A belt conveyor running state intelligent video detection method

Abstract

The application relates to a kind of belt conveyor operating state intelligent video detection method, belong to artificial intelligence field.The application uses time series picture and mean superposition figure, based on information entropy principle measurement the uncertain degree between picture pixel value, input convolution main network extracts feature, and finally identifies the operating state of belt conveyor, improves the precision degree of belt conveyor operating state detection.Input picture is composed of three channel data, a large number of improves convolution network layer parameter, uses DropBlock method to mask pixel area value in the training process, prevents model overfitting.The application directly outputs the operating state of belt conveyor, without introducing other image processing methods, has faster recognition speed.

Description

Technical Field

[0001] The invention belongs to the field of artificial intelligence and relates to an intelligent video detection method for the operating status of a belt conveyor. Background Art

[0002] Currently, belt conveyors are widely used for automated long-distance coal transportation. Belt conveyors are essential bulk material transport equipment due to their low operating costs, ease of operation, high efficiency, and long-distance capabilities. Monitoring the operating status of belt conveyors is crucial, especially for idle mines, to identify potential violations.

[0003] However, current belt conveyor operation monitoring relies primarily on manual inspections, sensor detection, and video detection. Manual inspections are labor-intensive and time-consuming, resulting in low detection efficiency, which can easily lead to oversights and safety issues, and hinders belt conveyor efficiency. Sensors are also fragile, and the large number of sensors makes them difficult to maintain. Current video-based detection methods primarily identify faults such as belt deviation and foreign objects, but fail to identify operational status.

[0004] Currently, there are few technologies for detecting the operating status of conveyors, and none of them describe in detail the specific methods for determining the operation of the belt. Summary of the Invention

[0005] In view of this, an object of the present invention is to provide an intelligent video detection method for the operating status of a belt conveyor.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent video detection method for the running status of a belt conveyor, the method comprising:

[0008] Collect the image of the belt conveyor running or stopping as input data, and convert the input image a and images b Convert to a single-channel grayscale image using the geometric mean method:

[0009] (1)

[0010] Indicates the color channels, Indicates the total number of channels used for calculation, 、 、 Represent the three color channels of the real image respectively;

[0011] Input image a and images bWhen the images are at the same moment, the output value in the Decision Network is 1, indicating that the images are the same; otherwise, it is 0, indicating that the images are different.

[0012] Adding an image to the input a and images b The mean overlay of the grayscale image and the mean of the corresponding elements are calculated:

[0013] (2)

[0014] 、 They represent the first Liedi The element value of the row;

[0015] Based on the uncertainty principle of information entropy measurement, the mean value is superimposed on the graph. c As input:

[0016] (3)

[0017] (4)

[0018] is an image a and images b The difference between Calculated by formula (4), the larger the value, the better the image a and images b In the Liedi The greater the difference in the element values ​​of the rows; M Indicates the total number of columns of the image pixel matrix; N Indicates the total number of rows of the image pixel matrix;

[0019] When the image a and images b When they are completely equal, a=b=c, =0, input image information entropy =0, the smaller the information entropy, the image a and b The less confusion there is, a and b The higher the similarity, the greater the information entropy, indicating that the image a and b The greater the confusion, a and b The lower the similarity;

[0020] image a ,image b and mean overlay plot cSynthesize a three-channel image and input it into the network convolution layer to extract color, texture, shape, and image topology features. The input is a three-channel image composed of three grayscale images. Each convolution layer has three feature maps. Each neuron in the convolution layer is connected to multiple neurons in the previous layer in a similar area. Each neuron in the convolution layer is connected to multiple neurons in the previous layer in a similar area.

[0021] (5)

[0022] in It is l +1 convolution layer output, i.e. feature map; Indicates the l The output of the convolution layer, is the Kronecker product of tensor operations, It is l +1 layer network parameter value, It is the bias term of the layer parameters. The three-channel superposition of the image will increase the complexity of the model. To prevent overfitting of the model, during the convolution process, the convolution layer is subjected to Drop Block regularization, which sets the pixel value of a region block composed of adjacent region units in the feature map to 0.

[0023] (6)

[0024] The probability value controls the amount of feature masking, feat_size and block_size represent the feature map and area block size, keep_prob is the retention ratio, and the Drop Block operation is implemented in each convolutional layer through formula (6);

[0025] Result after convolution operation Z Output to the fully connected layer D :

[0026] (7)

[0027] express D Middle i The output of a neuron; Represents the weight parameter, connecting the j Input features to i neurons; Represents the input feature value, the output feature map from the convolutional layer j elements; Indicates the i The bias term of the neuron parameter; the activation function output result after Dropout in the Decision Network And measure the model effect through the loss function:

[0028] (8)

[0029] Y is the true value of the sample, S is the number of samples, square loss function combined L 2 Regularization: evaluate the model effect during the training phase and control the training time.

[0030] The beneficial effects of the present invention are:

[0031] (1) Using time series images and mean overlay images, the uncertainty between image pixel values ​​is measured based on the principle of information entropy, and the features are input into the convolutional backbone network to extract features, and finally the operating status of the belt conveyor is identified, thereby improving the accuracy of the belt conveyor operating status detection.

[0032] (2) The input image consists of three-channel data, which greatly improves the parameters of the convolutional network layer. The DropBlock method is used to mask the pixel area values ​​during training to prevent the model from overfitting.

[0033] (3) This method directly outputs the operating status of the belt conveyor without introducing other image processing methods and has a faster recognition speed.

[0034] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0036] Figure 1 This is a schematic diagram of the principle of the present invention. DETAILED DESCRIPTION

[0037] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0038] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0039] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0040] See also Figure 1 The operating status of a belt conveyor is usually divided into running and stopped. Combined with the coal recognition algorithm, it can be divided into running with coal, running without coal, stopped with coal, and stopped without coal. The main method is to calculate the similarity of belt conveyor images in time series, identify whether the image has changed, and judge the belt operating status.

[0041] Collect the running or stopping images of the belt conveyor as input data, and convert the input images into a and b Convert to a single-channel grayscale image using the geometric mean method:

[0042] (1)

[0043] Indicates the color channels, Indicates the total number of channels used for calculation, 、 、 Represent the three color channels of the real image respectively;

[0044] Input image a and images b When the images are at the same moment, the output value in the Decision Network is 1, indicating that the images are the same; otherwise, it is 0, indicating that the images are different.

[0045] Adding an image to the input a and images bThe mean overlay of the grayscale image and the mean of the corresponding elements are calculated:

[0046] (2)

[0047] 、 They represent the first Liedi The element value of the row;

[0048] The introduction of the mean superposition map c as input is based on the principle of information entropy measurement uncertainty:

[0049] (3)

[0050] (4)

[0051] When the picture a and b When they are completely equal, a=b=c, then =0, input image information entropy =0. The smaller the information entropy, the better the image a and b The smaller the confusion between a and b The more similar, the more vice versa; M Indicates the total number of columns of the image pixel matrix; N Indicates the total number of rows in the image pixel matrix.

[0052] picture a 、 b 、 c To synthesize a three-channel image, without introducing a twin network, the input is fed into the network convolutional layer to extract color, texture, shape, and image topology features. The input is a three-channel image composed of three grayscale images. Each convolutional layer has three feature maps. Each neuron in a convolutional layer is connected to multiple neurons in a similar area in the previous layer.

[0053] (5)

[0054] in It is l +1 convolution layer output, i.e. feature map; Indicates the l The output of the convolution layer, is the Kronecker product of tensor operations, It is l +1 layer network parameter value, It is the bias term of the layer parameters. The three-channel superposition of the image will increase the complexity of the model. To prevent overfitting of the model, during the convolution process, the convolution layer is subjected to Drop Block regularization, and the pixel values ​​of a region block composed of adjacent region units in the feature map are set to 0.

[0055] (6)

[0056] The probability value controls the amount of feature masking, feat_size and block_size represent the feature map and area block size, keep_prob is the retention ratio, and the DropBlock operation is implemented in each convolutional layer through formula (6).

[0057] Result after convolution operation Z Output to the fully connected layer D:

[0058] (7)

[0059] express D Middle i The output of a neuron, Represents the weight parameter, connecting the j Input features to i neurons; Represents the input feature value, the output feature map from the convolutional layer j elements; Indicates the i The bias term of the neuron parameter; the activation function output result after Dropout in the Decision Network And measure the model effect through the loss function:

[0060] (8)

[0061] Square loss function combined L 2 Regularization: evaluate the model effect during the training phase and control the training time.

[0062] Combining coal recognition algorithms with belt conveyor monitoring images, the model determines whether the conveyor is operating or transporting coal. This provides supervisors with an efficient detection method for coal mine safety.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent video detection method for the running status of a belt conveyor, characterized by: The method is: Collect the image of the belt conveyor running or stopping as input data, and convert the input image a and images b Convert to a single-channel grayscale image using the geometric mean method: （1） Indicates the color channels, Indicates the total number of channels used for calculation, 、 、 Represent the three color channels of the real image respectively; Input image a and images b When the images are at the same moment, the output value in the Decision Network is 1, indicating that the images are the same; otherwise, it is 0, indicating that the images are different. Adding an image to the input a and images b The mean overlay of the grayscale image and the mean of the corresponding elements are calculated: （2） 、 They represent the first Liedi The element value of the row; Based on the uncertainty principle of information entropy measurement, the mean value is superimposed on the graph. c As input: （3） （4） is an image a and images b The difference between Calculated by formula (4), the larger the value, the better the image a and images b In the Liedi The greater the difference in the element values ​​of the rows; M Indicates the total number of columns of the image pixel matrix; N Indicates the total number of rows of the image pixel matrix; When the image a and images b When they are completely equal, a=b=c, =0, input image information entropy =0, the smaller the information entropy, the image a and b The less confusion there is, a and b The higher the similarity; The larger the information entropy, the better the image a and b The greater the confusion, a and b The lower the similarity; image a ,image b and mean overlay plot c Synthesize a three-channel image and input it into the network convolution layer to extract color, texture, shape, and image topology features. The input is a three-channel image composed of three grayscale images. Each convolution layer has three feature maps. Each neuron in the convolution layer is connected to multiple neurons in the previous layer in a similar area. Each neuron in the convolution layer is connected to multiple neurons in the previous layer in a similar area. （5） in It is l +1 convolution layer output, i.e. feature map; Indicates the l The output of the convolution layer; is the Kronecker product of tensor operations, It is l +1 layer network parameter value, It is the bias term of the layer parameters. The three-channel superposition of the image will increase the complexity of the model. To prevent overfitting of the model, the convolution layer is subjected to DropBlock regularization during the convolution process. The pixel values ​​of a region block composed of adjacent region units in the feature map are set to 0. （6） The probability value controls the amount of feature masking, feat_size and block_size represent the feature map and area block size, keep_prob is the retention ratio, and the Drop Block operation is implemented in each convolutional layer through formula (6); Result after convolution operation Z Output to the fully connected layer D : （7） express D Middle i The output of a neuron; Represents the weight parameter, connecting the j Input features to i neurons; Represents the input feature value, the output feature map from the convolutional layer j elements; Indicates the i The bias term of the neuron parameters; Activation function output results after Dropout in Decision Network And measure the model effect through the loss function: （8） is the true value of the sample, is the number of samples, square loss function combined L 2 Regularization: evaluate the model effect during the training phase and control the training time.

Images