KR20220147925A - Image detection system - Google Patents

Abstract

The present invention relates to an image detection system capable of detecting falling coal, which determines risk for a coal falling area. According to the present invention, the image detection system comprises: a camera photographing a monitoring area for detecting fallen coal accumulated by falling powdered coal on a conveyor belt capable of transferring the powdered coal; a display unit displaying the monitoring area captured by the camera; and a control unit detecting at least one coal falling area among the monitoring areas displayed on the display unit, setting a fall detection area and a critical area ratio of the coal falling area detected with respect to the fall detection area on the basis of the height and width of the detected coal falling area, and determining that the coal falling area is dangerous when the area ratio of the fall detection area with respect to the coal falling area detected afterward exceeds the critical area ratio.

Description

Image detection system that can detect camels

The present invention relates to an image sensing system, and more particularly, to an image sensing system capable of detecting fallen coal that is transferred to a conveyor belt and falls off the conveyor belt.

There is also a method of transporting coal using trains or vehicles by unloading imported coal to the unloading pier through a ship at a place where coal is used as fuel or a supplier that supplies coal, but it is generally widely used to transport large amounts of coal. The method used is a conveyor belt. In the case of using a conveyor belt, there may be falling and leakage occurring during the transport of coal. Coal exfoliated at the lower part of the belt accumulates as camel over time and is scattered as dust, which causes spontaneous ignition or dust explosion. To prevent this, a camel disposer is deployed to prevent accidents through circuit inspections. However, there is no way to check the status of camels that occur in all places due to the wide range of transport facilities, despite the fact that camels that cause obstacles to operation of equipment must be removed immediately after they occur.

Recently, studies have been conducted on the analysis of fall accidents of workers or the detection of fire occurrence in facilities using cameras, but there is no method of monitoring the fall that is a direct cause of the fire.

Accordingly, it is an object of the present invention to provide an image sensing system capable of detecting a fallen camel on a conveyor and judging a risk to a camel region.

An image sensing system according to the present invention for achieving the above object includes: a camera for photographing a monitoring area for detecting the accumulated coal that has fallen on a conveyor belt capable of transporting the powdered coal; a display unit for displaying the surveillance area photographed by the camera; and detecting at least one camel region among the monitoring regions displayed on the display unit, and based on the detected height and width of the camel region, a critical area ratio of the camel detection region and the detected camel region to the camel detection region and a control unit for determining that the camel area is dangerous when the area ratio of the camel area detected thereafter to the camel detection area exceeds the critical area ratio. After detecting a camel, the critical area ratio of the detected camel detection area and the detected camel detection area is set. You can judge what can be done.

Here, a notification unit that can notify the danger of a camel; and a communication unit capable of transmitting and receiving data to and from an external device, wherein the control unit provides a camel manager capable of removing the camel by using at least one of the notification unit and the communication unit when it is determined that the camel area is dangerous. can inform

and a temperature sensing unit capable of sensing the temperature of the camel region, wherein the control unit stores a critical temperature of the camel region, and when the temperature of at least one region of the camel region exceeds the threshold temperature, the It can be judged that the camel zone is dangerous.

Here, it further comprises a gas detection unit capable of detecting the harmful gas of the camel region, wherein the control unit stores information on the harmful gas of the camel region, and when the harmful gas is detected by the gas detection unit, the camel The area may be considered dangerous.

And an external temperature sensing unit for sensing the external temperature; an external humidity sensing unit for sensing external humidity; and a wind volume sensing unit for sensing the air volume toward the camel area, wherein the control unit sets the external critical temperature, critical humidity, and critical air volume toward the camel area, wherein the outside temperature exceeds the threshold temperature In this case, if at least one of a case in which the external humidity is less than the critical humidity and a case in which the amount of air toward the fallen area exceeds the critical air amount, it may be determined that the fallen area is dangerous.

According to the present invention, after detecting a camel, the critical area ratio of the detected camel detection area and the camel detection area is set, and after that, if the area ratio of the camel area to the camel detection area exceeds the critical area ratio, the camel Since it is possible to determine which areas can accumulate and become dangerous, it is effective in preventing accidents.

1 is an exemplary view of shooting a camel area with an image sensing system according to the present invention.
2 is an exemplary diagram illustrating detecting an object;
3 and 4 are exemplary views of detecting a camel and setting it as a camel region.
5 is an exemplary view of setting a camel detection area for a camel area.
6 is an exemplary view for confirming the ratio of the camel region to the camel detection region.
7 is an exemplary view for determining the height of the camel.
8 is a control block diagram.

Hereinafter, an image sensing system 1 according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is an exemplary view of photographing a camel area with the image detection system 1 according to the present invention, FIG. 2 is an exemplary view showing detecting an object, and FIGS. 3 and 4 are detecting a camel and setting it as a camel area It is an exemplary view, FIG. 5 is an exemplary view of setting a camel detection area for the camel area, FIG. 6 is an exemplary view for confirming the ratio of the camel area to the camel detection area, and FIG. 7 is an exemplary view for determining the camel height and FIG. 8 is a control block diagram.

The image detection system 1 includes a camera 10, a display unit 20, a user input unit 30, a notification unit 40, a communication unit 50, a sensing unit 60, a lighting unit 70, a thermal imaging camera ( 80) and a control unit 80 .

The camera 10 captures an image of a monitoring area for detecting the fallen and accumulated coals on a conveyor belt capable of transporting the powdered coals. The camera 10 may be provided as a plurality of cameras. Each camera may have a lens, an imaging sensor, and a lens driver.

The lens refracts light coming from a point desired by the user to be imaged on the imaging sensor. The imaging sensor transmits an image of the light introduced through the lens to the display unit 20 or the control unit 90 to be described later. The lens driving unit moves the lens in the front-rear direction to focus according to the distance to the area to be photographed.

The display unit 20 displays the surveillance area image captured by the camera 10 . The display unit 20 may display an image having a ratio of a horizontal plane to a vertical plane.

The user input unit 30 is a means for inputting a user's command and may be provided with a keyboard, a mouse, a touch panel, a microphone for voice recognition, a camera capable of recognizing the user's motion, and the like, and the user input unit 30 ) to set the fell detection area using a drag operation on the screen of the display unit 20 .

The notification unit 40 is provided to notify the danger of a camel. The notification unit 40 may be provided as a communication unit to transmit data on the danger of a camel to a user or an administrator. The notification unit 40 may be provided with a speaker, a light, an alarm, and the like.

The communication unit 50 enables transmission and reception of data with the outside. It may be provided by wire or may be provided wirelessly.

The sensing unit 60 may detect various situations and states of the camel area, the area around the camel, and the outside. The sensing unit 60 may include a temperature sensing unit 61 , a gas sensing unit 62 , an external temperature sensing unit 63 , an external humidity sensing unit 64 , and an air volume sensing unit 65 .

The temperature sensing unit 61 may sense the temperature of the camel region. The temperature sensing unit 61 may sense the temperature of the camel region by outputting a laser toward the camel and using a return signal. The temperature sensing unit 61 may sense the temperature using the image. The temperature sensing unit 61 may sense the temperature using a temperature measuring device installed in the vicinity of the camel area.

The gas detection unit 62 may detect harmful gas in the camel region. The gas detection unit 62 is installed in the vicinity of a plurality of the camel region, and is provided to detect the harmful gas by sucking the nearby air.

The external temperature sensing unit 63 may be inside or outside the conveyor belt of the camel region, and may sense an external temperature far away from the camel region.

The external humidity sensing unit 64 may be inside or outside the conveyor belt of the camel area, and may sense external humidity far away from the camel area.

The air volume detecting unit 65 may be installed around the camel area to detect the air volume toward the camel area.

The lighting unit 70 is provided to illuminate the light on the camel area designated within the sensing area. In particular, when it is dark, the captured image can be clearly and accurately made by illuminating the camel region using the lighting unit 70 .

The thermal imaging camera 80 is installed at a position similar to the camera 10 so as to accurately grasp the temperature of the camel area.

The control unit 90 detects at least one camel region among the monitoring regions displayed on the display unit 20, and based on the height and width of the detected camel region, the threshold of the detected camel detection region and the camel detection region After setting the area ratio, if the area ratio of the detected area to the camel detection area exceeds the critical area ratio, it is determined that the camel area is dangerous.

When it is determined that the camel area is dangerous, the controller 90 may notify the camel manager who can remove the camel using at least one of the notification unit 40 and the communication unit 50 . In addition, when it is determined that the camel area is dangerous, the controller 90 may use the communication unit 50 to move the scraper 3 to the camel area to remove the camel from the camel area to the extent that the camel area is not dangerous. The control unit 90 may remotely control the scraper 3 .

The controller 90 may store the critical temperature of the camel region, and when the temperature of at least one of the camel regions exceeds the critical temperature, it may be determined that the camel region is dangerous. The control unit 90 may set a temperature lower than the ignition temperature at which the camel can be ignited as the critical temperature so that a fire can be prevented in advance.

The control unit 90 may store information on the harmful gas in the camel area, and when the noxious gas is detected by the gas detection unit 62, it may be determined that the camel area is dangerous.

The control unit 90 sets an external critical temperature, a critical humidity, and a critical wind volume toward the camel region, and when the external temperature exceeds the critical temperature, when the external humidity is less than the critical humidity, and the air flow toward the camel region is a critical wind volume If at least one of the cases exceeding

Because there is a risk of fire when the temperature is high or the humidity is low and the air volume is large, it can be continuously set and updated using the existing data and new data on the ignition temperature, humidity and air volume. Here, it is supposed to grasp the temperature, humidity, and air volume, but is not limited thereto, and data determined to be the cause of spontaneous ignition may be received and applied.

Here, the camera may be fixed, but may be rotated to detect a plurality of camel areas because the camel area may increase. This may further include a camera driving unit. The camera driving unit drives the rotation of the camera 10 . The camera driving unit may include a support for supporting the camera 10, a turntable having a horizontal rotation motor for rotating the support in a horizontal direction, and a vertical rotation unit having a vertical rotation motor for rotating the turntable in a vertical direction. A camera driving unit may be provided for each camera 10 to drive each of the cameras 10 .

2 is an exemplary diagram illustrating detection of an object.

Object detection technology is an operation of performing localization and classification on a target category existing in an image in a given input image. As shown in FIG. 2 , localization is a regression operation for specifying a bounding box by finding the location of an object, and classification operation is an operation for determining what an object is within a bounding box. Object detection technology is largely divided into single object detection technology and multi-object detection technology.

Among them, the single object detection technique is not suitable for this study because camels are simultaneously piled up in various locations at the bottom of the conveyor. In the multi-object detection technology before deep learning, a method of traversing an area where an object may exist has been predominantly searched using a patch of a predetermined size specified by the user. However, in the case of this method, the number of areas to be searched is too large and the calculation time is excessively consumed because features between each area are not shared. To solve this problem, a region proposal algorithm that detects an object by selecting a target region rather than searching all regions has been proposed. The region proposal algorithm is a method of extracting candidate regions of an object by using characteristic similarities such as color, pattern, size, and shape in the image. The algorithm calculates an object candidate area through three steps: initial area generation, large area generation through integration of small areas, and candidate area generation based on the integrated area. However, in the case of this method, there is a limitation that parameters such as color similarity, feature points, and size similarity must be set by a person for the method of classifying similar regions, and optimal parameters must be found and set for each application environment.

Here, deep learning-based object detection techniques using CNN, which have been actively studied recently, are being studied so that they can be used universally without being limited to the environment. The deep learning-based object detector can be divided into a 1-stage detector that simultaneously processes a localization problem and a classification problem, and a 2-stage detector that sequentially processes the two problems. Representative 1-stage detectors include R-CNN, Fast R-CNN, and Faster RCNN, and as 2-stage detectors, algorithms such as YOLO and SSD exist. In case of 1-stage detector, it shows relatively high accuracy compared to 2-stage detector, and 2-stage detector shows relatively fast execution speed compared to 1-stage detector, and the object detection result is returned in the form of bounding box. . However, the bounding box-based object detection has a limitation in that it cannot specify the shape of an object existing within the bounding box. Recently, studies have been conducted to solve this problem. As a representative object region segmentation study, the Mask-RCNN model that replaces the RoI pooling step of Faster-RCNN with RoI align and adds the image region segmentation function was proposed. In the case of the model, there is an advantage in that the shape of the detected object can be specified along with the robust detection performance of the existing 1-stage detector. Due to the excessively high amount of computation, high hardware specifications are required, or there is a limit to analyzing multiple Carrera images at the same time. Accordingly, the present invention is a method for monitoring the width and height of a camel in real time with a low computational amount by using the previously studied 1-Stage neural network.

3 and 4 are exemplary views of detecting a camel and setting it as a camel region.

3 (a) In order to specify the area of the camel, the detection of an object is prioritized. Representative 1-stage detectors for detecting the object include YOLO and SSD. Both algorithms are algorithms that perform the localization process of an object using a predefined anchor box in a feature map obtained as a result of image convolution.

3 (b) and 4, the structure of such a neural network optimizes the amount of computation by specifying the shape of the anchor box for detection of camels deposited in the form of a wide cone, and the image input of the neural network according to the position or magnification of the installed camera It is a neural network suitable for the proposed system because its size can be freely adjusted.

5 is an exemplary diagram of setting a camel detection area for a camel area.

The detection area for detecting the height of the camel may be formed of a combination of n squares. The size and position are set according to the structure in the camera image, and the stage of the detection area can be dynamically set.

The detection area on the left consists of A and B, the middle detection area consists of A, B, and C, and the rightmost side consists of 100 to 500. The detection area may be automatically set by the controller in consideration of the size, width and height of the camel area, or may be set by the user.

6 is an exemplary view for confirming the ratio of the camel region to the camel detection region.

개의 프레임을 샘플링하여 수행한다. 샘플링된 매 프레임마다 입력된 카메라 영상을 이용하여 낙탄의 영역을 검출하고, 검출된 낙탄의 영역과 설정된 각 검지 영역이 겹쳐지는 넓이의 비율을 구한다. 그 후, 각 영역마다 겹쳐지는 넓이가 낙탄 검출 임계값보다 큰 경우에 해당 영역에 낙탄이 존재하는 것으로 판단한다. 상기 작업을

번 반복하여 각 영역마다 낙탄이 검출된 횟수를 계수한다. 도 6 은 각 검지 영역에 대하여,

번 낙탄의 검출 정도를 확인하는 작업에 대한 흐름도이다.The height of the camel is measured per second

This is done by sampling the frames. The area of the camel is detected using the camera image input for each sampled frame, and the ratio of the overlapping area between the detected area of the camel and each set detection area is obtained. Thereafter, when the overlapping area for each area is greater than the camel detection threshold, it is determined that the camel is present in the corresponding area. the above work

The number of times that a camel is detected is counted in each area by repeating. 6 shows each detection area,

It is a flowchart for the operation of confirming the detection degree of the burnt camel.

7 is an exemplary view for determining the height of the camel.

개의 프레임을 샘플링하여 수행한다. 지정된 프레임 수만큼의 반복 후 얻어진 낙탄 검지 영역별 검출 값,

를 이용하여 최종 낙탄의 높이를 판단한다. 높은곳에 위치한, 우선순위가 높은 영역부터 시작하여, 해당 영역의

값이 사전에 설정된 높이 검출 프레임 임계값을 넘는 경우 해당 높이만큼의 낙탄이 존재하는 것으로 판단한다. 도 7은 낙탄의 높이 판단 과정을 나타낸 흐름도이다.The height of the camel is measured per second

This is done by sampling the frames. The detection value for each camel detection area obtained after repeating the specified number of frames,

is used to determine the height of the final camel. Starting with the high-priority area located at the top,

When the value exceeds the preset height detection frame threshold, it is determined that a camel corresponding to the corresponding height exists. 7 is a flowchart illustrating a process of determining the height of a camel.

Modifiable embodiments other than the above embodiments will be described.

The controller 90 may store various shapes and shapes of the camel area. It is also possible to determine the ignition risk of the camel area if the shape and shape of the currently measured camel area are similar to the shape and shape of the camel area with high firing frequency by identifying the frequency of ignition occurrence for the shape and shape of the camel area. In this case, data of the temperature of the camel area, the presence or absence of harmful gas, external temperature, external humidity, and air volume may be applied.

The control unit 90 may determine the risk of the camel area by determining the formation time of the camel area and the rate of increase of the camel area. For example, if the formation time of the camel region is faster than the average formation time of the camel region by a predetermined time or more, or if the rate of increase of the camel region is faster than the increase rate of the average camel region by more than a predetermined time, the probability that the camel region will ignite is high. can judge

It may further include a water sprayer capable of spraying water toward the camel area. This water sprayer may spray water into the camel area before being removed by the scraper when it is determined that the camel area is in a dangerous state under the control of the controller 90 .

Due to the above image detection system (1), after detecting a camel, the critical area ratio of the detected camel area to the camel detection area and the camel detection area is set, and then the area ratio of the camel area to the camel detection area is If the critical area ratio is exceeded, it can be determined that the camel area can be piled up and become dangerous.

1: Image detection system 2: Conveyor belt
3: Scraper 4: Camel
10: camera 20: display unit
30: user input unit 40: notification unit
50: communication department
60: sensing unit 61: temperature sensing unit
62: gas sensing unit 63: external temperature sensing unit
64: external humidity sensing unit 65: air volume sensing unit
70: lighting unit 80: thermal imaging camera
90: control unit

Claims (5)

In the image sensing system,
A camera for photographing a monitoring area for detecting the fallen coal that has fallen on a conveyor belt capable of transporting the powdered coal;
a display unit for displaying the surveillance area photographed by the camera; and
Detects at least one camel area among the monitoring areas displayed on the display unit, and determines a critical area ratio of a camel detection area and the detected camel area with respect to the camel detection area based on the detected height and width of the camel area and a controller configured to determine that the camel area is dangerous when the area ratio of the detected area to the camel detection area exceeds the threshold area ratio. The method of claim 1,
A notification unit that can inform the danger of a camel; and
Further comprising a communication unit capable of transmitting and receiving data to an external device,
The control unit is
When it is determined that the camel area is dangerous, the image sensing system notifies a camel manager capable of removing the camel using at least one of the notification unit and the communication unit. The method of claim 1,
Further comprising a temperature sensing unit capable of sensing the temperature of the camel region,
The control unit is
The image sensing system stores the critical temperature of the camel region, and determines that the camel region is dangerous when the temperature of at least one of the camel regions exceeds the threshold temperature. The method of claim 1,
Further comprising a gas detection unit capable of detecting the harmful gas of the camel region,
The control unit is
An image sensing system that stores information on the harmful gas in the camel area, and determines that the camel area is dangerous when noxious gas is detected by the gas sensing unit. The method of claim 1,
an external temperature sensing unit for sensing an external temperature;
an external humidity sensing unit for sensing external humidity; and
Further comprising a wind volume sensing unit for sensing the air volume toward the camel area,
The control unit is
Set the external critical temperature, critical humidity, and critical wind volume toward the camel area, and when the outside temperature exceeds the critical temperature, when the outside humidity is less than the critical humidity, and the air volume toward the camel area is the critical air volume An image sensing system that determines that the camel area is dangerous if at least one of the cases exceeds.

Images