TWI783608B - Smart system for monitoring dust raising - Google Patents

Abstract

The present invention provides a smart system for monitoring dust raising and the related monitoring method. The system has a picture capturing device for taking a first image of an area, a processing unit for quantifying the first image by converting the first image into a first pixel array. After being processed, the first pixel array is calculated and its noise is removed by setting a threshold value so that the thus formed image pixel array has a pixel intensity for a first place and a second place. The pixel intensity can be used with the comparison of a particulate matter concentration and its threshold concentration so that a precise determination can be obtained to improve the identification efficiency and accuracy.

Description

Smart Dust Monitoring and Management System and Method

The present invention relates to a dust monitoring and management system and method, in particular to a smart dust monitoring and management system and method.

Flying dust pollution refers to the air pollution caused by particles generated during construction and material stacking. Particles with larger particle sizes in the flying dust are often blocked in the upper respiratory system. When the particle size of the flying dust is below 10 μm (PM10) It will enter the lower respiratory tract, and if the particle size is below 2.5 μm (PM2.5), it will accumulate in the alveoli, causing a series of diseases, and in severe cases, it can lead to lung failure and death.

Therefore, during the construction of the construction site, if the prevention and control measures are not installed in time, it is very easy to cause particulate matter dust pollution, which will affect the air quality in the vicinity. Set up various air pollution prevention and control facilities, including engineering fences, dust-proof nets, dust-proof cloths, car washing equipment, etc., and strengthen protection against vehicle paths and bare ground that may generate fugitive pollution.

For example, construction projects often use engineering materials such as sand, gravel and earthwork with dust dissipating properties. Preventive measures such as covering dust-proof cloths, dust-proof nets or regularly spraying chemical stabilizers should be taken. It is necessary to carry out greening, surface compaction and watering to suppress the generation of dust.

In order to solve these problems in recent years, a lot of manpower has been spent on inspections. When the dust is too large, it is necessary to use manpower to spray on the dust, or install air quality detection devices for monitoring. However, these detection devices are currently installed in construction sites. Limited by the power supply, it needs to be set in a fixed way. However, the general construction site power supply has not yet been fully planned, and it is not easy to set it according to the polluted hot zone.

In addition, the construction site is vast, and there are often multiple constructions at the same time, and each construction site is not consistent. Therefore, the current air quality monitoring devices cannot be installed flexibly, resulting in ineffective monitoring. In addition, some construction sites have video recording equipment, but the video recording equipment is difficult to broadcast the dust situation in real time and judge the amount of dust on the site and the scope of influence, and it is impossible to deal with it in a timely manner if it is judged manually. There are many factors that may lead to the occurrence of dust in the construction site, so finding the exact source of pollution can effectively solve the problem of dust.

The smart dust monitoring and management system and method of the present invention combine image capture devices to capture images of air quality conditions and construction environments, which are further converted into a matrix of pixel values by an arithmetic processing unit, and after deducting reference values, the preset threshold is used as Noise is filtered out, and identification results can be generated. This identification result can be compared with the suspended particle concentration value obtained by the air detection device, and the judgment result can be accurately produced, which greatly improves the identification efficiency and accuracy, and can be further improved. Use Convolutional Neural Networks to learn the collected pollution and environmental data, understand the dust transmission path and trace the dust pollution source, and then use the Sobel edge detection algorithm to calculate the dust spread and influence range from the pixel value matrix, and then warn personnel Apply cloth cover or other disposal measures.

The main purpose of the invention is to provide a smart dust monitoring and management system. The sensing device is combined with the image capture device to capture the image of air quality and construction environment, and can further use the The collected pollution and environmental data is learned by using a neural network to understand the dust transmission path and trace back to the source of dust pollution.

The first image of the area is captured by an image capture device, and the image is converted into an image pixel value matrix by an operation processing unit for quantization, and then the image pixel matrix is calculated and the preset threshold value is used to filter out noise signals, and further Use the suspended particle concentration value to confirm whether the standard is met, and obtain the judgment result.

In order to achieve the above-mentioned purpose, one embodiment of the present invention discloses a smart dust monitoring and management system, including: an image capture device, set in an area, and captures a first position and a second position of the area at a time A first image of the location; a sensing device, arranged in the area, to sense a suspended particle concentration value in the area at the time; and an arithmetic processing unit, respectively communicating with the image capturing device and the sensing device for signals connecting, converting the first image into a first image pixel value matrix, and the first image pixel value matrix includes a first pixel value intensity corresponding to the first position and a second pixel value intensity corresponding to the second position , and receive the aerosol concentration value; wherein, the arithmetic processing unit subtracts a reference image pixel value matrix from the first image pixel value matrix to generate a second image pixel value matrix, when the second image pixel value matrix corresponds to When the intensity of the first pixel value at the first position and the intensity of the second pixel value at the second position are both less than zero, a first judgment result is generated, and when the pixel value matrix of the second image corresponds to the first When the intensity of the first pixel value at the position or the intensity of the second pixel value at the second position is greater than zero, a preset threshold value is subtracted from the second image pixel value matrix to generate a third image pixel value matrix. When the intensity of the third pixel value of the third image pixel value matrix corresponding to the first position and the intensity of the fourth pixel value of the second position are both less than zero, a second judgment result is generated, when the third image pixel When the intensity of the third pixel value of the value matrix corresponding to the first position or the intensity of the fourth pixel value of the second position is greater than zero, and the aerosol concentration value is less than a threshold value of aerosol concentration, a first 3. Judgment result, when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix or the fourth pixel value of the second position When the pixel value intensity is greater than zero, and the suspended particle concentration value is greater than the suspended particle concentration threshold value, a fourth judgment result is generated; wherein, the first judgment result, the second judgment result and the third judgment result are the There is no dust in the first position and the second position; wherein, the fourth judgment result includes: when the intensity of the third pixel value of the third image pixel value matrix corresponding to the first position is greater than zero, and corresponding to the third pixel value matrix The intensity of the fourth pixel value at the second position is less than zero, and the concentration value of the suspended particles is greater than the threshold value of the concentration of suspended particles, then it is judged that there is dust in the first position, and the concentration value of the suspended particles has reached the upper limit, and the second position No dust; when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix is less than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particles If the concentration value is greater than the threshold value of the suspended particle concentration, it is judged that there is no dust in the first position, and there is dust in the second position, and the concentration value of the suspended particle has reached the upper limit; and when the pixel value matrix of the third image corresponds to the When the intensity of the third pixel value at the first position is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particle concentration value is greater than the suspended particle concentration threshold value, then it is determined that the first There is dust at the location and the second location, and the concentration of suspended particulates has reached the upper limit.

Preferably, the suspended particle concentration value is the concentration value of fine suspended particles (PM2.5) and suspended particles (PM10) in the air.

Preferably, the operation processing unit uses a Sobel edge detection algorithm and the fourth judgment result to obtain a gradient map, and integrates according to the gradient map to generate a dust pollution range and a range corresponding to the dust pollution range. 1. Dust pollution intensity.

Preferably, it further includes a sprinkler unit, which is connected with the signal of the operation processing unit, and the operation processing unit generates a sprinkler signal according to the fourth judgment result and corresponds to a sprinkler position with dust to the sprinkler unit. The sprinkler unit turns on the sprinkler device to sprinkle water according to the sprinkler signal and the sprinkler position.

Preferably, the arithmetic processing unit further obtains wind direction information and a wind speed value sensed by the sensing device in the area at the time, and substitutes the fourth judgment result, the wind direction information and a wind speed value into a convolutional neural network model The wind speed value is compared and calculated to generate a comparison result. The comparison result includes the dust generated at the first location, and correspondingly affects the first location, or the dust generated at the first location, and the second location The flying dust in the first location comes from the first location, or the flying dust generated in the second location, and the flying dust in the first location comes from the second location, or the flying dust generated in the second location, and affects the second location accordingly.

Another object of the present invention is to provide a smart dust monitoring and management method, which implements the method with a smart dust monitoring and management system to capture the first image of the area and convert it into a matrix of pixel values. The suspended particle concentration value further judges whether the suspended particle concentration threshold value is exceeded, and a judgment result is generated to accurately and effectively confirm the dust raising situation.

In order to achieve the above object, another embodiment of the present invention discloses a smart dust monitoring and management method, the steps include: an image capture device captures a first position of an area and a first position of a second position at a time An image is sent to an operation processing unit; the operation processing unit converts the first image into a first image pixel value matrix, and subtracts a reference image pixel value matrix from the first image pixel value matrix to generate a second Image pixel value matrix; when the intensity of the first pixel value corresponding to the first position in the second image pixel value matrix and the intensity of the second pixel value in the second position are both less than zero, a first judgment is generated As a result, or when the intensity of the first pixel value corresponding to the first position in the second image pixel value matrix or the intensity of the second pixel value in the second position is greater than zero, the operation processing unit according to the second Deducting a preset threshold value from the image pixel value matrix to generate a third image pixel value matrix; when the third pixel value intensity corresponding to the first position and one of the second position in the third image pixel value matrix When the intensity of the fourth pixel value is less than zero, a second judgment result is generated, or when the intensity of the third pixel value corresponding to the first position or the fourth pixel value of the second position in the third image pixel value matrix When the pixel value intensity is greater than zero, the arithmetic processing unit obtains a suspended particle concentration value at the time from a sensing device; and when the suspended particle concentration value is less than a suspended particle concentration threshold value, generates a first Three judgment results, or when the suspended particle concentration value is greater than the suspended particle concentration threshold value, a fourth judgment result is generated; wherein, the first judgment result, the second judgment result and the third judgment result are the first position and the second position without dust; wherein, the fourth judgment result includes: when the intensity of the third pixel value of the third image pixel value matrix corresponding to the first position is greater than zero, and corresponding to the second position If the intensity of the fourth pixel value is less than zero, and the concentration of suspended particles is greater than the threshold value of the concentration of suspended particles, it is judged that there is dust in the first position, and the concentration of suspended particles has reached the upper limit, and there is no dust in the second position ; when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix is less than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particle concentration value is greater than the threshold value of the suspended particle concentration, it is judged that there is no dust in the first position, and there is dust in the second position, and the concentration of suspended particles has reached the upper limit; and when the pixel value matrix of the third image corresponds to the first When the intensity of the third pixel value at the position is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the concentration value of the suspended particles is greater than the threshold value of the concentration of suspended particles, it is determined that the first position and There is dust at the second location, and the concentration of suspended particulates has reached the upper limit.

Preferably, the suspended particle concentration value is the concentration value of fine suspended particles (PM2.5) and suspended particles (PM10) in the air.

Preferably, it further includes the steps: the operation processing unit performs operation with a Sobel edge detection algorithm and the fourth judgment result to obtain a gradient map; and the operation processing unit integrates according to the gradient map to generate a dust pollution range And one of the dust pollution intensity corresponding to the dust pollution range.

Preferably, it includes the steps of: the arithmetic processing unit generates a sprinkling signal and a sprinkling position corresponding to the dust to a sprinkling unit according to the fourth judgment result; and the sprinkling unit generates a sprinkling signal according to the sprinkling signal and the The water position turns on the sprinkler to sprinkle water.

Preferably, the sensing device senses wind direction information and a wind speed value in the area at the time, and sends them to the computing processing unit; and the computing processing unit substitutes a convolutional neural network model into the fourth judgment result , The wind direction information and the wind speed value are compared and calculated to generate a comparison result; wherein, the comparison result includes the dust generated by the first location, and correspondingly affects the first location, or the dust generated by the first location Dust in the second location comes from the first location, or the dust generated in the second location, and the dust in the first location comes from the second location, or the dust generated in the second location, and the corresponding Affects this second position.

The beneficial effect of the present invention is to carry out its image recognition in an automatic and high-precision manner. After the image is converted into a matrix of pixel values by an arithmetic processing unit, the noise is further filtered out with a preset threshold value. When the pixel value is determined by When the pixel value intensity of the corresponding first position or second position is greater than zero, it is further confirmed that the suspended particle concentration value is greater than its suspended particle concentration threshold value, and the judgment result is obtained, and can be calculated by the wind direction and wind speed value obtained by the sensing device Identify the source of the dust, its identification efficiency is excellent, and the accuracy is high, and it improves the conventional problem that even if the suspended particle detection data is obtained, it is impossible to know where the dust is generated, and it does not require a lot of manpower to check the dust one by one. Elimination, greatly increasing its efficiency in improving dust.

1: Image capture device

2: Sensing device

3: Operation processing unit

4: sprinkler unit

5: Warning unit

S1-S19: Steps

Figure 1: It is a schematic diagram of the system of an embodiment of the present invention; Figure 2A: It is a flow chart of a method of an embodiment of the present invention; Figure 2B: It is a flow chart of a method of another embodiment of the present invention ; and Fig. 2C: it is a partial method flow chart of one embodiment of the present invention.

The features and technical contents of the relevant patent applications of the present invention will be clearly presented in the following detailed description of preferred embodiments with reference to the drawings.

Please refer to Fig. 1, which is a system diagram of an embodiment of the present invention. As shown in the figure, the smart dust monitoring and management system of the present invention includes: an image capture device 1, a sensing device 2, an arithmetic processing unit 3, a sprinkler unit 4 and a warning unit 5, wherein the image capture device 1 and the sensing device The device 2, the sprinkler unit 4 and the warning unit 5 are signal-connected with the computing processing unit 3, and its image capture device 1 and sensing device 2 can be installed on a portable device for easy portability and installation.

In an embodiment of the present invention, please refer to FIG. 2A , which is a flow chart of a method of an embodiment of the present invention. As shown, the method includes the following steps.

Step S1: An image capture device captures a first image of a first position and a second position in an area at a time, and sends it to an operation processing unit; Step S3: The operation processing unit converts the first image The image is a first image pixel value matrix, and a reference image pixel value matrix is subtracted from the first image pixel value matrix to generate a second image pixel value matrix; Step S5: when the corresponding When the intensity of the first pixel value at the first position and the intensity of the second pixel value at the second position are both less than zero, a first judgment result is generated, or when the pixel value in the second image matrix corresponding to the When the intensity of the first pixel value at the first position or the intensity of the second pixel value at the second position is greater than zero, the arithmetic processing unit subtracts a preset threshold value from the second image pixel value matrix to generate a third image Pixel value matrix; step S7: when the third pixel value intensity corresponding to the first position and the fourth pixel value intensity corresponding to the second position in the third image pixel value matrix are both less than zero, generate a first Two judgment results, or when the intensity of the third pixel value corresponding to the first position in the third image pixel value matrix or the intensity of the fourth pixel value in the second position is greater than zero, the arithmetic processing unit performs a the sensing device obtains a value of the aerosol concentration at that time; and Step S9: Generate a third judgment result when the suspended particle concentration is less than a suspended particle concentration threshold, or generate a fourth judgment result when the suspended particle concentration is greater than the suspended particle concentration threshold.

As described in step S1, the image capture device 1 captures the first image of the first position and the second position of the area at a time, and sends it to the processing unit 3. In this embodiment, in the first image, there are The corresponding first position and the second position, as well as the pixel value intensity corresponding to the first position and the pixel value intensity corresponding to the second position, but not limited thereto, multiple positions can also be marked in the first image, and their corresponding The intensity of the pixel value is convenient for subsequent judgment processing.

As described in step S3, the arithmetic processing unit 3 converts the first image so that the first image corresponds to generate a first image pixel value matrix, that is, the image is quantized into a matrix value. In this embodiment, the image size is 1024*768 as an example. But not limited to this, it means that the image consists of a length with a width of 1024 pixels and a length with a width of 768 pixels, a total of 786,432 pixels, so whether it is the first position or the second position, there are different RGB The vector matrix is formed, and the reference image pixel value matrix stored in the operation processing unit 3 is subtracted from the first pixel value matrix to generate the second image pixel value matrix.

As described in step S5, in the second image pixel value matrix, when the intensity of the pixel value corresponding to the first position and the second position is less than zero, the first judgment result is directly generated that the first position and the second position have no dust , then do not continue to the next step. On the contrary, when the intensity of the first pixel value corresponding to the first position or the intensity of the second pixel value corresponding to the second position is greater than zero, the preset threshold value is deducted according to the second image pixel value matrix , and generate the third image pixel value matrix.

The second pixel value matrix obtained above is close to the grayscale image, and the noise is further filtered out with the preset threshold, which will make the judgment more accurate. Therefore, the generated third image pixel value matrix is close to the grayscale image. In the embodiment, dust is characterized by white mist or white smoke. When the preset threshold value is set lower (that is, closer to 0), it means that the sensitivity is higher. Conversely, when the preset threshold value is set higher (that is, it is closer to 0 Closer to 255) means the lower the sensitivity, the accuracy of the judgment will be different, and it should be set according to the needs of the construction site.

As described in step S7, in the third image pixel value matrix, when the intensity of the third pixel value corresponding to the first position and the intensity of the fourth pixel value at the second position are both less than zero, the second judgment result is directly generated as the first If there is no dust in the first position and the second position, do not continue to the next step. On the contrary, when the intensity of the pixel value at any one of the first position or the second position is greater than zero, then continue to the next step. When the intensity of the third pixel value at a position or the intensity of the fourth pixel value corresponding to the second position is greater than zero, it indicates that there may be dust in the first position or the second position. The suspended particle concentration value corresponding to the time.

As described in step S9, the arithmetic processing unit 3 compares the suspended particle concentration value with the suspended particle concentration threshold value. The location is free of dust, and when the concentration of suspended particles is greater than the threshold value of the concentration of suspended particles, the fourth judgment result will be produced, and in this embodiment, the concentration of suspended particles refers to the fine suspended particles in the air (PM2.5) and the concentration of suspended particulates (PM10), but not limited thereto. The threshold value of suspended particulate concentration can be changed according to the needs of the construction site. If the concentration of suspended particulates in the construction site is always high, the relative threshold value of suspended particulates must be increased. , to avoid misjudgment, otherwise, reduce it.

Wherein the fourth judgment result includes the following in detail: the first case is when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is less than zero , and the suspended particle concentration value is greater than the suspended particle concentration threshold value, it is judged that there is dust in the first position, and the suspended particle concentration value has reached the upper limit, and there is no dust in the second position; the second case is when the third image pixel value matrix When the intensity of the third pixel value corresponding to the first position is less than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particle concentration value is greater than the suspended particle concentration threshold value, it is judged that there is no dust in the first position, and there is dust at the second location, and the concentration of suspended particulates has reached the upper limit; and The third situation is when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the concentration value of suspended particles is greater than Particle concentration threshold value, it is judged that there is dust in the first position and the second position, and the concentration of suspended particles has reached the upper limit; if the intensity of the third pixel value corresponding to the first position in the third image pixel value matrix is less than zero, and the corresponding When the intensity of the fourth pixel value at the second position is less than zero, it is determined that there is no dust in the first position and the second position.

When the fourth judgment result is generated, it means that there may be dust in a certain place, and steps S91-S93 can be further included: Step S91: the calculation processing unit performs calculation with a Sobel edge detection algorithm and the fourth judgment result, Obtain a gradient map; and step S93: the calculation processing unit integrates according to the gradient map to generate a dust pollution range and a dust pollution intensity corresponding to the dust pollution range.

In one embodiment, the Sobel edge detection algorithm is performed on the dusty image, and the formula is as follows:

為Sobel水平方向遮罩；

為Sobel垂直方向遮罩。以前述遮罩於一影像內任一位置覆蓋像素點，所取得的灰階值，此一步驟需先計算出遮罩的中心所在像素點梯度，經式子(A)計算後取得梯度值，再將遮罩中心移至下一個像素點，進而重複前述計算過程，而可依序取得所有梯度值，並依據該些梯度值取得影像，即與原來影像同樣大小的梯度圖，而求出影像的邊界，換言之，當運算處理單元3之第四判斷結果為任一位置有揚塵污染時，則可進一步以Sobel 邊緣檢測演算法取得揚塵污染範圍，並進一步利用揚塵污染範圍及其像素值進行積分取得揚塵污染強度。 in,

Mask for the Sobel horizontal direction;

Mask for the Sobel vertical direction. To cover the pixel at any position in an image with the aforementioned mask, and obtain the gray scale value, this step needs to first calculate the gradient of the pixel point where the center of the mask is located, and obtain the gradient value after calculation by formula (A), Then move the center of the mask to the next pixel, and then repeat the above calculation process, so that all the gradient values can be obtained in sequence, and the image can be obtained according to these gradient values, that is, the gradient map with the same size as the original image, and the image can be obtained In other words, when the fourth judgment result of the operation processing unit 3 is that there is dust pollution at any position, the Sobel edge detection algorithm can be used to further obtain the dust pollution range, and further use the dust pollution range and its pixel value for integration Obtain the dust pollution intensity.

In order to provide an automatic solution, please refer to FIG. 2B , which is a flow chart of another embodiment of the present invention, which further includes a sprinkler unit 4 . As shown in the figure, this embodiment further includes the following steps S11-S13 after step S9.

Step S11: the arithmetic processing unit generates a water sprinkling signal and a water sprinkling position corresponding to the dust to a water sprinkling unit according to the fourth judgment result; and Step S13: the water sprinkling unit generates water according to the water sprinkling signal and the water sprinkling unit location for watering.

As described in step S11, the arithmetic processing unit 3 generates the sprinkler signal and the corresponding sprinkler position with dust to the sprinkler unit 4 according to the fourth judgment result. In this embodiment, the aforementioned situation is divided into three situations, the first In the first case, when the result of the fourth judgment is that there is dust in the first position and there is no dust in the second position, the corresponding sprinkler position with dust is the first position; the second case is when the result of the fourth judgment is that there is dust in the second position. If there is no dust at one location and there is dust at the second location, the corresponding sprinkler location with dust is the second location; and the third situation is when the fourth judgment result is that there is dust at the first location and the second location When there is also dust, then the corresponding sprinkling positions with dust are the first position and the second position.

Finally, as described in step S13, the sprinkler unit 4 sprinkles water according to the sprinkler signal and the sprinkler position, that is, sprinkles water at the corresponding first position or the second position according to the fourth judgment result of the previous step.

To determine exactly which pile or construction place the dust is caused by, please refer to FIG. 2C , which is a flow chart of a method according to an embodiment of the present invention. As shown in the figure, this embodiment further includes steps S15-S17, which are described as follows: Step S15: the sensing device senses a wind direction information and a wind speed value in the area at this time, and sends them to the calculation processing unit; and step S17: the operation processing unit substitutes a convolutional neural network model into the fourth judgment result, the wind direction information and the wind speed value to perform a comparison operation to generate a comparison result; As described in step S15, the arithmetic processing unit 3 obtains wind direction information and wind speed value through the sensing device 2 at the corresponding time, wherein in this embodiment, the sensing device 2 includes a thermometer, a hygrometer, an anemometer and an air detection device , but not limited to.

As described in step S17, the calculation processing unit 3 uses the convolutional neural network model to substitute the fourth judgment result, wind direction information, and wind speed value to perform a comparison operation to generate a comparison result, wherein the comparison result includes the following situations, the first situation Dust is generated at the first location, and the first location is affected accordingly, which means that the first location judges that there is dust, and the dust only affects the range of the first location; the second situation is that the first location generates dust, and the second location The dust comes from the first location, which means that the first location judges that there is dust, and due to the relationship between the wind direction and wind speed value, it flies to the range of the second location, and then affects the second location; the third situation is that the second location generates dust, and The dust in the first location comes from the second location, which means that the second location judges that there is dust, and due to the relationship between wind direction and wind speed, it flies to the range of the first location, thereby affecting the first location; the fourth situation is the second location Generate dust, and correspondingly affect the second location, which means that the second location judges that there is dust, and the dust only affects the range of the second location; finally, both the first location and the second location generate dust, resulting in the first location and the second location. The second positions all affect each other.

Among them, the calculation processing unit 3 compares the wind direction information and wind speed value according to the fourth judgment result, which is to carry out convolutional neural network (Convolutional Neural Networks, CNN) learning, and uses the chain derivation rule to calculate the nodes in the hidden layer. Derivation, that is, gradient descent plus chain derivation law, connecting multiple convolutional layers together is a feature that gradually increases in abstraction. Through input and output training methods, the weight value and partial weight between nodes are constantly adjusted. value, so that the output calculated by the network is the target output, which is used to understand the dust pollution transmission path, and the image pixel value matrix can be used to calculate the dust spread and influence range by using the Sobel edge detection algorithm.

Generally speaking, the CNN structure consists of two layers, one is the feature extraction layer, the input of each neuron is connected to the part of the receptive field of the previous layer, and the features of this part are extracted, once the part of the feature is extracted, it is connected with other The orientation relationship between features is also determined; the second is the feature map layer. Each calculation layer of the network is composed of multiple feature maps. Each feature map is a plane, and the weights of all neurons on the plane are equal. The feature map structure uses the sigmoid function with a small influence function kernel as the activation function of the convolutional network, so that the feature map has displacement invariance. Each unit of the neural network can also be called a Logistic regression (Logistic regression) model. When multiple units are combined and have a hierarchical structure, a neural network model is formed, showing a hidden layer. Convolutional neural network, the corresponding formula is as follows:

為例，其係將輸入層的參數x ₁ 、x ₂ 、x ₃經加權值

、

、

及偏權值

累加後，經式子(C)中的活化函數轉換

、

、

後，取得輸出值h _W,b (x)，最後以此輸出值與觀測結果進行比較，如果不相符再次遞迴進行權重的改變，直到與輸出值與結果接近，此時就可以得到一組權重值與偏權值，用來預測之後之結果，倘若再與後續結果有出入時，則再次進行學習得出新的權重值與偏權值，如此只要經過龐大數據的學習，就會得到較精準的行為模式和預測結果。 First, according to formula (B),

For example, it is the weighted value of the parameters x ₁ , x ₂ , x ₃ of the input layer

,

,

and partial weight

After accumulation, it is transformed by the activation function in formula (C)

,

,

After that, the output value h _W,b ( x ) is obtained, and finally the output value is compared with the observation result. If it does not match, recursively change the weight again until it is close to the output value and the result. At this time, a set of The weight value and partial weight value are used to predict the subsequent results. If there is a discrepancy with the subsequent results, the new weight value and partial weight value will be learned again. In this way, as long as a large amount of data is learned, it will be relatively accurate. Precise behavioral patterns and predicted outcomes.

For this reason, with the known pollution and atmospheric data such as wind direction and wind speed, substitute the learning value of CNN into the observation value of dust pollution that has occurred in a certain place, and obtain a set of weight values and partial weight values, and then based on this After the parameters and real-time monitoring data, that is, the fourth judgment result in the calculation processing unit 3 are substituted into the model, the subsequent pollution transmission path and pollution source can be predicted when the conditions are met.

When comparing the results to determine where is the source of the dust, please refer to Figure 2C, which is a partial method flow chart of an embodiment of the present invention, after step S17, further includes step S19: Step S19: the calculation processing unit generates a warning message to the warning unit according to the comparison result.

As described in step S19, the calculation processing unit 3 will generate warning information to the warning unit 5 according to the comparison result. In this embodiment, the warning unit 5 can further push the warning information to the mobile device of the staff responsible for taking measures, Or the warning unit 5 is an electronic device of a construction monitoring unit, which notifies the construction personnel in the form of lights or texts, where they need to be covered with cloth or other preventive measures, but not limited thereto, wherein, in this embodiment , the warning information includes dust pollution intensity, dust pollution range, pollution source and pollution trajectory, but not limited thereto.

To sum up, the intelligent dust monitoring and management system and method of the present invention uses the image capture device to capture the image of the construction area, and then performs image processing and calculation through the calculation processing unit, and the concentration value of suspended particles obtained by the sensing device. Comparing, and then generating judgment results, to effectively solve the conventional knowledge that it is only possible to judge the situation of suspended particles, but it is difficult to know exactly where the dust is, and it must rely on a large amount of manpower to check, resulting in the lack of low efficiency in improving dust, and the aforementioned The method and its system can capture images at a fixed time for judgment in an automatic and highly accurate manner, so the purpose of the present invention can be achieved indeed.

But the above-mentioned ones are only preferred embodiments of the present invention, and should not limit the implementation scope of the present invention, that is, any simple equivalent changes and modifications made according to the scope of the patent application for the present invention and the contents of the description of the invention, All still belong to the scope covered by the patent of the present invention.

1: Image capture device

2: Sensing device

3: Operation processing unit

4: sprinkler unit

5: Warning unit

Claims (10)

A smart dust monitoring and management system, comprising: an image capture device, set in an area, to capture a first image of a first position and a second position in the area at a time; a sensing device, set in The area senses a suspended particle concentration value in the area at the time; and an arithmetic processing unit is respectively connected to the image capture device and the sensing device for signal conversion, and converts the first image into a first image pixel value matrix, and the first image pixel value matrix includes a first pixel value intensity corresponding to the first position and a second pixel value intensity corresponding to the second position, and receives the suspended particle concentration value; wherein, the operation process The unit subtracts a reference image pixel value matrix from the first image pixel value matrix to generate a second image pixel value matrix. When the second image pixel value matrix corresponds to the first pixel value intensity corresponding to the first position and the When the intensity of the second pixel value at the second position is less than zero, a first judgment result is generated, when the intensity of the first pixel value of the second image pixel value matrix corresponding to the first position or the intensity of the second position When the second pixel value intensity is greater than zero, a preset threshold value is subtracted from the second image pixel value matrix to generate a third image pixel value matrix. When the third image pixel value matrix corresponds to the first position When the intensity of a third pixel value and the intensity of a fourth pixel value at the second position are both less than zero, a second judgment result is generated, when the third pixel of the third image pixel value matrix corresponding to the first position value intensity or the fourth pixel value intensity of the second position is greater than zero, and the aerosol concentration value is less than a aerosol concentration threshold value, then a third judgment result is generated, when the corresponding pixel value matrix of the third image When the intensity of the third pixel value at the first position or the intensity of the fourth pixel value at the second position is greater than zero, and the aerosol concentration value is greater than the aerosol concentration threshold value, a fourth judgment result is generated; Wherein, the first judgment result, the second judgment result and the third judgment result are that there is no dust in the first position and the second position; wherein, the fourth judgment result includes: when the third image pixel value matrix The intensity of the third pixel value corresponding to the first position is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is less than zero, and the suspended particle concentration value is greater than the suspended particle concentration threshold value, then it is determined that the There is dust at the first location, and the suspended particle concentration value has reached the upper limit, and there is no dust at the second location; when the intensity of the third pixel value corresponding to the first location of the third image pixel value matrix is less than zero, and When the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particle concentration value is greater than the suspended particle concentration threshold value, it is judged that there is no dust at the first position, and there is dust at the second position, and the The suspended particle concentration value has reached the upper limit; and when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero , and the suspended particle concentration value is greater than the suspended particle concentration threshold value, it is determined that there is dust at the first location and the second location, and the suspended particle concentration value has reached the upper limit. According to the smart dust monitoring and management system described in claim 1, the suspended particle concentration value is the concentration value of fine suspended particles (PM _2.5 ) and suspended particles (PM ₁₀ ) in the air. According to the smart dust monitoring and management system described in Claim 1, wherein, the operation processing unit uses a Sobel edge detection algorithm and the fourth judgment result to perform calculations to obtain a gradient map, and integrates according to the gradient map to generate a Dust pollution range and one of the dust pollution intensity corresponding to the dust pollution range. According to the smart dust monitoring and management system described in Claim 1, it further includes a sprinkler unit, which is connected with the signal of the operation processing unit, and the operation processing unit generates A sprinkling signal and a sprinkling position corresponding to the dust are sent to the sprinkling unit, and the sprinkling unit performs sprinkling according to the sprinkling signal and the sprinkling position. According to the smart dust monitoring and management system described in Claim 1, wherein, the computing processing unit further obtains wind direction information and a wind speed value sensed by the sensing device in the area at the time, and uses a convolutional neural network model Substituting the fourth judgment result, the wind direction information and the wind speed value to perform a comparison operation to generate a comparison result, the comparison result includes the dust generated at the first location, and correspondingly affects the first location, or the second location. Dust generated at one location, and the dust at the second location comes from the first location, or the dust generated at the second location, and the dust at the first location comes from the second location, or the dust generated at the second location dust, and correspondingly affects the second location. A smart dust monitoring and management method, the steps include: an image capture device captures a first image of a first position and a second position in an area at a time, and transmits it to an operation processing unit; the operation processing unit converting the first image into a first image pixel value matrix, and subtracting a reference image pixel value matrix according to the first image pixel value matrix to generate a second image pixel value matrix; when the second image pixel value matrix When the intensity of the first pixel value corresponding to the first position and the intensity of the second pixel value of the second position are both less than zero, a first judgment result is generated, or when the pixel value in the second image matrix corresponds to When the intensity of the first pixel value at the first position or the intensity of the second pixel value at the second position is greater than zero, the arithmetic processing unit subtracts a preset threshold value from the second image pixel value matrix to generate a third Image pixel value matrix; when the intensity of a third pixel value corresponding to the first position and the intensity of a fourth pixel value in the second position in the third image pixel value matrix are both less than zero, a second judgment is generated As a result, or when the third pixel value intensity corresponding to the first position in the third image pixel value matrix or the first When the intensity of the fourth pixel value at the second position is greater than zero, the arithmetic processing unit obtains a suspended particle concentration value at the time from a sensing device; and when the suspended particle concentration value is less than a suspended particle concentration threshold value, generates A third judgment result, or when the suspended particle concentration value is greater than the suspended particle concentration threshold value, a fourth judgment result is generated; wherein, the first judgment result, the second judgment result and the third judgment result are the There is no dust in the first position and the second position; wherein, the fourth judgment result includes: when the intensity of the third pixel value of the third image pixel value matrix corresponding to the first position is greater than zero, and corresponding to the third pixel value matrix The intensity of the fourth pixel value at the second position is less than zero, and the concentration value of the suspended particles is greater than the threshold value of the concentration of suspended particles, then it is judged that there is dust in the first position, and the concentration value of the suspended particles has reached the upper limit, and the second position No dust; when the intensity of the third pixel value corresponding to the first position of the third image pixel value matrix is less than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particles If the concentration value is greater than the threshold value of the suspended particle concentration, it is judged that there is no dust at the first location, and there is dust at the second location, and the concentration of the suspended particle has reached the upper limit; and when the pixel value matrix of the third image corresponds to the When the intensity of the third pixel value at the first position is greater than zero, and the intensity of the fourth pixel value corresponding to the second position is greater than zero, and the suspended particle concentration value is greater than the suspended particle concentration threshold value, then it is judged that the first There is dust at the location and the second location, and the concentration of suspended particulates has reached the upper limit. According to the smart dust monitoring and management method described in Claim 6, wherein the suspended particle concentration value is the concentration value of fine suspended particles (PM _2.5 ) and suspended particles (PM ₁₀ ) in the air. According to the smart dust monitoring and management method described in claim 6, it further includes steps: The calculation processing unit performs calculation with a Sobel edge detection algorithm and the fourth judgment result to obtain a gradient map; and the calculation processing unit performs integration according to the gradient map to generate a dust pollution range and a corresponding to the dust pollution range 1. Dust pollution intensity. According to the smart dust monitoring and management method described in Claim 6, wherein, when the suspended particle concentration value is greater than the suspended particle concentration threshold value, after the step of generating a fourth judgment result, a step is further included: the operation processing unit according to the The judgment result generates a watering signal and a watering unit corresponding to a watering position with dust; and the watering unit performs watering according to the watering signal and the watering position. According to the intelligent dust monitoring and management method described in claim 6, it further includes the steps: the sensing device senses wind direction information and a wind speed value in the area at the time, and sends them to the computing processing unit; and the computing processing unit Substituting a convolutional neural network model into the fourth judgment result, the wind direction information, and the wind speed value to perform a comparison operation to generate a comparison result; wherein, the comparison result includes the dust generated at the first location, and corresponds to Affect the first location, or the dust generated by the first location, and the dust at the second location comes from the first location, or the dust generated at the second location, and the dust at the first location comes from the second location location, or the dust generated by the second location, and correspondingly affects the second location.

Images