A method for detecting boron ore blockage based on machine vision

Abstract

The invention relates to a boron ore blockage detection method based on machine vision. The method comprises the following steps: first, calling a camera device to obtain a picture of the actual working state of a feed port, and collecting a working picture of the feed port of an on-site crusher; second, performing grayscale processing on the collected picture to reduce the amount of calculation; third, performing an image enhancement operation on the grayscale image; fourth, performing a mask operation on the picture after the enhancement processing to shield an area outside the feed port; fifth, simultaneously performing straight line detection according to rule 1 and profile detection according to rule 2 at the feed port; and sixth, judging whether a blockage occurs at the current feed port according to a blockage image counter and a threshold value, and issuing a blockage alarm. The method collects on-site working images by a camera device, and pre-processes the images, thereby laying a foundation for subsequent detection. The method uses straight line detection and profile detection to judge whether a blockage occurs at the boron ore feed port, thereby reducing the frequency of the crusher's blockage operation and improving the working efficiency of the crusher.

Description

Technical Field

[0001] The present invention relates to the technical field of blockages, and in particular to a boron ore blockage detection method based on machine vision. Background Art

[0002] With the modernization of industrial technology, various minerals have become indispensable and important resources. Moreover, throughout the geological cycle, boron ore formation can range from magma to surface, and both endogenous and exogenous sources can produce industrial enrichment. Boron ore has a wide range of applications in various industries. Currently, most boron ore processing involves crushing the ore before transportation or testing. First, the mined ore is transported to a designated location and crushed by specific equipment. However, when the ore enters the crushing equipment, due to the uncertain size of the ore and the fixed size of the crushing equipment entrance, it is easy to cause blockage in the feed port when facing some large ores, preventing the remaining ore from entering the crusher. Generally speaking, most feed ports are unattended. Once blockage occurs, it will seriously affect the operation and production efficiency of the equipment. Due to the variable size and shape of the ore at the feed port, coupled with the influence of external factors such as the mine environment, traditional target recognition methods are not very effective in dealing with this problem, and the false alarm rate is high.

[0003] Chinese patent publication number CN112046957A discloses a method and device for monitoring and processing ore blockages, in which a control device receives target blockage information and target blockage image data at the raw material bin outlet; the control device obtains target blockage position image area data based on the target blockage information and target blockage image data at the raw material bin outlet and transmits it to a processing device; the processing device receives the target blockage position image area data and performs processing within the target blockage position image area; this solves the problems of requiring dedicated personnel for on-site processing, high equipment management costs, low equipment operating efficiency, and potential safety hazards; however, due to the complexity of the on-site conditions, the collected images are easily affected, the images are unclear or damaged, and the target blockage information cannot be determined, resulting in on-site misjudgment.

[0004] Chinese patent publication number CN110294286A discloses a material blockage detection device, including a sensing probe rod with one end extending into the material flow; a rotating shaft rotatably connected to the sensing probe rod; a proximity switch that selectively contacts the other end of the sensing probe rod; and a controller electrically connected to the proximity switch and including a timing module and an alarm module. The device is conducive to timely detection of abnormal conditions in the process flow at or upstream of the material blockage detection device, helps to avoid the expansion of accidents, and improves the smoothness of the process flow. The sensing probe rod is not easily damaged, and the material blockage detection device has a long service life. However, the sensing probe rod cannot detect the size of material particles, and material blockage detection may result in misjudgment and delay, thereby causing inaccurate blockage detection. Summary of the Invention

[0005] The present invention provides a boron ore blockage detection method based on machine vision. The method preprocesses an image, extracts real-time parameters of the boron ore feed opening by using straight line detection and profile detection, and compares the parameters with a set threshold value to detect blockage at the boron ore feed opening.

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

[0007] A method for detecting boron ore blockage based on machine vision comprises the following steps:

[0008] Step 1: Call the camera device to obtain the actual working status picture of the feed port and collect the working picture of the feed port of the crusher on site;

[0009] Step 2: Process the collected images in grayscale to reduce the amount of calculation;

[0010] Step 3: Perform image enhancement operation on the grayscale image;

[0011] Step 4: Perform a mask operation on the enhanced image to shield the area outside the feed port;

[0012] Step 5: Perform straight line detection according to rule 1 and cross-section detection according to rule 2 at the feed port simultaneously;

[0013] Step 6: Count the number of feature points that are not revealed under Rule 1 and the number of profiles when the profile variance under Rule 2 is greater than the threshold, and calculate their percentage of all feature points respectively, and set the initial value of the percentage to 0; select the blocking threshold of the feature area of the current image, and compare this threshold with the percentage. If the percentage is greater than this threshold, the current image is judged to be in a blocking state, otherwise it is in a normal state; set the blocking image counter flag, the initial value is 0, when the percentage is greater than the threshold, flag = flag + 1, if the percentage is less than the threshold, the counter is set to 0; select the threshold M of the blocking image number flag, if flag ≥ M, it can be judged that the current discharge port is blocked and a blocking alarm is issued, otherwise it returns to a normal state.

[0014] Furthermore, the rule 1 in step five is to select 16 groups of feature points at the feed port, which are recorded as H1~H16. Straight lines will be detected on the crossbeam of the boron ore crusher feed port and at the junction of the feed port and the surrounding fine sand. They will be used as feature points for straight line detection. When unloading, the rolling ore will cover the originally selected feature points. As the ore enters the crusher, some feature points will reappear. The feature points that are not exposed cannot detect the original straight line, and their number is counted and recorded as f1.

[0015] Furthermore, the rule 2 in step five is to draw multiple section lines at the feed inlet. As the sunlight moves, shadows will be generated at different positions of the feed inlet. The shadow parts will cause misjudgment of the detection. Therefore, multiple section lines should be drawn at the feed inlet, and the variance t1 of each section line is extracted. The corresponding threshold T1 is set for the variance of each section line. Section line detection in shadow areas that are prone to shadows will result in false alarms. Such false alarms are included in the fault tolerance range. During the unloading process, the number of section lines with t1>T1 is counted and recorded as f2.

[0016] Furthermore, the step six selects a blocking threshold for the feature area of the current image, and compares this threshold with the percentage. If the percentage is greater than this threshold, the current image is judged to be in a blocking state, otherwise it is in a normal state. 42 feature points are selected as the basis for judgment. If all the ore enters the crusher, all the feature points will be redetected, and their parameters will not differ from the original settings. If not all of them enter the crusher, some feature points will be blocked. At the beginning, the number of feature points covered at the feed port is 0, that is, no material is discharged. From the start of feeding, the coverage rate of the feed port begins to increase. When the feed port is completely covered by ore, the feature points are completely covered. Then the ore gradually enters the crusher, and the curve of the percentage of feature point coverage over time begins to decline. When the ore has not completely fallen into the crusher, feeding begins again, and the curve rises again. The maximum point of the curve indicates that the feed port is completely covered, the minimum point indicates that the ore has not completely fallen into the crusher, and the minimum point indicates that all the ore has fallen into the crusher.

[0017] Furthermore, the blocking image number flag in step 6 cannot determine whether the current state is blocking or unloading based on a single image. Only if the consecutive images are all judged to be in blocking state can it be determined that blocking has occurred at the feed port.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1) Collect on-site work images through the camera device and pre-process the images to lay the foundation for subsequent detection;

[0020] 2) Use straight line detection and profile detection to determine whether the boron ore feeding port is blocked;

[0021] 3) Reduce the frequency of crusher blockage operation and improve the working efficiency of the crusher. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic flow chart of the method of the present invention.

[0023] Figure 2 This is the flow chart of the feed port linear detection according to Rule 1 of the present invention.

[0024] Figure 3 This is the flow chart of feed port profile detection according to Rule 2 of the present invention.

[0025] Figure 4 It is a top view of the feed port of the present invention.

[0026] Figure 5 It is a schematic diagram of the linear detection of the feed port according to the present invention.

[0027] Figure 6 It is a schematic diagram of the feed port cross-sectional detection according to the present invention.

[0028] Figure 7 It is a curve showing the percentage of coverage of the feed port characteristic points according to the present invention changing with time. DETAILED DESCRIPTION

[0029] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0030] See Figure 1 , is a schematic flow chart of the method of the present invention. The present invention provides a boron ore blockage detection method based on machine vision, comprising the following steps:

[0031] Step 1: Call the camera device to obtain the actual working status picture of the feed port, and call the Hikvision device SDK file to collect the working picture of the crusher feed port on site;

[0032] Step 2: Use NIVISION to perform grayscale processing on the collected images. Grayscale processing is the most basic image processing method. It converts the color images collected on site into grayscale images, reduces the color channels, and reduces the amount of calculation. The formula is as follows:

[0033] Gray(i,j)=[R(i,j)+G(i,j)+B(i,j)] / 3(1)

[0034] Where (i, j) is the pixel coordinate, Gray(i, j) is the grayscale value at (i, j) in the image, and R, G, and B are the three color channels;

[0035] Step 3: Use histogram equalization to perform image enhancement processing on the grayscale image. By changing the histogram of the feed inlet image, the grayscale of each pixel in the image is changed, the contrast of the image is enhanced, and the histogram of the original feed inlet image is transformed into a uniformly distributed form.

[0036] Step 4: Based on the coordinates of the pixel points around the feed port, create a mask for the area outside the feed port, block the area outside the feed port, and change its pixel value to 0, that is, display it as black, to control the area of image processing. The formula is as follows:

[0037] I(i,j)=5*I(i,j)-[I(i-1,j)+I(i+1,j)+i(i,j-1)+I(i,j+1)](2)

[0038] Where (i, j) is the pixel coordinate and I is the mask matrix;

[0039] Step 5: Perform straight line detection according to rule 1 and cross-section detection according to rule 2 at the feed port simultaneously;

[0040] Step 6: Count the number of feature points that are not revealed under Rule 1 and the number of profiles when the profile variance under Rule 2 is greater than the threshold, and calculate their percentage of all feature points respectively, and set the initial value of the percentage to 0; select the blocking threshold of the feature area of the current image, and compare this threshold with the percentage. If the percentage is greater than this threshold, the current image is judged to be in a blocking state, otherwise it is in a normal state; set the blocking image counter flag, the initial value is 0, when the percentage is greater than the threshold, flag = flag + 1, if the percentage is less than the threshold, the counter is set to 0; select the threshold M of the blocking image number flag, if flag ≥ M, it can be judged that the current discharge port is blocked and a blocking alarm is issued, otherwise it returns to a normal state.

[0041] See Figure 2 、 Figure 4 and Figure 5Furthermore, the rule 1 in step five is to select 16 groups of feature points at the feed port, which are recorded as H1~H16. The feed port of the boron ore crusher is circular as a whole, and the outer ring is covered with a layer of fine sand to buffer the pressure of the falling ore. A thicker rectangular beam is installed in the middle, and an eccentric sleeve is installed in the center of the beam. A conical crushing cone is installed below the eccentric sleeve and penetrates into the bottom of the crusher to crush large pieces of ore. When selecting the ROI area, the feed port beam can detect a straight line, and a clear dividing line will be formed at the junction of the feed port and the surrounding fine sand. Straight lines can be detected within a smaller range. These straight lines can be used as feature points for detection. When unloading, the rolling ore will cover the originally selected feature points. As the ore enters the crusher, some feature points will reappear. The feature points that are not exposed cannot detect the original straight lines, and their number is counted and recorded as f1.

[0042] See Figure 3 、 Figure 4 , further, the rule 2 in the step 5 is to draw a profile line on the feed port part. The line profile is mainly used to display the grayscale value corresponding to the pixel point on a line in the image, and can output the maximum value, minimum value, average value, variance, and standard deviation of the grayscale value on the line. The boron ore feed port works 24 hours a day. Due to the different positions of the sun, shadows will be generated at different positions. When the sun is at position A, area 7 and area 8 will be shadowed; when the sun is at position C, area 5 and area 6 will be shadowed; when the sun is at position B, area 5 and area 7 will be shadowed; when the sun is at position D, area 6 and area 8 will be shadowed;

[0043] See Figure 6 Taking the case when the sun is at position A as an example, when the sun is at position A, shadows will appear in areas A and C, and the overall color is dark. When no ore falls, the pixel value of the section line inside the area is low. However, in area B, regardless of whether there is sunlight shadow, the color of area B itself is darker and the pixel value is low, and the influence of the shadow is negligible. Therefore, the judgment results of areas A and B may not be completely used as the basis for judging whether there is material blockage, and false alarms are prone to occur. Therefore, multiple section lines should be drawn in the area of the material discharge port, and the variance t1 of each section line should be extracted. The corresponding threshold T1 is set for the variance of each section line, and the section lines of areas A and B are used as a certain basic fault tolerance rate. During the material discharge process, the number of section lines with t1>T1 is counted and recorded as f2.

[0044] Furthermore, in step 6, 42 feature points are selected as the basis for judgment. If all the ore enters the crusher, all the feature points will be re-detected, and their parameters will not be different from the original settings; if not all the ore enters the crusher, some feature points will be blocked;

[0045] See Figure 7 At the beginning, the number of feature points covered at the feeding port is 0, which means no material is being fed; from the start of feeding, the coverage rate of the feeding port begins to increase, and all the feature points are covered. Then the ore gradually enters the crusher, and the curve of the percentage of feature point coverage changing with time begins to decrease. When the ore has not completely fallen into the crusher, feeding begins again, and the curve rises again. The maximum point of the curve indicates that the feeding port is completely covered, the minimum points 1, 2, and 3 indicate that the ore has not completely fallen into the crusher, and the minimum point 4 indicates that the ore has completely fallen into the crusher; F is selected as the blocking threshold of the current image feature area, and F is compared with f. If f>F, the current image is judged to be in a blocking state, otherwise it is in a normal state.

[0046] Furthermore, the initial value of the number of blocking images flag in step six is 0, and the threshold value M of the number of blocking images flag is selected. A single image cannot determine whether the current state is blocking or unloading. Only if the continuously judged images are in blocking state can it be determined that a blockage has occurred in the feed port. In the present invention, a picture is captured every 5 seconds, so the threshold value M can be selected as 30, that is, 150 seconds. According to the actual situation on site, if it is continuously judged as blocking within 150 seconds, it can be proved that a blockage has indeed occurred in the feed port. If flag≥M, it can be determined that a blockage has occurred in the current unloading port and a blockage alarm can be issued. The staff is reminded of the blockage by the alarm music, and the alarm log is saved at the same time so that the maintenance personnel can perform maintenance and adjust the parameters according to the alarm log. Otherwise, it returns to normal.

[0047] The above embodiments are implemented under the premise of the technical solution of the present invention, and detailed implementation methods and specific operation processes are given, but the protection scope of the present invention is not limited to the above embodiments. The methods used in the above embodiments are conventional methods unless otherwise specified.

Claims

1. A boron ore blockage detection method based on machine vision, characterized in that: The steps include: Step 1: Call the camera device to obtain the actual working status picture of the feed port and collect the working picture of the feed port of the crusher on site; Step 2: grayscale the collected images to reduce the amount of calculation; Step 3: Perform image enhancement operation on the grayscale image; Step 4: Perform a mask operation on the enhanced image to shield the area outside the feed port; Step 5: Perform straight line detection according to rule 1 and cross-section detection according to rule 2 at the feed port simultaneously; Step 6: Count the number of feature points that are not revealed under Rule 1 and the number of profiles when the profile variance under Rule 2 is greater than the threshold, and calculate their percentage of all feature points respectively, and set the initial value of the percentage to 0; select the blocking threshold of the feature area of the current image, and compare this threshold with the percentage. If the percentage is greater than this threshold, the current image is judged to be in a blocking state, otherwise it is in a normal state; set the blocking image counter flag, the initial value is 0, when the percentage is greater than the threshold, flag = flag + 1, if the percentage is less than the threshold, the counter is set to 0; select the threshold M of the blocking image number flag, if flag ≥ M, it can be judged that the current discharge port is blocked and a blocking alarm is issued, otherwise it returns to a normal state.

2. The method for detecting boron ore blockage based on machine vision according to claim 1, characterized in that: Rule 1 in step five is to select 16 groups of feature points at the feed port, denoted as H1 to H16. Straight lines will be detected on the crossbeam of the boron ore crusher feed port and at the junction of the feed port and the surrounding fine sand. These straight lines will be used as feature points for straight line detection. When unloading, the rolling ore will cover the originally selected feature points. As the ore enters the crusher, some feature points will reappear. The feature points that are not exposed cannot detect the original straight line, and their number is counted and recorded as f1.

3. The method for detecting boron ore blockage based on machine vision according to claim 1, characterized in that: Rule 2 in step five is to draw multiple section lines at the feed inlet. As the sunlight moves, shadows will be generated at different positions of the feed inlet. The shadow parts will cause misjudgment of the detection. Therefore, multiple section lines should be drawn at the feed inlet, and the variance t1 of each section line is extracted. The corresponding threshold T1 is set for the variance of each section line. Section line detection in shadow-prone areas may result in false alarms. Such false alarms are included in the fault tolerance range. During the unloading process, the number of section lines with t1>T1 is counted and recorded as f2.

4. The method for detecting boron ore blockage based on machine vision according to claim 1, wherein: As described in step six, a blocking threshold of the feature area of the current image is selected, and the threshold is compared with the percentage. If the percentage is greater than the threshold, the current image is judged to be in a blocking state, otherwise it is in a normal state; 42 feature points are selected as the basis for judgment. If all the ore enters the crusher, all the feature points will be re-detected, and their parameters will not differ from the original settings; if not all of them enter the crusher, some feature points will be blocked; at the beginning, the number of feature points covered at the feeding port is 0, that is, no material is discharged; from the start of feeding, the coverage rate of the feeding port begins to increase. When the feeding port is completely covered by ore, the feature points are completely covered. Then the ore gradually enters the crusher, and the curve of the percentage of feature point coverage over time begins to decline. When the ore has not completely fallen into the crusher, feeding begins again, and the curve rises again. The maximum point of the curve indicates that the feeding port is completely covered, the minimum point indicates that the ore has not completely fallen into the crusher, and the minimum point indicates that all the ore has fallen into the crusher.

5. The method for detecting boron ore blockage based on machine vision according to claim 1, characterized in that: The blocking image number flag in step 6 cannot determine whether the current state is blocking or unloading based on a single image. Only if the consecutive images are all judged to be in blocking state can it be determined that blocking has occurred in the feed port.

Images