CN117456439A - Violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation - Google Patents
Violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation Download PDFInfo
- Publication number
- CN117456439A CN117456439A CN202311205384.6A CN202311205384A CN117456439A CN 117456439 A CN117456439 A CN 117456439A CN 202311205384 A CN202311205384 A CN 202311205384A CN 117456439 A CN117456439 A CN 117456439A
- Authority
- CN
- China
- Prior art keywords
- image
- image data
- dangerous task
- early warning
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 29
- 238000000605 extraction Methods 0.000 claims abstract description 22
- 238000007781 pre-processing Methods 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000013145 classification model Methods 0.000 claims abstract description 6
- 238000004364 calculation method Methods 0.000 claims description 18
- 108090000623 proteins and genes Proteins 0.000 claims description 15
- 230000035772 mutation Effects 0.000 claims description 13
- 230000006399 behavior Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 239000011159 matrix material Substances 0.000 claims description 9
- 230000014759 maintenance of location Effects 0.000 claims description 7
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 2
- 238000012544 monitoring process Methods 0.000 abstract description 4
- 238000004519 manufacturing process Methods 0.000 description 10
- 230000008859 change Effects 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000002068 genetic effect Effects 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 208000027418 Wounds and injury Diseases 0.000 description 3
- 230000006378 damage Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 208000014674 injury Diseases 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000003278 mimic effect Effects 0.000 description 2
- 238000013486 operation strategy Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 206010064571 Gene mutation Diseases 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/84—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using probabilistic graphical models from image or video features, e.g. Markov models or Bayesian networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for warning illegal operation and predicting dangerous task level based on man-machine collaborative operation, which comprises the following steps: acquiring image data of a worker and a machine in a cooperative operation; performing image preprocessing on the image data, and dividing an image data set; performing feature extraction on the preprocessed image data based on a W-BA algorithm; classifying the extracted features by using CART tree, constructing a rule-breaking feature classification model, and identifying rule-breaking operationBehavior and early warning are carried out; based onAnd training and calculating the posterior probability of each feature classification, selecting the category with the highest posterior probability as a dangerous task level prediction result, and carrying out early warning according to the dangerous task level. According to the method, the illegal operation behavior is identified in real-time monitoring, early warning and intervention are timely carried out, accidents caused by the illegal operation are avoided, the capability of processing complex operation scenes is achieved, and the safety of an operation environment is remarkably improved.
Description
Technical Field
The invention relates to the technical field of man-machine cooperation, in particular to a violation operation early warning and dangerous task level prediction method based on man-machine cooperation operation.
Background
In traditional man-machine collaborative operation, the machine executes operation according to a preset instruction, and a human operator is mainly responsible for supervision and adjustment, however, the mode is easily affected by human errors, negligence and illegal operation, so that accidents and problems on a production line are caused, especially in a complex production environment, the capacity of identifying illegal operation and evaluating dangerous tasks is limited, and the operator can generate illegal operation for various reasons, so that serious potential safety hazards and production loss are caused.
Some tasks are related to high-risk operation or special environments, potential threats are formed on the safety of operators, on one hand, the traditional image feature extraction method is limited by the features of manual design, key information under a complex scene is difficult to capture, so that the recognition performance is poor, on the other hand, the traditional risk assessment method has delay in terms of real-time prediction and feedback, accurate judgment and response cannot be timely made when the illegal operation occurs, so that potential safety hazards are generated in high-risk tasks, and the traditional system is too simplified for risk level assessment, lacks individuation and real-time risk level judgment, and cannot comprehensively and accurately judge the risk task level.
Therefore, a technology for identifying the illegal operation and evaluating the dangerous level of the task and actively intervening if necessary to ensure the operation safety and efficiency is urgently needed for the man-machine cooperative work.
Disclosure of Invention
In order to overcome the defects and shortcomings in the prior art, the invention provides a violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation, which improves the operation efficiency and safety by combining deep learning and machine learning technologies, can extract features from working images, and realizes accurate recognition of violation operation by optimizing a model through a W-BA algorithm; the method has the advantages that the illegal operation behaviors are identified in real-time monitoring, early warning and intervention are timely carried out, accidents caused by the illegal operation are prevented, personnel injury and production loss are reduced, the capability of processing complex operation scenes is achieved, and the safety of an operation environment is remarkably improved.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the invention provides a violation operation early warning and dangerous task level prediction method based on man-machine collaborative operation, which comprises the following steps:
acquiring image data of a worker and a machine in a cooperative operation;
performing image preprocessing on the image data, and dividing an image data set;
performing feature extraction on the preprocessed image data based on a W-BA algorithm;
classifying the extracted features by using a CART tree, constructing a rule-breaking feature classification model, identifying rule-breaking operation behaviors and carrying out early warning;
based onAnd (3) Bayes training is carried out to calculate the posterior probability of each feature classification, the category with the highest posterior probability is selected as a dangerous task level prediction result, and early warning is carried out according to the dangerous task level.
As a preferable technical scheme, the image preprocessing for the image data specifically includes:
and denoising the image, converting the color image into a gray image, enhancing the contrast of the image, and sharpening the image.
As a preferable technical scheme, the feature extraction is performed on the preprocessed image data based on a W-BA algorithm, and specifically includes:
carrying out fitness calculation on the preprocessed image data;
calculating the fitness value of each image characteristic, selecting the image characteristic with the largest retention ratio according to roulette, and selecting the image characteristic as a parent individual to perform intersection and mutation operation;
selecting genes from the parents based on random crossover for combining to produce offspring individuals;
introducing random variation of genes in generating new individuals according to Weight variation;
updating the population, generating a new generation of individual set, and extracting to obtain image features.
As a preferable technical scheme, the fitness value calculation formula is expressed as:
yα 1 x1+α 2 x2α 3 x3…α n xn
wherein x1, x2, x3,..xn represents the image characteristic data after preprocessing, α 1 、α 2 、α 3 、α n All are expressed as preset super parameters for coordinating the image feature sizes, and y represents the fitness value.
As a preferable technical scheme, the image characteristics with the largest retention ratio are selected according to the roulette selection, and the roulette selection calculation formula is expressed as follows:
where y represents the fitness value of the current feature, Σy represents the fitness values of all features, prob represents the predicted probability value.
As a preferred technical solution, the combination is performed from the parent selection genes based on random crossover, and the random crossover calculation formula is expressed as:
wherein,representing randomly selecting m results from all the n roulette selection results as output, de-weight representing performing de-duplication operation on the selected m results, and forming new data by reserved data permutation and combination, and extracting to obtain features.
As a preferred technical scheme, according to the random variation of the genes introduced by Weight variation when generating new individuals, the Weight variation calculation formula is expressed as:
Var＝W*x B
wherein W and B represent matrices, the matrix W is used for assigning weights to the features x, the matrix B is used for adjusting the size of the matrix, x represents the features left after selection and crossing, and Var represents the final variation value.
As a preferred technical solution, the extracted features are classified by using CART tree, and the specific steps include:
labeling the image data set to obtain an image data set containing characteristics and labels, wherein the characteristics are used for representing the attribute for classification, and the labels are used for representing the predicted category;
preprocessing an image dataset comprising features and labels, including processing missing values, normalizing or normalizing the features, and encoding the labels into digital values;
training by using a CART tree algorithm, selecting the optimal characteristics of an unrepeace criterion and a threshold value, recursively dividing a data set, and constructing a tree structure until a stopping condition is reached;
after training, classifying the new image data set by using the CART tree, starting from the root node, dividing along branches of the tree according to the characteristics and the threshold value, and finally reaching the leaf node, wherein the class or class probability of the leaf node is used as a prediction result of the image data set.
As a preferred technical proposal, based onThe Bayes training calculates the posterior probability of each feature class, and the specific steps include:
at the position ofIn the training process of Bayes, the prior probability of each category is calculated from the image data training set, the classification prediction probability is carried out, and the conditional probability under each category is calculated for each category characteristicClassifying the new image dataset based on +.>Bayes calculates the posterior probability of the image dataset under each category.
The invention also provides a system for warning illegal operation and predicting dangerous task level based on man-machine collaborative operation, which comprises the following steps: the system comprises an image data acquisition module, an image preprocessing module, a feature extraction module, a feature classification module, an early warning module and a dangerous task grade prediction module;
the image data acquisition module is used for acquiring image data of the cooperative operation of workers and machines;
the image preprocessing module is used for carrying out image preprocessing on the image data and dividing an image data set;
the feature extraction module is used for carrying out feature extraction on the preprocessed image data based on a W-BA algorithm;
the feature classification module is used for classifying the extracted features by using the CART tree, constructing an offence feature classification model and identifying offence operation behaviors;
the dangerous task grade prediction module is used for being based onThe Bayes training calculates the posterior probability of each feature classification, and the class with the highest posterior probability is selected as a dangerous task level prediction result;
the early warning module is used for early warning the illegal operation behaviors and early warning according to the dangerous task grade.
Compared with the prior art, the invention has the following advantages and beneficial effects:
(1) The invention combines deep learning and machine learning technologies to be applied to a man-machine collaborative operation system, recognizes the illegal operation behavior in real-time monitoring, and timely performs early warning and intervention, thereby being beneficial to preventing accidents caused by the illegal operation, reducing personnel injury and production loss, having the capability of processing complex operation scenes and remarkably improving the safety of the operation environment.
(2) According to the method, the illegal operation and the high-risk task are automatically identified, production line stagnation and waste caused by misoperation or improper operation can be avoided, the operation strategy can be automatically adjusted according to the characteristics and the risk level of the task, the operation efficiency is improved to the greatest extent, and the production efficiency and the product quality are optimized.
(3) The method and the device learn and optimize from a large amount of data, thereby better adapting to different production scenes and changes, extracting the characteristics based on the W-BA algorithm, optimizing the characteristic set, selecting the most representative characteristics, and improving the recognition accuracy and generalization capability of the W-BA model.
(4) The intelligent intervention mechanism not only focuses on the identification of illegal operations, but also focuses on the prediction and evaluation of dangerous tasks, can automatically analyze the characteristics of the tasks, evaluate the dangerous grades of the tasks, take corresponding measures to ensure the safety of operators, can automatically trigger alarms according to the illegal operations or the high-risk tasks, and perform corresponding interventions, such as stopping the operations or replacing the operators to execute the tasks, and can prevent misoperation and potential accidents.
Drawings
FIG. 1 is a flow chart of a method for warning illegal operations and predicting dangerous task levels based on man-machine collaborative operation;
FIG. 2 is a schematic flow chart of feature extraction based on a W-BA algorithm;
fig. 3 is a schematic application flow chart of the method for warning illegal operation and predicting dangerous task level based on man-machine collaborative operation.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Example 1
As shown in fig. 1, the embodiment provides a method for warning illegal operations and predicting dangerous task levels based on man-machine collaborative operation, which comprises the following steps:
s1: during the data acquisition phase of the system, the key task is to acquire image and video data from the job site, which will be used to train models, analyze operations, and improve system performance. The data acquisition is the basis of the whole system and has important influence on the accuracy and the reliability of the system.
In the data acquisition stage, the primary task is to set a data acquisition point, for example, a camera is set on a robot arm to take data without affecting the work, which involves arranging the camera or other sensors on a work site to capture real-time situations of the cooperative operation of workers and machines, and the devices need to be capable of capturing various operation scenes, angles and actions so as to fully reflect the real working environment, and once the data acquisition device is set, the real-time recording and storage of the data is performed, and the data includes various information such as positions, actions, postures and the like of workers and machines; in addition, environmental information of the working area, such as illumination, temperature, humidity, etc., which also affect the safety and accuracy of the operation, an effective data storage strategy needs to be implemented in order to ensure the usability and consistency of the data, which involves using a database or cloud storage solution to manage a large amount of image and video data, and the data should be classified and stored according to standards of time, operation type, worker identity, etc., and stored in the form of a table or database for later analysis and backtracking.
S2: preprocessing the image obtained by the steps, including: first, denoising, which is to perform denoising processing on an image, so as to reduce noise interference in the image. The denoising method comprises median filtering, gaussian filtering, mean filtering and the like; secondly, graying, namely converting the color image into a gray image, removing color information and only retaining brightness information; thirdly, the contrast is enhanced, and the contrast of the image is adjusted, so that details in the image are more obvious; fourthly, sharpening the image, and enhancing the edge and detail of the image to make the image clearer; fifth, the size of the image is adjusted as needed, and the image can be scaled, cut, etc.
Finally, the data set is divided into a training set, a verification set and a test set, wherein the training set is used for training and parameter optimization of the model, the verification set is used for selecting and debugging the model, and the test set is used for evaluating the performance and generalization capability of the model.
S3: extracting features based on a W-BA algorithm;
the image feature extraction stage has a key effect in a system based on man-machine collaborative operation, extracts features from a working image through a W-BA algorithm, and provides a basis for subsequent model training and operation prediction.
As shown in fig. 2, feature extraction based on the W-BA algorithm includes the following sub-steps:
(1) Acquiring image data;
the preprocessed image obtained after step S2 is input into the W-BA algorithm for subsequent operations.
(2) Calculating the fitness;
assessing fitness is a crucial step in the W-BA algorithm. It quantifies its performance in the problem space by applying an objective function of the problem to each individual solution. The fitness function not only reflects the quality of the individual solutions, but also provides basis for the selection process, so that the excellent solutions are more likely to be selected. In each generation of evolution, the fitness function helps the algorithm to distinguish the merits of the solutions, so as to guide the crossover and mutation operations and gradually drive the population to develop towards the better solutions. Through fitness evaluation, the W-BA algorithm can simulate natural selection, focus on a promising area in the potential solution space, and therefore the optimal solution of the problem is effectively searched.
The fitness function calculation formula in the feature extraction of the W-BA algorithm is specifically expressed as follows:
yα 1 x1+α 2 x2α 3 x3…α n xn⑴
wherein alpha is 1 、α 2 、α 3 、α n All are shown as preset super parameters for coordinating the feature sizes, x1, x2, x3 and xn all represent the image feature data after preprocessing, y represents the fitness function value,to characterize the importance of the current feature.
In the embodiment, fitness calculation is performed according to given super parameters, the difference between the predicted result and the real result is calculated, the size of each parameter is updated according to difference feedback, and repeated repeatedly is performed continuously, so that parameter adjustment is realized;
the fitness function selected in the embodiment increases the super parameter α, that is, the feature at each moment is not equal weight, but the value of the feature with high importance is larger, and the value of the feature with low importance is smaller, so that the final result is only affected by important information, and the accuracy is improved.
(3) Roulette selection;
the selection operation is a core step in the W-BA genetic algorithm and is used for selecting parent individuals from the current population for subsequent crossing and mutation operations according to the fitness value of the individuals. The goal of the selection is to make it more likely that the more adaptable individual will be selected to mimic the principle of survival of the fittest in natural evolution. The selection method of the embodiment is roulette selection, which regards fitness values as scales on the roulette, and the probability that an individual is selected is proportional to the fitness thereof.
The roulette selection calculation formula in the W-BA algorithm feature extraction is specifically expressed as:
where y represents the fitness value of the current feature, Σy represents the fitness values of all features, prob represents the predicted probability value.
The method adopted by the embodiment squares the ratio, and has the advantages that on one hand, negative number generation is avoided, on the other hand, the squared data with smaller values are smaller, the occupied probability is smaller, and the selection operation is facilitated; the selection operation not only guides better individuals to enter the next generation, but also ensures the diversity of the population and avoids the premature convergence to the local optimal solution.
In this embodiment, the following operations are performed by calculating the fitness value of each feature, and selecting the data with the largest retention ratio according to the roulette selection calculation formula.
(4) Uniformly crossing;
crossover is an important operation in the W-BA algorithm for generating new individual solutions by combining the genetic information of parent individuals to generate offspring. The goal of crossover is to mimic the genetic recombination process in biological genetics, introduce diversity and promote the evolution of populations. During crossover, certain gene segments in the selected pair or pairs of parent individuals will be interchanged, thereby forming offspring individuals.
The crossover method used in this example is random crossover, which randomly combines the selection genes from the parent to produce offspring individuals.
The random cross calculation formula in the feature extraction of the W-BA algorithm is specifically expressed as:
wherein,and randomly selecting m results from all the n results as output, wherein n represents the data reserved after the previous selection, de-weight represents the de-duplication operation on the m selected results, and the reserved data are arranged and combined to form new data to finish feature extraction.
The cross algorithm of the embodiment adopts a random selection mode, so that the diversity of samples is ensured, and the result is more reliable.
The crossover operation is not only helpful to generate new solution space, but also can avoid population trapping in the local optimal solution. However, the manner of interleaving and probability selection have a significant impact on the performance of the algorithm. The crossover probability can be adjusted to maintain diversity, and too high crossover probability can cause too slow convergence speed, so that the crossover, selection, mutation and other operations are comprehensively used, and the genetic algorithm can gradually search for excellent solutions through multi-generation evolution and is suitable for various optimization problems.
(5) Weight variation;
variation is a critical operation in the W-BA algorithm for introducing random changes in genes when generating new individuals. The method is a process of simulating gene mutation in natural evolution, is beneficial to maintaining population diversity and avoids the algorithm from falling into a local optimal solution.
Mutation is achieved by randomly modifying the individual gene values, usually involving separate gene loci, with the rate of mutation being a parameter that controls the probability of mutation, usually small, to ensure that most genes remain unchanged, the purpose of mutation being to introduce enough perturbation, but not to completely upset the individual's characteristics.
The Weight variation calculation formula in the feature extraction of the W-BA algorithm is specifically expressed as follows:
Var＝W*x B⑷
wherein W and B represent matrices, where W is used to assign weights to x, B is used to adjust the size of the matrix, x represents the features left after selection and crossing, and Var represents the final variance value.
The variation of this embodiment adopts Weight variation, that is, features are randomly changed through a random matrix, and the characteristics after the change are adjusted through a B matrix, and whether the amount of change meets the set requirement is judged based on loss calculation, if the change is too large, the amount of change needs to be reduced, and if the change is too small, the change value needs to be increased through B; the mutation operation can help the algorithm jump out of the local optimal solution and explore a wider solution space.
(6) A population of offspring;
updating populations is an important step in genetic algorithms that occur after selection, crossover, mutation, etc., to generate a new generation set of individuals. In the updating population stage, the offspring individuals generated through the operations of selection, crossover, mutation and the like are replaced with some individuals in the original population. In general, the replacement strategy may be elite retention (Elitism) or a complete replacement method. The selection of the update strategy depends on the nature of the problem and the goal of the algorithm. Elite retention can maintain excellent solutions but reduce diversity of populations. Complete substitution then helps introduce new gene combinations but destroys the excellent solutions that have been obtained. Combining the operations of selection, crossing, mutation, updating and the like, the genetic algorithm gradually improves the population through multi-generation evolution to find a better solution of the problem.
In this embodiment, the more accurate and better the extracted data requires prediction, the W-BA algorithm works.
S4: training a CART tree;
in training the CART tree, it builds a tree structure by recursively partitioning the dataset into less impure subsets. Each internal node of the tree represents a feature and its threshold for partitioning data into different branches. The leaf nodes represent the final prediction result and may be classification labels or regression values. The CART tree training process starts with the root node, selects the best features and thresholds for segmentation, using criteria based on the non-purity, such as the genie non-purity (for classification) or the square error (for regression). The segmented subsets continue to be recursively segmented until a predefined stopping condition is reached, such as the maximum depth of the tree or the number of samples in the node reaching a threshold. In the embodiment, the CART tree is adopted to classify the features after the step S3, and the CART tree is trained to accurately distinguish the illegal features.
The specific implementation process is as follows: in classifying the features after step S3 using the CART tree, a dataset is first prepared comprising features representing the attributes used for classification and labels representing the categories to be predicted, typically the categories, such as violations and non-violations. Next, the data is pre-processed, including processing missing values, normalizing or normalizing features, and encoding the tags into digital values. The CART tree algorithm is then used to train, select the best features and thresholds based on the criteria of non-purity such as genie non-purity to recursively divide the data set, building a tree structure until a stopping condition is reached, such as maximum depth of the tree or minimum number of samples in the nodes. After training is completed, the CART tree may be used to classify new data samples, starting from the root node, segmenting along branches of the tree according to features and thresholds, and finally reaching leaf nodes whose class or class probability will be the prediction result of the samples. Finally, an independent test data set can be used for evaluating the classification performance of the CART tree, such as indexes of accuracy, recall, F1 score and the like, so as to ensure the effectiveness and accuracy of the model. By this process, a model can be created that accurately classifies the offending features.
S5：Bayes training
Nave Bayes is a classification algorithm based on probability statistics and is commonly used for tasks such as text classification, spam filtering, emotion analysis and the like. It is based on bayesian theorem and "naive" assumptions, i.e. assuming that the features are independent of each other, simplifying the calculation process. In the naive bayes training process, first, the prior probabilities of the respective categories and the conditional probabilities of each feature under the respective categories are calculated from the training data. This probability information is used to predict the classification of the new sample. When a new sample is given, the algorithm calculates the posterior probability of that sample under each category and selects the category with the highest posterior probability as the prediction result. When the operation is performed in an unknown place, the risk level assessment operation can be performed according to the real-time flow image, so that the casualties are prevented, and the working safety degree is greatly improved.
The specific implementation is as follows: at the position ofIn the training process of Bayes, the prior probability of each category is calculated from training data, namely, the classification prediction probability is carried out on the images captured through high-speed shooting. Then, for each class feature, a conditional probability under the respective class, i.e. the probability of the feature occurring given a certain class, is calculated. These probability information constituteParameters of the Bayes model. When (when)When a new image sample needs to be classified, < +.>The Bayes algorithm calculates the posterior probability of the sample under each category. The posterior probability reflects the probability that a given sample belongs to each class. Then (I)>The Bayes algorithm will select the class with the highest posterior probability as the predicted outcome, i.e. the class with the highest probability as the final outcome, i.e. the risk level mentioned herein.
S6: judging the violation operation by the CART tree;
as shown in fig. 3, when a worker performs work, the illegal operation behavior is identified in real-time monitoring, and early warning and intervention are performed in time. This helps to prevent accidents caused by illegal operations, reduces personnel injuries and production losses, and creates a safer work environment by allowing workers to find problems with greater accuracy and sensitivity.
S7：Bayes predicts the risk level;
as shown in fig. 3, the trained model is input into the picture during actual construction to predict probability, predict danger level, alarm if danger exists, automatically learn the type of dangerous operation, automatically adjust operation strategy according to the task characteristics and danger level, and maximally improve operation efficiency, thereby not only reducing production cost, but also improving consistency and quality of products, and being beneficial to improving market competitiveness of enterprises.
Example 2
The technical contents are the same as in example 1 except for the following technical contents;
the embodiment provides a violation operation early warning and dangerous task level prediction system based on man-machine collaborative operation, which comprises the following steps: the system comprises an image data acquisition module, an image preprocessing module, a feature extraction module, a feature classification module, an early warning module and a dangerous task grade prediction module;
in this embodiment, the image data acquisition module is configured to acquire image data of a worker operating in cooperation with a machine;
in this embodiment, the image preprocessing module is configured to perform image preprocessing on image data and divide an image dataset;
in this embodiment, the feature extraction module is configured to perform feature extraction on the preprocessed image data based on a W-BA algorithm;
in this embodiment, the feature classification module is configured to classify the extracted features by using a CART tree, construct a rule-breaking feature classification model, and identify rule-breaking operation behaviors;
in this embodiment, the dangerous task level prediction module is configured to be based onThe Bayes training calculates the posterior probability of each feature classification, and the class with the highest posterior probability is selected as a dangerous task level prediction result;
in this embodiment, the early warning module is used for early warning the illegal operation behavior and early warning according to the dangerous task level.
The above examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above examples, and any other changes, modifications, substitutions, combinations, and simplifications that do not depart from the spirit and principle of the present invention should be made in the equivalent manner, and the embodiments are included in the protection scope of the present invention.
Claims (10)
1. A violation operation early warning and dangerous task grade prediction method based on man-machine cooperative operation is characterized by comprising the following steps:
acquiring image data of a worker and a machine in a cooperative operation;
performing image preprocessing on the image data, and dividing an image data set;
performing feature extraction on the preprocessed image data based on a W-BA algorithm;
classifying the extracted features by using a CART tree, constructing a rule-breaking feature classification model, identifying rule-breaking operation behaviors and carrying out early warning;
based onAnd (3) Bayes training is carried out to calculate the posterior probability of each feature classification, the category with the highest posterior probability is selected as a dangerous task level prediction result, and early warning is carried out according to the dangerous task level.
2. The method for warning illegal operation and predicting dangerous task level based on man-machine cooperative operation according to claim 1, wherein the image preprocessing is performed on the image data, and specifically comprises the following steps:
and denoising the image, converting the color image into a gray image, enhancing the contrast of the image, and sharpening the image.
3. The method for warning and predicting the level of a dangerous task according to claim 1, wherein the method for predicting the level of the illegal operation based on the man-machine collaborative operation is characterized by extracting features of the preprocessed image data based on a W-BA algorithm, and specifically comprises the following steps:
carrying out fitness calculation on the preprocessed image data;
calculating the fitness value of each image characteristic, selecting the image characteristic with the largest retention ratio according to roulette, and selecting the image characteristic as a parent individual to perform intersection and mutation operation;
selecting genes from the parents based on random crossover for combining to produce offspring individuals;
introducing random variation of genes in generating new individuals according to Weight variation;
updating the population, generating a new generation of individual set, and extracting to obtain image features.
4. The method for warning and predicting the level of a dangerous task according to claim 3, wherein the fitness value calculation formula is expressed as:
y＝α 1 x1+α 2 x2+α 3 x3+…α n xn
wherein x1, x2, x3,..xn represents the image characteristic data after preprocessing, α 1 、α 2 、α 3 、α n All are expressed as preset super parameters for coordinating the image feature sizes, and y represents the fitness value.
5. The method for warning and predicting the level of a dangerous task according to claim 3, wherein the image features with the largest retention ratio are selected according to the roulette selection, and the roulette selection calculation formula is expressed as follows:
where y represents the fitness value of the current feature, Σy represents the fitness values of all features, prob represents the predicted probability value.
6. The method for warning and predicting the level of a dangerous task according to claim 3, wherein the genes are selected from the parents based on random crossover, and the random crossover calculation formula is expressed as:
wherein,representing randomly selecting m results from all the n roulette selection results as output, de-weight representing de-duplication operation of the selected m results, forming new data by the reserved data permutation and combination, extracting to obtain features。
7. The method for warning and predicting the level of a dangerous task according to claim 3, wherein the random variation of genes is introduced when a new individual is generated according to Weight variation, and the Weight variation calculation formula is expressed as:
Var＝W*x±B
wherein W and B represent matrices, the matrix W is used for assigning weights to the features x, the matrix B is used for adjusting the size of the matrix, x represents the features left after selection and crossing, and Var represents the final variation value.
8. The method for warning and predicting the level of a dangerous task according to claim 1, wherein the extracted features are classified by using CART trees, and the specific steps include:
labeling the image data set to obtain an image data set containing characteristics and labels, wherein the characteristics are used for representing the attribute for classification, and the labels are used for representing the predicted category;
preprocessing an image dataset comprising features and labels, including processing missing values, normalizing or normalizing the features, and encoding the labels into digital values;
training by using a CART tree algorithm, selecting the optimal characteristics of an unrepeace criterion and a threshold value, recursively dividing a data set, and constructing a tree structure until a stopping condition is reached;
after training, classifying the new image data set by using the CART tree, starting from the root node, dividing along branches of the tree according to the characteristics and the threshold value, and finally reaching the leaf node, wherein the class or class probability of the leaf node is used as a prediction result of the image data set.
9. The method for warning illegal operation and predicting dangerous task level based on man-machine cooperative operation according to claim 1, wherein the method is based onThe Bayes training calculates the posterior probability of each feature class, and the specific steps include:
at the position ofIn the training process of Bayes, the prior probability of each category is calculated from the image data training set, the classification prediction probability is carried out, the conditional probability under each category is calculated for each category characteristic, and when a new image data set is classified, the classification is based on +.>Bayes calculates the posterior probability of the image dataset under each category.
10. A violation operation early warning and dangerous task level prediction system based on man-machine collaborative operation is characterized by comprising: the system comprises an image data acquisition module, an image preprocessing module, a feature extraction module, a feature classification module, an early warning module and a dangerous task grade prediction module;
the image data acquisition module is used for acquiring image data of the cooperative operation of workers and machines;
the image preprocessing module is used for carrying out image preprocessing on the image data and dividing an image data set;
the feature extraction module is used for carrying out feature extraction on the preprocessed image data based on a W-BA algorithm;
the feature classification module is used for classifying the extracted features by using the CART tree, constructing an offence feature classification model and identifying offence operation behaviors;
the dangerous task grade prediction module is used for being based onThe Bayes training calculates the posterior probability of each feature classification, and the class with the highest posterior probability is selected as a dangerous task level prediction result;
the early warning module is used for early warning the illegal operation behaviors and early warning according to the dangerous task grade.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311205384.6A CN117456439A (en) | 2023-09-18 | 2023-09-18 | Violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311205384.6A CN117456439A (en) | 2023-09-18 | 2023-09-18 | Violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117456439A true CN117456439A (en) | 2024-01-26 |
Family
ID=89589959
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311205384.6A Pending CN117456439A (en) | 2023-09-18 | 2023-09-18 | Violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117456439A (en) |
-
2023
- 2023-09-18 CN CN202311205384.6A patent/CN117456439A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111832608B (en) | Iron spectrum image multi-abrasive particle identification method based on single-stage detection model yolov3 | |
| CN112383052B (en) | Power grid fault repairing method and device based on power internet of things | |
| CN117421684B (en) | Abnormal data monitoring and analyzing method based on data mining and neural network | |
| CN109544399B (en) | Power transmission equipment state evaluation method and device based on multi-source heterogeneous data | |
| CN111401418A (en) | Employee dressing specification detection method based on improved Faster r-cnn | |
| CN111126820A (en) | Electricity stealing prevention method and system | |
| CN111237988A (en) | Control method and system for subway vehicle-mounted air conditioning unit | |
| CN110636066A (en) | Network security threat situation assessment method based on unsupervised generative reasoning | |
| CN115865483A (en) | Abnormal behavior analysis method and device based on machine learning | |
| CN116052082A (en) | Power distribution station room anomaly detection method and device based on deep learning algorithm | |
| CN116884192A (en) | Power production operation risk early warning method, system and equipment | |
| CN117032165A (en) | Industrial equipment fault diagnosis method | |
| CN117315380A (en) | Deep learning-based pneumonia CT image classification method and system | |
| CN116756688A (en) | Public opinion risk discovery method based on multi-mode fusion algorithm | |
| CN113743461B (en) | Unmanned aerial vehicle cluster health degree assessment method and device | |
| CN118446652A (en) | LED track lamp production control method and system based on Internet of things | |
| CN118038021A (en) | Transformer substation operation site foreign matter intrusion detection method based on improvement yolov4 | |
| CN117768618A (en) | Method for analyzing personnel violation based on video image | |
| CN117372723A (en) | Intelligent substation violation operation early warning system | |
| CN117523499A (en) | Forest fire prevention monitoring method and system based on Beidou positioning and sensing | |
| CN117372144A (en) | Wind control strategy intelligent method and system applied to small sample scene | |
| CN109636194B (en) | Multi-source cooperative detection method and system for major change of power transmission and transformation project | |
| CN117456439A (en) | Violation operation early warning and dangerous task grade prediction method based on man-machine collaborative operation | |
| CN115438190A (en) | Power distribution network fault decision-making assisting knowledge extraction method and system | |
| CN118096131B (en) | Operation and maintenance inspection method based on electric power scene model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |