A belt empty detection method based on image processing

Abstract

This invention relates to an image processing-based method for detecting belt idling, comprising the following steps: Step 1: acquiring a belt surface image using the SDK file of a camera device; Step 2: performing image enhancement, grayscale processing, and filtering / denoising on the acquired belt image; Step 3: determining the presence or absence of an external light source, and using variance and expectation determination to determine the idling status of the image; Step 4: selecting features of the belt image based on its idling and material-containing states, wherein the material-containing state can be categorized into blocky and fine powder states based on the characteristics of the ore, and the idling state can be categorized into black-and-white features, gradient features, and no obvious features based on the belt surface characteristics; Step 5: after selecting the belt image features, determining idling status according to the feature selection rules, specifying a flag value as the number of consecutive idling states in the belt image; utilizing image processing of the video for variance and expectation determination effectively completes the detection of belt idling status, saving energy and reducing consumption.

Description

Technical Field

[0001] This invention relates to the field of belt inspection technology, and in particular to a belt unloaded detection method based on image processing. Background Technology

[0002] Belt conveyors, as a common tool for transporting minerals, have advantages such as long transport distance, large transport capacity, and long continuous operation time, and are widely used in modern industrial technology. When the belt is in an unloaded state, it may run off-center due to uneven stretching of the inner and outer sides, uneven tension, and unstable center of gravity, leading to a series of adverse consequences. Prolonged unloaded operation also generates unnecessary belt wear and unnecessary energy consumption. When the belt is loaded, restarting from a stopped state can easily cause breakage at the joint, or may overload or prevent the motor from starting, or even burn out the motor, or cause coal blockage. Therefore, it is necessary to design a system to detect belt unload, thereby reducing losses, improving belt conveying efficiency, and reducing the company's labor costs.

[0003] Chinese patent CN110519566B discloses a method for monitoring the motion state of a belt based on video processing. The method includes: firstly, preprocessing the video, then detecting moving objects; then setting a flag based on the area of ​​the moving region in the video to determine whether the belt in the current video frame is dynamic or static; defining a state machine to realize the dynamic transformation of four motion states of the belt; and then judging the state of the current video frame based on the flag and the accumulated value of the flag, thereby making a judgment on the motion state of the belt and providing a real-time alarm. However, this patent may affect the monitoring results and status in complex working conditions or environments where the camera is obstructed.

[0004] Chinese patent CN102951428A discloses an energy-saving system for belt conveyors and its energy-saving control method, including a system start button, a belt conveyor, a material detection device, a signal transmitter, a belt conveyor drive motor, a frequency converter, and a programmable controller; it effectively reduces the belt conveyor's idle time and idle energy consumption, thereby reducing the system's mechanical losses during these events and saving energy; however, this patent controls energy saving by detecting the belt conveyor's idle time through hardware connection and linkage, and the tension between belt conveyors and other effects result in poor actual detection of belt conveyor idle time. Summary of the Invention

[0005] This invention provides a belt unload detection method based on image processing. By using image processing of video to perform variance determination and expectation determination rules, the method effectively completes the detection of belt unload status, thereby improving belt transportation efficiency, reducing industrial energy consumption, and reducing enterprise costs.

[0006] To achieve the above object, the present invention is implemented by the following technical solutions:

[0007] A belt no-load detection method based on image processing, comprising the following steps:

[0008] Step 1: Install a camera device, collect the belt image through the camera device, and obtain the belt surface image by calling the SDK file of the camera device;

[0009] Step 2: Preprocess the collected belt image, perform image enhancement, grayscale processing, and filtering to remove noise;

[0010] Step 3: Determine whether there is a light source in the outside world. It is stipulated that a = 1 represents the no-load state of the belt, and a = 0 represents the state of the belt with materials. And use two methods of variance determination and expectation determination to determine the no-load state of the picture;

[0011] Step 4: Select the characteristics of the belt image according to the no-load state and the state with materials of the belt. The state with materials can be divided into blocky and fine powder according to the ore characteristics. The no-load state can be divided into black and white characteristics, gradient characteristics, and no obvious characteristics according to the belt surface characteristics;

[0012] Step 5: After the belt image characteristics are selected, the system selects appropriate determination rules according to its characteristics for no-load determination. It is stipulated that the value of flag is the number of consecutive no-load states of the belt image.

[0013] Further, when there is a light source in Step 3, Rule 1 determination criteria are used for determination. The variance determination method of Rule 1 is as follows:

[0014] When it is judged to be daytime, the variance threshold D0 selects the daytime variance threshold d1;

[0015] When it is judged to be night, the variance threshold D0 selects the night variance threshold d2;

[0016] Compare the variance value D1 of the current belt image with the variance threshold D0. When D1 < D0, judge that the current belt image is in the no-load state, that is, a = 1; when D1 > D0, judge that the current belt image is in the state with materials, that is, a = 0;

[0017] Expectation determination method:

[0018] When it is judged to be daytime, the expectation difference threshold E0 selects the daytime expectation threshold e1;

[0019] When it is judged to be night, the expectation difference threshold E0 selects the night expectation threshold e2;

[0020] If the first two characteristic situations are met, judge that the current belt image is in the no-load state; if the first two characteristic situations are not met, then perform expectation difference determination.

[0021] Further, when there is no light source in Step 3, the determination is made according to the determination criterion of Rule 2. The variance determination method of Rule 2 is as follows:

[0022] Let the variance threshold D0 = d0;

[0023] Compare the variance value D1 of the current belt image with the variance threshold D0. When D1 < D0, it is determined that the current belt image is in the no-load state, that is, a = 1; when D1 > D0, it is determined that the current belt image is in the with-material state, that is, a = 0;

[0024] Expected determination method:

[0025] Let the expected threshold E0 = e0;

[0026] Judge whether the belt image conforms to the black-and-white feature or the gradient feature. If it conforms to the first two feature cases, it is determined that the current belt image is in the no-load state; if it does not conform to the first two feature cases, the expected difference judgment is performed.

[0027] Further, in Step 4, when there is interference from an external light source in the belt image, draw a feature line and make a mark, and complete the no-load determination through variance determination and expected determination of the made feature line.

[0028] Further, in Step 5, it is stipulated that the value of flag is the number of consecutive no-load belt images. Set the initial value of flag to 0, that is, when the system starts running, the number of no-load belt images is 0. When each consecutive 1 image is in the no-load state, that is, a = 1, let flag = flag + 1. When the system continuously determines that more than or equal to 30 images are in the no-load state, at this time the flag value is greater than or equal to 30, the system will give a no-load alarm. If the image is determined to be in the with-material state, that is, a = 0, during the determination process, the flag value will be reset to 0; after the no-load determination of the image, the process can be ended; if not selected, the system will continue to collect the current image and perform the no-load determination.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] 1) By collecting the belt image through the camera device and performing image processing, the feature extraction of the belt is realized, so as to realize the no-load detection of the belt;

[0031] 2) The accuracy and efficiency of the no-load detection of the belt are improved through the variance determination and the expected determination method;

[0032] 3) Further improve the belt transportation efficiency, reduce industrial energy consumption, and reduce the cost of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is the flow schematic diagram of the present invention.

[0034] Figure 2 This is a flowchart of the belt unload detection rules under external light source interference as described in this invention.

[0035] Figure 3 This is a flowchart of the belt unload detection rules under the condition of no external light source interference as described in this invention.

[0036] Figure 4 This is a schematic diagram illustrating the selection of features for determining whether a belt image is empty or loaded, as described in this invention.

[0037] Figure 5 It is a belt image with black and white features, gradient features, and material state as described in this invention. Detailed Implementation

[0038] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0039] See Figure 1 This is a flowchart illustrating the present invention. The present invention relates to an image processing-based method for detecting belt unloaded traffic, comprising the following steps:

[0040] Step 1: Install a camera device to capture images of the conveyor belt. Obtain images of the conveyor belt surface by calling the camera device's SDK file. Capture conveyor belt video through the camera device at the work site, obtain the camera's SDK package, and select conveyor belt images and videos of different times, different workshops, and different states for cropping. To reduce the computational load of image processing, select to crop one image every 3 seconds. At the same time, perform image classification, consider the brightness of different times and whether the conveyor belt is indoors or outdoors, and further analyze the characteristics of the conveyor belt in the unloaded and loaded states.

[0041] Step 2: Preprocess the acquired belt images by performing image enhancement, grayscale conversion, and filtering for noise reduction;

[0042] Step 3: Based on the light source issue, the belt unload determination rules are divided into two types: one is the belt unload determination rule under the condition of external light source interference, and the other is the belt unload determination rule under the condition of no external light source interference. It is stipulated that a=1 is the belt unloaded state, and a=0 is the belt loaded state. The determination methods are divided into two types: variance determination method and expectation determination method. If either determination method determines that the belt is unloaded, then the belt image is determined to be in an unloaded state.

[0043] Step 4: Select features of the belt image based on the belt's unloaded and loaded states. The loaded state can be divided into blocky and fine powder based on the characteristics of the ore. The unloaded state can be divided into black and white features, gradient features, and no obvious features based on the belt surface characteristics.

[0044] Step 5: After the belt image features are selected, the system selects appropriate determination rules according to its features for no-load determination, and stipulates that the value of flag is the number of consecutive no-load states of the belt image.

[0045] See Figure 2 、 Figure 4 and Figure 5 , further, when there is a light source in Step 3, Rule 1 determination criteria are used for determination. Rule 1 is the no-load determination of the belt under the interference of external light sources. The specific content is as follows:

[0046] First, judge the time of the image, and judge according to the gray values of the pixel points on the first feature line. Select to draw the feature value at the place where the light changes, such as at L0 where the sunlight changes or at L0' where the light changes. Judge the current time according to the gray values of the pixel points on the feature lines of L0 and L0'. Let the two gray values be R1 and R2 respectively. First, according to data statistics and analysis, set appropriate thresholds R0 and R0' for sunlight and light to distinguish day and night. Since the smaller the gray value, the darker the color, if the first feature line is drawn at L0, the threshold is R0, and it is stipulated that R1 > R0 is daytime and R1 < R0 is night; if the first feature line is drawn at L0', the threshold is R0', and it is stipulated that R2 < R0' is daytime and R2 > R0' is night;

[0047] Variance determination method:

[0048] When it is judged as daytime, the variance threshold D0 selects the daytime variance threshold d1;

[0049] When it is judged as night, the variance threshold D0 selects the night variance threshold d2;

[0050] Let the variance value D1 of the current belt image be compared with the variance threshold D0. When D1 < D0, as shown in A0, it is judged that the current belt image is in the no-load state, that is, a = 1; when D1 > D0, as shown in A3, it is judged that the current belt image is in the state with material, that is, a = 0;

[0051] Expected value determination method:

[0052] When it is judged as daytime, the expected difference threshold E0 selects the daytime expected threshold e1;

[0053] When it is judged as night, the expected difference threshold E0 selects the night expected threshold e2;

[0054] First, determine whether the belt image conforms to the black-and-white feature shown in A1 or the gradient feature shown in A2. If it conforms to the first two feature cases, determine that the current belt image is in the no-load state; if it does not conform to the first two feature cases, then compare the expected difference E2 - E1 of the current belt image with the expected difference threshold E0. When E2 - E1 < E0, as shown in A0, determine that the current belt image is in the no-load state, that is, a = 1; when E2 - E1 > E0, as shown in Figure 5 A3 of

[0055] See Figure 3 , Figure 4 and Figure 5 , further, when there is no light source in step three, the determination is made according to the determination criterion of rule 2. The variance determination method of rule 2 is as follows:

[0056] Variance determination method:

[0057] Set the variance threshold D0 = d0;

[0058] Compare the variance value D1 of the current belt image with the variance threshold D0. When D1 < D0, determine that the current belt image is in the no-load state, that is, a = 1; when D1 > D0, as shown in Figure 5 A3 of

[0059] Expected determination method:

[0060] Set the expected threshold E0 = e0;

[0061] First, determine whether the belt image conforms to the black-and-white feature shown in A1 or the gradient feature shown in A2. If it conforms to the first two feature cases, determine that the current belt image is in the no-load state; if it does not conform to the first two feature cases, then compare the expected difference E2 - E1 of the current belt image with the expected difference threshold E0. When E2 - E1 < E0, as shown in A0, determine that the current belt image is in the no-load state, that is, a = 1; when E2 - E1 > E0, as shown in A3, determine that the current belt image is in the loaded state, that is, a = 0.

[0062] See Figure 4 , further, when there is interference from an external light source in step four, the first feature line is selected to be drawn at window 1, marked as L0 or at light 3, marked as L0'. Feature regions are selected on belt A0, marked as L1 to L35; according to the different features and states of the belt image, L1 to L35 have different functions, and the specific functions are as follows:

[0063] Variance determination situation:

[0064] During on-site operations, when the conveyor belt is loaded with material, the ore can exist in two states: either in lumps or in fine powder. When the ore is in lumps, the variance of the grayscale values ​​of pixels on the conveyor belt with ore will be much larger than the variance of the grayscale values ​​of pixels when the conveyor belt is unloaded. In this case, the mean variance of five sets of feature lines (15 feature lines) L12~L14, L17~L19, L22~L24, L27~L29, and L32~L34 is compared with the threshold D0 to complete the unload determination.

[0065] Expected judgment situation:

[0066] When the ore is in the form of fine powder, the variance of the grayscale value of the pixel where there is ore on the belt will be smaller than the variance of the grayscale value of the pixel when the belt is unloaded. However, the difference between the two is small. At this time, the expected mean of five sets of feature lines (L1, L6, L2, L7, L3, L8, L4, L9, L5, L10) on the edge of the belt is selected and the expected mean of five sets of feature lines (L12~L14, L17~L19, L22~L24, L27~L29, L32~L34) (15 feature lines) are used to perform difference calculation. The difference obtained is compared with the threshold E0 to complete the unload determination.

[0067] When the conveyor belt is interfered with by the external light source L0 at window 1, i.e. sunlight, in the material-carrying state, shadow areas will be formed on both sides of the material on the belt, while a reflective surface will be formed at the edge of the belt. At this time, the expected mean of the gray values ​​of each pixel on the five sets of feature lines (L1, L6, L2, L7, L3, L8, L4, L9, L5, L10) at the edge of the belt is compared with the expected mean of the gray values ​​of each pixel on the five sets of feature lines (L11, L15, L16, L20, L21, L25, L26, L30, L31, L35) on both sides. The difference obtained is compared with the threshold E0 to complete the no-load determination.

[0068] When the belt conveyor image is black and white, the variance of the grayscale values ​​of each pixel on five feature lines (L12-L14, L17-L19, L22-L24, L27-L29, and L32-L34) is selected, and the maximum and minimum values ​​of the 15 variance data are compared. When the minimum variance value is less than 0.7 times the maximum variance value, the belt conveyor image is considered to be black and white, thus completing the no-load determination.

[0069] When the belt conveyor image under no-load conditions shows a gradual change, the variance of the grayscale values ​​of each pixel on five feature lines (15 feature lines) L12~L14, L17~L19, L22~L24, L27~L29, and L32~L34 is selected. It is then determined whether the minimum value of these five sets of variance data is the variance of the grayscale values ​​of each pixel on the middle feature line (i.e., the variance of L13, L18, L23, L28, and L33). If the middle value is the minimum, the belt conveyor image is considered to show a gradual change, thus completing the no-load determination.

[0070] Furthermore, step five specifies that the value of flag is the number of consecutive belt images in an unloaded state. The initial value of flag is set to 0, meaning that there are 0 belt images in an unloaded state when the system starts running. For each consecutive image in an unloaded state (a=1), flag is set to flag+1. When the system continuously determines that 30 or more images are in an unloaded state, the flag value is greater than or equal to 30, and the system will issue an unloaded alarm. If the image is determined to be loaded (a=0) during the determination process, the flag value is reset to 0. After the image is determined to be unloaded, the process can be terminated. If not, the system will continue to collect the current image and perform unload determination.

[0071] The above embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the above embodiments. Unless otherwise specified, the methods used in the above embodiments are conventional methods.

Claims

1. A belt unload detection method based on image processing, characterized in that, It includes the following steps: Step 1: Install a camera device, collect belt images through the camera device, and obtain the belt surface image by calling the SDK file of the camera device; Step 2: Preprocess the collected belt images, perform image enhancement, grayscale processing, and filtering to remove noise; Step 3: Determine whether there is a light source in the outside world; Step 4: Select the features of the belt image according to the no-load state and the loaded state of the belt. The loaded state is divided into lump and fine powder according to the ore characteristics, and the no-load state is divided into black-and-white characteristics, gradient characteristics, and no obvious characteristics according to the belt surface characteristics; Step 5: After the belt image features are selected, the system makes a no-load determination according to the determination rules. It is stipulated that the value of flag is the number of consecutive no-load states of the belt image. It is stipulated that a = 1 represents the no-load state of the belt, and a = 0 represents the loaded state of the belt, and two methods of variance determination and expectation determination are used to determine the no-load state of the picture; Specifically include: (1) When there is a light source, make a determination according to the determination standard of Rule 1. The variance determination method of Rule 1 is: When it is judged to be daytime, the variance threshold D0 selects the daytime variance threshold d1; When it is judged to be night, the variance threshold D0 selects the night variance threshold d2; Compare the variance value D1 of the current belt image with the variance threshold D0. When D1 < D0, judge that the current belt image is in the no-load state, that is, a = 1; when D1 > D0, judge that the current belt image is in the loaded state, that is, a = 0; Expectation determination method: When it is judged to be daytime, the expectation difference threshold E0 selects the daytime expectation threshold e1; When it is judged to be night, the expectation difference threshold E0 selects the night expectation threshold e2; First, judge whether the belt image conforms to the black-and-white characteristics or the gradient characteristics. If it conforms to the black-and-white characteristics or the gradient characteristics, judge that the current belt image is in the no-load state; if it does not conform to the black-and-white characteristics or the gradient characteristics, then compare the expectation difference E2 - E1 of the current belt image with the expectation difference threshold E0. When E2 - E1 < E0, judge that the current belt image is in the no-load state, that is, a = 1; when E2 - E1 > E0, judge that the current belt image is in the loaded state, that is, a = 0; (2) When there is no light source, make a determination according to the determination standard of Rule 2. The variance determination method of Rule 2 is: Set the variance threshold D0 = d0; Compare the variance value D1 of the current belt image with the variance threshold D0. When D1 < D0, judge that the current belt image is in the no-load state, that is, a = 1; when D1 > D0, judge that the current belt image is in the loaded state, that is, a = 0; Expectation determination method: Set the expectation threshold E0 = e0; First, judge whether the belt image conforms to the black-and-white characteristics or the gradient characteristics. If it conforms to the black-and-white characteristics or the gradient characteristics, judge that the current belt image is in the no-load state; if it does not conform to the black-and-white characteristics or the gradient characteristics, then compare the expectation difference E2 - E1 of the current belt image with the expectation difference threshold E0. When E2 - E1 < E0, judge that the current belt image is in the no-load state, that is, a = 1; when E2 - E1 > E0, judge that the current belt image is in the loaded state, that is, a = 0.

2. The belt unload detection method based on image processing according to claim 1, characterized in that, Step five specifies that the value of flag is the number of consecutive belt conveyor images in an unloaded state. The initial value of flag is set to 0, meaning that there are 0 belt conveyor images in an unloaded state when the system starts running. For each consecutive image in an unloaded state (a=1), flag is incremented by 1. When the system continuously determines that 30 or more images are in an unloaded state, the flag value is greater than or equal to 30, and the system will issue an unloaded alarm. If the image is determined to be loaded (a=0) during the determination process, the flag value is reset to 0. After the image is determined to be unloaded, the process can be terminated. If not, the system will continue to collect the current image and perform unload determination.

Images