KR102494578B1 - Method and System for Artificial Intelligence based Quality Inspection in Manufacturing Process using Machine Vision Deep Learning - Google Patents

Abstract

The present invention relates to an AI-based detachable safety detection method and safety device for agricultural machinery using deep learning. It is about an AI-based detachable safety detection method and safety device for agricultural machinery using deep learning, which can work more safely by notifying with an alarm and protect agricultural machinery operators and people around them.

Description

AI-based detachable safety detection method and safety device for agricultural machinery using deep learning {Method and System for Artificial Intelligence based Quality Inspection in Manufacturing Process using Machine Vision Deep Learning}

The present invention relates to an AI-based detachable safety detection method and safety device for agricultural machinery using deep learning. It is about an AI-based detachable safety detection method and safety device for agricultural machinery using deep learning, which can work more safely by notifying with an alarm and protect agricultural machinery operators and people around them.

Every year, the population returning to farming is increasing, interest in agriculture is growing, and the age group engaged in agriculture is also getting younger. Currently, the mechanization rate of rice paddy agriculture is 97%, and the mechanization rate of field agriculture is 58%. Safety measures for the use of agricultural machinery are required according to the mechanization of agriculture.

In general, automobiles are well equipped with safety devices such as front/rear assist sensors and inter-vehicle distance control, but safety measures are not provided for agricultural machinery. Since the agricultural environment has many obstacles invisible to the naked eye during work, it is very different from the road environment, so a sensor suitable for the agricultural environment is required. Although various systems such as ADAS and autonomous driving are being developed in autonomous driving technology, there is no system for work vehicles that are only installed in expensive vehicles and require more detailed driving.

Between 2014 and 2018, a total of 6,981 accidents occurred with agricultural machinery, and 492 deaths occurred, with about 4 safety accidents occurring per day. The need for detection sensors is highlighted in that 74% of the causes of cultivator accidents are overturning and falling.

In addition, it is necessary to create a low-cost detachable safety device to bring safety and income to farmers at the same time.

In addition, if a strong ND magnet and a case suitable for agricultural machinery are produced with a 3D printer, it can be attached to and detached from all agricultural machines with one low-cost safety device. It is expected that the sensor can be fabricated.

Also, in the agricultural environment, there are environments made of grain, grass, mud, etc. (safe objects) that can be stepped on or bumped into during work, but in environments such as stones and trees that are difficult to visually identify during work, even if simply bumped into, it will lead to a fall/fall. There are potential risk factors.

Meanwhile, the early Raspberry Pi was manufactured through a licensed manufacturing agreement with Element 14/Premier Panel and RS Components.

Raspberry Pis all have the same Videocore IV GPU, single-core ARMv6 compatible CPU or the newer ARMv7 compatible quad-core (Raspberry Pi 2), 1 GB of RAM (Raspberry Pi 2), 512 MB (Raspberry Pi 1 B and B+), or 256 MB (models A and A+, older model B) of memory. They have an SD card slot (models A and B) or microSDHC for bootable media and persistent information storage.

In 2014, the Raspberry Pi Foundation released a 'compute module' for use as part of an embedded system with the same computational power as the original Raspberry Pi. In early 2015, the next-generation Raspberry Pi, the Raspberry Pi 2, was released. This new computer board initially came in one format (the Model B), with a quad-core ARM Cortex-A7 CPU and 1GB of RAM, with the rest of the specs similar to the Model B+.

The Raspberry Pi Foundation provides downloads for the Debian and Arch Linux ARM distributions, promotes the use of Python as a major programming language, and supports BBC Basic. Languages such as C, C++, Java, Perl, Ruby, and Squeak Smalltalk are available.

However, a device for agricultural machinery using a raspberry pi having such a variety of functions has not yet been developed in various ways.

On the other hand, image processing techniques used in machine vision are capable of high accuracy and defect rate detection under conditions where there is no change in the target item and the external factors are exactly the same and controlled. It has limitations in that it requires trying to learn a new model or redesigning the discrimination algorithm for learning.

In addition, existing agricultural machines are made in such a way that the operator visually checks the workplace with the naked eye. However, agricultural machines such as cultivators and tractors have a problem in that the ground for plowing or weeding is covered by the machine, and human error occurs due to various factors such as organization and environment in visual inspection.

JP 2014-153560 A (published on 2014.08.25.) KR 10-1711073 B1 (registration date 2017.02.22.) KR 10-2073162 B1 (registration date 2020.01.29.) JP 6653929 B1 (registration date 2020.01.31.)

The present invention has been made to solve the above problems, and an object of the present invention is to provide a detachable safety detection device specialized for agricultural machinery using deep learning.

In addition, the present invention is in the form of attachment and detachment that anyone can easily mount on agricultural machinery, and intuitive notifications such as sound and screen vibration are possible, so that the safety of workers in aging rural areas can be maintained.

In addition, the present invention can be used in various agricultural machines by making a case that can be attached to and detached from each agricultural machine such as a tractor and a cultivator.

In addition, the present invention distinguishes through deep learning an object that does not matter if it is stepped on or collided with, such as grain, grass, mud, etc. (safe object), and an object that is damaged when collided with, such as a stone or tree, and only dangerous objects are detected and notified by the sensor make it possible

In addition, in the present invention, grass, trees, etc. can be distinguished through deep learning learning because the shape of the object is different, thereby eliminating human error, minimizing the cost of quality control, and improving productivity.

The present invention for achieving the above object is an AI-based detachable safety detection method for agricultural machinery equipped with a deep learning function controlled by an artificial intelligence unit, through a camera in which the artificial intelligence unit processes and transmits images of the ground in real time processing and analyzing an input image frame by frame; learning a learning model using an artificial intelligence algorithm for an image component related to a dangerous situation in the image frame analyzed by the artificial intelligence unit; recognizing one or more objects from the collected images based on the learned algorithm by the artificial intelligence unit, and determining a dangerous situation based on states of the one or more objects; and generating an alarm according to the dangerous situation by the artificial intelligence unit.

The step of determining the dangerous situation by the artificial intelligence unit may include: extracting characteristics of rocks, soil lumps, trees, and crop types in order to diagnose a ground situation; converting the extracted one-dimensional image signal into a two-dimensional gray-scale image by the artificial intelligence unit, and extracting features for each type by confirming a correlation between samples; and diagnosing the state of the obstacle through the learning model that has passed the verification by the artificial intelligence unit.

In the step of processing and analyzing the image of the ground input through the camera in frame units, the artificial intelligence unit pans, tilts, and zooms the camera to the place where the dangerous situation occurs to focus and stores the image, Switching to the direct monitoring state; further includes.

The step of determining the dangerous situation includes determining the dangerous situation when the obstacle is different from a preset reference value.

The artificial intelligence algorithm includes at least one of machine learning, neural network, genetic, deep learning, and image classification algorithms.

Alarms according to the dangerous situation are classified visually or audibly.

Learning the learning model may include capturing an image of the captured obstacle and collecting data; extracting a keypoint score from the collected data through an autoencoder; learning an obstacle model with the extracted feature points; and determining whether the obstacle belongs to the normal category.

The step of determining that the dangerous situation may include classifying the acquired image into a normal image and a bad image through cos-similarity between images in the image using a feature point detection algorithm; Extracting and matching keypoints through comparison between different images; Calculating a keypoint scoring (KPS) representing a degree of similarity between each image; classifying each image with the feature point score calculated in the calculating step; and outputting the classification number and probability value classified in the classifying step as a determination result.

The method further includes collecting images of the defective images and re-learning them. At this time, the re-learning step is performed by a remote server.
The re-learning step is performed by a remote server.
The present invention includes a camera that processes and transmits images of the ground in real time; and an artificial intelligence unit which learns a classifier for distinguishing a stone from a soil lump with the image generated by the camera and calculates a rate at which soil lumps are judged to be stones as defective using the learned classifier.
The present invention is an AI-based detachable safety detection device for agricultural machinery using deep learning in which an image of a camera is learned by an artificial intelligence unit and controlled by the artificial intelligence unit, wherein the artificial intelligence unit measures data related to the image of a soil lump or stone is measured in real time, the measurement data is collected, a learning model of the collected data is created, and then the learning model is verified, and for the safety diagnosis of the obstacle of the input image, the characteristics of the type are extracted, , If the feature of the type is a vibrating body, the extracted 1-dimensional vibration signal generated by the obstacle is converted into a 2-dimensional gray image, and the correlation in the 1-dimensional image is confirmed to characterize the obstacle by type. After extracting, preventive measures can be taken by diagnosing and analyzing the condition of the obstacle through the learning model that has passed the verification of the verification unit.
The camera includes a subminiature control board 220 for controlling to operate at least one sensor that generates a sensing value at a preset period; A magnetic body 300 providing magnetism so that the sensor unit and the microcontroller board are attached to one side of the iron plate; is additionally connected.
The camera captures external images in real time, decomposes them into clip images for analysis by deep learning, and transmits them.
The magnetic body 300 is made of an ND magnet that can be easily mounted on a certain part of an agricultural machine.
The camera is connected to a vibrating body for notification of sound and screen vibration.
The subminiature control board is a Raspberry Pi or Arduino based control board.
The subminiature control board classifies and detects a strong object with an impact amount of more than a certain value and a weak object with an impact amount of less than a certain value in the event of an accident through deep learning.
The camera is connected to a display device displaying a deep learning analysis result of the subminiature control board.
The sensor unit is connected to a harvesting IC used when necessary after finely harvesting and accumulating vibration energy from agricultural machinery.
The present invention is an AI-based detachable safety detection device for agricultural machinery equipped with a deep learning function, comprising: an output unit for outputting a visual and audible alarm; A communication unit for exchanging data with a plurality of cameras; and an artificial intelligence unit that controls image components related to dangerous situations to be learned using an artificial intelligence algorithm, wherein the artificial intelligence unit processes and analyzes images collected from the image collection device through the communication unit in frame units. and recognizes one or more objects from the collected images based on the learned algorithm, determines a dangerous situation based on the size of the one or more objects, and generates an alarm according to the dangerous situation through the output unit.
Control the camera to focus on the location where the dangerous situation occurs to start video recording, determine a dangerous situation based on the size of the one or more objects, generate an alarm according to the dangerous situation through the output unit, and directly monitor control to switch state.
The artificial intelligence unit learns image components related to dangerous situations using an artificial intelligence algorithm, processes and analyzes images collected from image collection devices connected through the network in frame units, and based on the learned algorithm Recognize one or more objects from the collected images, determine a first-level dangerous situation based on the movement of the one or more objects, generate an alarm according to the first-stage dangerous situation, and set the image collection device to the first-stage dangerous situation. Video recording is started by focusing on the place where the situation occurs, it is determined as a second-stage dangerous situation based on the amount of change in the movement of the one or more objects, an alarm is generated according to the second-stage dangerous situation, and a direct monitoring state is switched.

As described above, according to the embodiments of the present invention, unlike conventional machine vision inspection techniques, the automatic defect detection system in the AI-based detachable safety detection method for agricultural machinery using image deep learning has excellent field applicability and great differentiation. Since it has , it is possible to dramatically lower the introduction cost by enabling machine vision-based inspection even in areas where it is difficult to formalize defects and rely on the naked eye.

In addition, the present invention distinguishes objects that may cause problems in a collision using deep learning through Raspberry Pi and reinforces safety when using agricultural machinery.

In addition, the present invention should be in a form that can be easily installed by anyone, and intuitive notifications such as sound and screen vibration are possible.

In addition, according to embodiments of the present invention, human error can be eliminated through an automatic defect detection system, cost of quality control can be minimized, and productivity of machinery can be improved.

In addition, according to embodiments of the present invention, it is possible to provide an automated learning and optimization method even when the shapes of rocks and soil on the ground surface are frequently changed due to weather.

1 is a diagram showing the configuration of an AI-based detachable safety detection device for agricultural machinery using deep learning according to an embodiment of the present invention.
2 is a diagram showing a flowchart for determining a risk and generating an alarm in an AI-based detachable safety detection method for agricultural machinery using deep learning according to an embodiment of the present invention.
3 is a diagram showing a flow chart for diagnosing the state of an obstacle with a two-dimensional grayscale image in the AI-based detachable safety detection method for agricultural machinery using deep learning according to an embodiment of the present invention.
4 is a flowchart showing a transition to a focusing image storage state in an AI-based detachable safety detection method for agricultural machinery using deep learning according to another embodiment of the present invention.
5 is a flowchart showing a comparison between an operation value of an obstacle image and a reference value in an AI-based detachable safety detection method for agricultural machinery using deep learning according to another embodiment of the present invention.
6 is a diagram showing a plurality of algorithms used in an AI-based detachable safety detection method for agricultural machinery using deep learning according to another embodiment of the present invention.
7 is a view showing solving the problem of difficult determination of soil lumps and rocks in the AI-based detachable safety detection method for agricultural machinery using deep learning according to another embodiment of the present invention.

In order to fully understand the present invention, preferred embodiments of the present invention will be described with reference to the accompanying drawings. Embodiments of the present invention may be modified in various forms, and the scope of the present invention should not be construed as being limited to the examples described in detail below. This embodiment is provided to more completely explain the present invention to those skilled in the art. Therefore, the shapes of elements in the drawings may be exaggerated to emphasize a clearer explanation. It should be noted that in each drawing, the same members are sometimes indicated by the same reference numerals. In addition, detailed descriptions of well-known functions and configurations that may unnecessarily obscure the subject matter of the present invention are omitted.

Hereinafter, the present invention will be described in more detail with reference to the drawings.

As shown in FIG. 1 , the camera 210 includes at least one sensor that generates image values at a preset period, and controls the camera 210 to operate; A magnetic body 300 providing magnetism so that the camera 210 and the miniature control board 220 including the artificial intelligence unit are attached to one side of the iron plate for agricultural machinery.

The present invention includes a camera that processes and transmits images of the ground in real time; and an artificial intelligence unit which learns a classifier for distinguishing a stone from a soil lump with the image generated by the camera and calculates a rate at which soil lumps are judged to be stones as defective using the learned classifier.

The artificial intelligence unit collects the measurement data with the image processing and analysis unit 110 that measures measurement data on the image of the soil or stone in real time, creates a learning model of the collected data, and then verifies the learning model. Including, the risk situation determination unit 120 extracts the characteristics of the type for the safety diagnosis of the obstacle of the input image, and extracts it from the vibrating body of the approaching obstacle. The vibration signal of the 1D vibrating body is converted into a 2D gray image, and features are extracted for each type of obstacle by checking the correlation in the 1D image. Through this, preventive measures can be taken by diagnosing and analyzing the condition of obstacles.

The camera may include a sensor unit including at least one sensor that generates a sensing value at a preset period; an artificial intelligence unit controlling the sensor unit to operate; A magnetic material that provides magnetism so that the sensor unit and the microcontroller board are attached to one side of the iron plate; is additionally connected.

As shown in FIG. 5, deep learning is a technology used to cluster or classify objects or data. Computers cannot distinguish between rocks and lumps of dirt only with photos, but humans can distinguish them very easily.

To this end, a method called 'Machine Learning' was devised, which is a technique that inputs a lot of data into a computer and classifies similar ones. If a photo similar to the stored stone photo is entered, the computer classifies it as a stone photo.

Many machine learning algorithms have already appeared on how to classify such image data, but deep learning is a machine learning method proposed to overcome the limitations of artificial neural networks.

Deep learning creates a deep neural network by imitating a human neural network and learns and solves the weight problem with a network created by stacking layers between input and output. The deep learning used here shows strength in high-level abstraction of objects, which showed the most weakness in existing machine learning techniques, through a combination of several nonlinear transformation techniques.

In an embodiment of the present invention, an AI-based detachable safety detection method for agricultural machinery using deep learning is presented.

For example, in classifying soil and stones in a farm where plowing or weeding works, deep learning techniques are used to learn data such as stones, soil lumps, and trees of objects to be detected, and to automatically discriminate alarms or controls. is an automated way to do it.

In the quality inspection method according to an embodiment of the present invention, an optimized discrimination model is automatically generated by inputting learning data (images of rocks or soil lumps). If the image is irregular, it can be discriminated through learning even in an atypical pattern.

In addition, in the quality inspection method according to an embodiment of the present invention, by including noise in input data and learning, strength is shown in changes in external environmental factors, and re-learning and detection model replacement are possible through online user feedback information.

Here, automatic object detection is performed by the camera at regular time intervals on the generated image, and real-time processing can be performed by applying tracking technology using a deep learning network to the detected object, minimizing the area where the object is searched You can design and train a deep network by applying the method.

Hereinafter, the agricultural machine safety detection method will be described in detail with reference to FIGS. 2 to 5 .

The present invention is an AI-based detachable safety detection method for agricultural machinery equipped with a deep learning function controlled by an artificial intelligence unit, in which the artificial intelligence unit processes and transmits an image of the ground in real time, and transmits an image input through a camera frame by frame. processing and analyzing; learning a learning model using an artificial intelligence algorithm for an image component related to a dangerous situation in the image frame analyzed by the artificial intelligence unit; recognizing one or more objects from the collected images based on the learned algorithm by the artificial intelligence unit, and determining a dangerous situation based on states of the one or more objects; and generating an alarm according to the dangerous situation by the artificial intelligence unit.

The step of determining the dangerous situation by the artificial intelligence unit may include: extracting characteristics of rocks, soil lumps, trees, and crop types in order to diagnose a ground situation; converting the extracted one-dimensional image signal into a two-dimensional gray-scale image by the artificial intelligence unit, and extracting features for each type by confirming a correlation between samples; and diagnosing the state of the obstacle through the learning model that has passed the verification by the artificial intelligence unit.

In the step of processing and analyzing the image of the ground input through the camera in frame units, the artificial intelligence unit pans, tilts, and zooms the camera to the place where the dangerous situation occurs to focus and stores the image, Switching to the direct monitoring state; further includes.

The step of determining the dangerous situation includes determining the dangerous situation when the obstacle is different from a preset reference value.

The artificial intelligence algorithm includes at least one of machine learning, neural network, genetic, deep learning, and image classification algorithms.

Alarms according to the dangerous situation are classified visually or audibly.

Learning the learning model may include capturing an image of the captured obstacle and collecting data; extracting a keypoint score from the collected data through an autoencoder; learning an obstacle model with the extracted feature points; and determining whether the obstacle belongs to the normal category.

The step of determining that the dangerous situation may include classifying the acquired image into a normal image and a bad image through cos-similarity between images in the image using a feature point detection algorithm; Extracting and matching keypoints through comparison between different images; Calculating a keypoint scoring (KPS) representing a degree of similarity between each image; classifying each image with the feature point score calculated in the calculating step; and outputting the classification number and probability value classified in the classifying step as a determination result.

The method further includes collecting images of the defective images and re-learning them. At this time, the re-learning step is performed by a remote server.

So far, AI-based detachable safety detection method and safety device for agricultural machinery using deep learning have been described in detail with preferred embodiments.

In the embodiment of the present invention, a deep learning algorithm is applied to classify products and discriminate defects in a manufacturing plant, and through deep learning, a defect detection model is automatically created through image data learning to be atypical rather than a simple pattern. It made it possible to analyze even complex video images.

In addition, in the embodiments as shown in FIG. 7, a detection model can be automatically created by learning data, and self-optimization through online learning through user feedback is possible to provide high flexibility, thereby providing vision inspection. It can lower system introduction cost and change cost.

On the other hand, the present invention is a camera for processing and transmitting the image of the ground in real time; and an artificial intelligence unit which learns a classifier for distinguishing a stone from a soil lump with the image generated by the camera and calculates a rate at which soil lumps are judged to be stones as defective using the learned classifier.
The present invention is an AI-based detachable safety detection device for agricultural machinery using deep learning in which an image of a camera is learned by an artificial intelligence unit and controlled by the artificial intelligence unit, wherein the artificial intelligence unit measures data related to the image of a soil lump or stone is measured in real time, the measurement data is collected, a learning model of the collected data is created, and then the learning model is verified, and for the safety diagnosis of the obstacle of the input image, the characteristics of the type are extracted, , If the feature of the type is a vibrating body, the extracted 1-dimensional vibration signal generated by the obstacle is converted into a 2-dimensional gray image, and the correlation in the 1-dimensional image is confirmed to characterize the obstacle by type. After extracting, preventive measures can be taken by diagnosing and analyzing the condition of the obstacle through the learning model that has passed the verification of the verification unit.
The camera includes a subminiature control board 220 for controlling to operate at least one sensor that generates a sensing value at a preset period; A magnetic body 300 providing magnetism so that the sensor unit and the microcontroller board are attached to one side of the iron plate; is additionally connected.
The camera captures external images in real time, decomposes them into clip images for analysis by deep learning, and transmits them.
The magnetic body 300 is made of an ND magnet that can be easily mounted on a certain part of an agricultural machine.
The camera is connected to a vibrating body for notification of sound and screen vibration.
The subminiature control board is a Raspberry Pi or Arduino based control board.
The subminiature control board classifies and detects a strong object with an impact amount of more than a certain value and a weak object with an impact amount of less than a certain value in the event of an accident through deep learning.
The camera is connected to a display device displaying a deep learning analysis result of the subminiature control board.
The sensor unit is connected to a harvesting IC used when necessary after finely harvesting and accumulating vibration energy from agricultural machinery.
The present invention is an AI-based detachable safety detection device for agricultural machinery equipped with a deep learning function, comprising: an output unit for outputting a visual and audible alarm; A communication unit for exchanging data with a plurality of cameras; and an artificial intelligence unit that controls image components related to dangerous situations to be learned using an artificial intelligence algorithm, wherein the artificial intelligence unit processes and analyzes images collected from the image collection device through the communication unit in frame units. and recognizes one or more objects from the collected images based on the learned algorithm, determines a dangerous situation based on the size of the one or more objects, and generates an alarm according to the dangerous situation through the output unit.
Control the camera to focus on the location where the dangerous situation occurs to start video recording, determine a dangerous situation based on the size of the one or more objects, generate an alarm according to the dangerous situation through the output unit, and directly monitor control to switch state.
The artificial intelligence unit learns image components related to dangerous situations using an artificial intelligence algorithm, processes and analyzes images collected from image collection devices connected through the network in frame units, and based on the learned algorithm Recognize one or more objects from the collected images, determine a first-level dangerous situation based on the movement of the one or more objects, generate an alarm according to the first-stage dangerous situation, and set the image collection device to the first-stage dangerous situation. Video recording is started by focusing on the place where the situation occurs, it is determined as a second-stage dangerous situation based on the amount of motion change of the one or more objects, an alarm is generated according to the second-stage dangerous situation, and a direct monitoring state is switched.
Of course, the technical idea of the present invention can also be applied to a computer-readable recording medium containing a computer program for performing the functions of the device and method according to the present embodiment. In addition, technical ideas according to various embodiments of the present invention may be implemented in the form of computer readable codes recorded on a computer readable recording medium. The computer-readable recording medium may be any data storage device that can be read by a computer and store data. For example, the computer-readable recording medium may be ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical disk, hard disk drive, and the like. In addition, computer readable codes or programs stored on a computer readable recording medium may be transmitted through a network connected between computers.

In addition, although the preferred embodiments of the present invention have been shown and described above, the present invention is not limited to the specific embodiments described above, and the technical field to which the present invention belongs without departing from the gist of the present invention claimed in the claims. Of course, various modifications are possible by those skilled in the art, and these modifications should not be individually understood from the technical spirit or perspective of the present invention.

100: artificial intelligence department
110: image processing and analysis unit
120: risk situation judgment unit
210: camera
220: subminiature control board
300: magnetic body

Claims (23)

The image of the camera is learned by an artificial intelligence unit, and the artificial intelligence unit measures measurement data on a soil lump or stone image in real time, collects the measurement data, creates a learning model of the collected data, and then learns the data. The verification of the model is carried out, and for the safety diagnosis of the obstacle of the input image, a tangible feature is extracted. gray image), extracts features for each type of obstacle by checking the correlation in the 1D image, diagnoses and analyzes the condition of the obstacle through a learning model that has passed the verification of the verification unit, and takes preventive action. In a method using an AI-based detachable safety detection device for agricultural machinery using running,
Processing and analyzing an image input by an artificial intelligence unit through a camera that processes and transmits an image of the ground in real time in frame units;
learning as a learning model using an artificial intelligence algorithm by separating only image components related to dangerous situations from the image frame analyzed by the artificial intelligence unit;
recognizing one or more objects from the input image based on the learned algorithm by the artificial intelligence unit, and determining a dangerous situation based on a state of the one or more objects;
Including; generating an alarm according to the dangerous situation by the artificial intelligence unit,
The step of determining that the artificial intelligence unit is in a dangerous situation,
Extracting characteristics of rocks, soil lumps, trees, and crop types to diagnose the condition of the ground;
converting the extracted one-dimensional image signal into a two-dimensional gray-scale image by the artificial intelligence unit, and extracting features for each type by confirming a correlation between samples;
Including; diagnosing the state of the obstacle through the learning model that has passed the verification by the artificial intelligence unit,
The step of learning the learning model,
Capturing an image of a photographed obstacle and collecting data;
extracting a keypoint score from the collected data through an autoencoder;
learning an obstacle model with the extracted feature points;
Determining whether the obstacle belongs to the normal category; AI-based detachable safety detection method for agricultural machinery using deep learning, characterized in that it comprises. delete The method of claim 1,
The step of processing and analyzing the image of the ground input through the camera in frame units
The artificial intelligence unit pans, tilts, and zooms the camera to the place where the dangerous situation occurs in the image storage device to focus, stores the image, and directly switches to a monitoring state; Deep learning characterized in that it further comprises An AI-based detachable safety detection method for agricultural machinery. The method of claim 1,
The step of determining the dangerous situation is,
An AI-based detachable safety detection method for agricultural machinery using deep learning, comprising the step of determining a dangerous situation when an obstacle is different from a preset reference value. The method of claim 1,
The artificial intelligence algorithm is an AI-based detachable safety detection method for agricultural machinery using deep learning, characterized in that it includes at least one of machine learning, neural network, gene, deep learning, and image classification algorithm. The method of claim 1,
An AI-based detachable safety detection method for agricultural machinery using deep learning, characterized in that the alarm according to the dangerous situation is visually or auditoryly distinguished. delete The method of claim 1,
The step of determining the dangerous situation is,
classifying the acquired image using a feature point detection algorithm into a normal image and a defective image through cos-similarity;
Extracting and matching keypoints through comparison between different images;
Calculating a keypoint scoring (KPS) representing a degree of similarity between each image;
classifying each image with the feature point score calculated in the calculating step;
An AI-based detachable safety detection method for agricultural machinery using deep learning, characterized in that consisting of: outputting the classification number and probability value classified in the classifying step as a determination result. The method of claim 8,
Re-learning by collecting images of the defective image; AI-based detachable safety detection method for agricultural machinery using deep learning, characterized in that it further comprises. The method of claim 9,
In the relearning step,
An AI-based detachable safety detection method for agricultural machinery using deep learning, characterized in that it is performed by a remote server. The image of the camera that processes and transmits the image of the ground in real time is learned by the artificial intelligence unit,
The artificial intelligence unit measures measurement data related to a soil or stone image in real time, collects the measurement data, creates a learning model of the collected data, and then proceeds with verifying the learning model to determine the quality of the input image. For the safety diagnosis of an obstacle, a tangible feature is extracted, and if the tangible feature is a vibrating body, the extracted 1D vibration signal generated by the obstacle is converted into a 2D gray image, By checking the correlation of, it is possible to extract features for each type of obstacle, diagnose and analyze the condition of the obstacle through a learning model that has passed the verification of the verification unit, and take preventive action.
AI-based agricultural machine detachable safety using deep learning, characterized in that a classifier for distinguishing stones and soil lumps is learned with the image generated from the camera, and a ratio of determining soil lumps as stones is defective using the learned classifier. detection device. In the AI-based agricultural machine detachable safety detection device using deep learning where the image of the camera is learned by the artificial intelligence unit and controlled by the artificial intelligence unit,
The artificial intelligence department,
Measuring data on a soil or stone image in real time, collecting the measurement data, generating a learning model of the collected data, and then verifying the learning model,
In order to diagnose the safety of obstacles in the input image, the characteristics of the type are extracted,
If the feature of the type is a vibrating body, the extracted 1-dimensional vibration signal generated by the obstacle is converted into a 2-dimensional gray image, and the correlation in the 1-dimensional image is confirmed to extract features for each type of obstacle. Therefore, it is an AI-based detachable safety detection device for agricultural machinery using deep learning that can diagnose and analyze the state of obstacles through a learning model that has passed the verification of the verification unit and take preventive measures. The method of claim 12,
In the camera,
A miniature control board 220 for controlling to operate at least one sensor that generates a sensing value at a preset period;
A magnetic body 300 that provides magnetism so that the sensor unit and the microcontroller board are attached to one side of the iron plate; AI-based detachable safety detection device for agricultural machinery using deep learning, characterized in that additionally connected.
delete delete delete delete delete delete delete delete delete delete

Images