate of the underwater temporary side slope. The optimized design slope rate is obtained on the basis of different designparameters and calculation working conditions by utilizing the convenience of numerical simulation calculation. When certain safety requirements are met, the earth excavation and backfill amount of the whole foundation trench is reduced, and damage to the seabed natural ecological environment is reduced to a certain extent.

The invention provides a sleeping posture detection method; when human body has different sleeping postures, a neck back and a shoulder can generate different pressure or pressure intensity on a human body support device; the method can realize detection distinguishing; a specific representation: when people lay on his / her back, the human body neck back, the human body support device and a head support device can form a natural cavity, and the human body support device in the cavity can receive very small pressure or pressure intensity from the neck back; when people lay on his / her side, the shoulder falls into the cavity, so the human body support device in the cavity can receive obviously bigger pressure or pressure intensity from the shoulder. The sleeping posture detection method is high in detection precision, is not limited by body weight, height, and sexes, and has excellent suitability; the invention also provides a pillow with an adjustable height, and the pillow is designed according to the said sleeping posture detection method.

The invention discloses an audio ground electromagnetic signal de-noising method based on signal-noise classification. The method comprises the following steps: extracting audio ground electromagneticsignal samples; respectively computing unevenness and irregularity of multifractal spectrum of each audio ground electromagnetic signal sample; training a preset classification model by utilizing theunevenness and irregularity of multifractal spectrum of each audio ground electromagnetic signal sample and a classification value of each audio ground electromagnetic signal sample; performing classification on a to-be-prepared measured audio ground electromagnetic signal according to a signal-noise classification mathematical model to obtain a signal segment not in strong interference and the signal segment in strong interference; performing matching tracing de-noising processing on the electromagnetic signal segment in strong interference; and combining the signal segment after de-noisingprocessing and the signal segment not in strong interference to obtain a reconstructed audio ground electromagnetic useful signal. Through the above method disclosed by the invention, the audio groundelectromagnetic signal with higher quality can be acquired, the condition that the signal not in the strong interference is filtered can be effectively avoided, and the de-noising precision is improved.

The invention discloses a wrong-pedaling-prevention protection method for an accelerator pedal of a novel battery electric vehicle. The method is characterized by comprising the steps that a sensor obtains the driving state and before-vehicle obstacle information of the vehicle, and transmits the state and information to a signal analysis and processing system through a CAN, the processing systemmakes a judgment and finally transmits the judgment result to a finished vehicle controller through the CAN, and the finished vehicle controller sends out a final instruction; a radar sensor is adopted for collecting information of the distance between an obstacle in front of the vehicle and the vehicle; meanwhile, a vehicle speed sensor collects vehicle speed information, the finished vehicle controller records the opening degree and time of the accelerator pedal in real time, and through derivation, the change rate of the accelerator pedal is obtained; moreover, a wrong-pedaling-prevention protection system for the accelerator pedal of the novel battery electric vehicle can preset threshold values of three parameters of the vehicle distance, the vehicle speed and the change rate of the accelerator pedal to serve as safe driving critical values. The method has the excellent advantages that the distinguishing accuracy is very high, the system stability is high, redundant cost is not added, and the probability of the occurrence of traffic accidents is lowered.

An image analysis method for generating data indicating a tumorigenic state of an image of a tissue or a cell. The image analysis method is an image analysis method for analyzing an image of a tissue or a cell using a deep learning algorithm of a neural network structure, analysis data are generated from the analysis target image including a tissue or cell to be analyzed, the analysis data are input to the deep learning algorithm, and data indicating the tumorigenic state of tissues or cells in the analysis target image are generated by the depth learning algorithm.

The invention discloses an aero-engine gas circuit fault diagnosis method based on deep learning and information fusion, and belongs to the field of power machinery fault diagnosis. The method comprises the following steps: preprocessing data; secondly, respectively training a convolutional neural network model and a feedforward neural network to be optimal by utilizing the processed sample data,so as to obtain judgment results of the two models on the category to which the sample data belongs; thirdly, in the construction stage of the basic probability distribution of the evidence bodies, judging results of the two models are regarded as the two evidence bodies, in the stage, scores of all categories to which the samples in the evidence bodies belong are converted into the basic probability distribution of the samples, and meanwhile the uncertainty of the samples is calculated; and finally, in the evidence body synthesis and decision-making stage, calculating probability distributionof the two evidence bodies after synthesis according to a synthesis rule of a D-S evidence theory, and giving a category to which each sample belongs according to a decision-making rule. According tothe invention, decision results given by two deep neural networks after learning sample data from different angles are fused, an aerial fault diagnosis model with high precision is obtained, and theproblems of unstable discrimination capability and low robustness of a non-deep learning model for aerial big data from different sources can be overcome.

The invention discloses an abnormal data discrimination method based on multi-criterion fusion. The method comprises the following steps: S1, preprocessing a sample data set; S2, according to four detection models in the abnormal data discrimination model, establishing sample data sets corresponding to the four detection models respectively; S3, inputting the input data set of the deep learning method into the long-short-term neural network model for training, and obtaining the trained long-short-term neural network model; S4, respectively inputting the sample data sets into the correspondingmodels, discriminating abnormal values, and obtaining the probability that each data point is discriminated as an abnormal point by the corresponding model; and S5, fusing the discrimination result probabilities of the data points discriminated by the corresponding models, and judging a fusion result according to a set judgment criterion to obtain a final abnormal data discrimination result. The data abnormal point discrimination precision is improved, the data accuracy and the available value are also improved, and accurate data guarantee is provided for operation detection business.

The invention relates to a rotary kiln firing state recognition method based on flame picture structure similarity. According to the technical scheme, the method includes the steps that a standard flame gray level image y, a normal flame gray level image storage FN and an abnormal flame gray level image storage FA are set; a to-be-monitored flame image P is obtained, and filtering and grey level transformation are carried out on the to-be-monitored flame image P to obtain a to-be-monitored flame gray level image x; average structural similarity coefficient calculation is carried out on the to-be-monitored flame gray level image x and all standard flame gray level images to obtain a+b MSSIM coefficients (x,y); the maximum value MAXssim of the MSSIM coefficients is selected, if the standard flame gray level image y corresponding to the maximum value MAXssim of the MSSIM coefficients belongs to the normal flame gray level image storage FN, the to-be-detected flame image P belongs to a normal state, and otherwise to-be-monitored flame image P belongs to an abnormal state. The method has the advantages of being high in precision, low in computation complexity, short in processing procedure and capable of achieving online real-time monitoring of flame state changes.

Provided is a geo-environmental simulation and prediction platform for coupled geo-environmental elements. The platform comprises an infrastructure layer, a data layer, a supporting layer, an application layer and a presentation layer; the application layer comprises a basic module, a data management module, a data mining module, a subject analysis module, a regional analysis module and a compressive assessment module, wherein the basic module is used for serving as basic auxiliary of the simulation and prediction of typical regions; the data management module is used for the simulation of typical regional elements; the data mining module performs ground-object recognition, change detection, trajectory analysis, hot point detection and visualized analysis; the subject analysis module conducts geopolitical pattern analysis, geopolitical conflict analysis and geopolitical relation analysis; the regional analysis module analyzes specific regions; and the comprehensive assessment module conducts terrorist attack risk evaluation, armed conflict risk evaluation and geo-environmental risk comprehensive evaluation. The geo-environmental simulation and prediction platform for the coupled geo-environmental elements provides a set of visualized and visual analysis platforms for geo-environmental researchers, decision-makers and managers, and can assist the researchers in analysis and decision-making.

The method includes steps: first, preparing histogram statistics for mismatched pixels of moving forwards compensation block in current input image; next, calculating statistic of heterotactic characters of histogram for mismatched pixels of moving forwards compensation block between adjacent frames; then, carrying out determining procedure for switching scenes based on the said statistic, and outputting the determined result. Based on difference detection technique of moving compensation, the invention overcomes disadvantages: sensitivity to noise and movement caused by the difference algorithm for pixels located at same positions in adjacent frames; and negligence of space information of video object occurred in conventional decision method of analyzing statistical histogram of pixel intensities. Advantages are: lowering probability of false alarm and probability failed to report in detection of switching video scenes as well as raising precision and correctness of determination.

In order to enable learning of identifiers for identifying the presence of a target cell in an image while a user performs cell-examination work, an image-feature-value calculation unit (18) extracts image feature values of images of cell-candidate regions. An NRBC identification unit (20) uses previously learned identifiers to identify, on the basis of the image feature values of the images of the cell-candidate regions, whether the target cell is depicted in the cell-candidate regions. It cases when it is identified that the target cell is depicted in a cell-candidate region, an identification-result display unit (26) displays the image of the cell-candidate region. Furthermore, in cases when it is identified that the target cell is depicted in the cell-candidate region, an identifier learning unit (34) performs, on the basis of a user input related to whether the target cell is depicted in the cell-candidate region, identifier learning in which the image feature value of the image of the cell-candidate region is used as a learning sample.

The invention discloses a special infrared reflection type yarn guide device for yarn breakage monitoring. The special infrared reflection type yarn guide device is characterized in that the upper lateral side, the lower lateral side and the periphery of a spinning frame yarn guide plate are provided with an infrared reflection type monitoring device respectively; the continuity of yarn production in the process of spun yarn spinning is sensed by every monitoring device, the intensity of a reflected light analog signal is converted into a digital signal, and the analysis is performed on the output digital signal so as to determine whether spun yarns of a ring spinning frame are broke yarns or not; the judgment on the yarn breakage of the spinning frame is based on whether the digital signal is produced or not within a certain period of time, the spun yarns are not broken if yes, and the spun yarns are broken otherwise. The special infrared reflection type yarn guide device for the yarn breakage monitoring is simple in structure, free of special requirement for raw materials and the yarn variety and applicable to the yarn breakage monitoring of the ring spinning of cotton spinning, wool spinning and color spinning and the like.