Ore sorting method, device and storage medium based on fusion of x-ray image and visible light image

Abstract

The application belongs to the technical field of ore sorting, and particularly relates to an ore sorting method and device based on X-ray image and visible light image fusion and a storage medium. The method comprises the following steps: collecting a first initial image set, the first initial image set comprising first X-ray images, second X-ray images and visible light images of each ore sample in an ore sample set, the first X-ray images corresponding to X-ray energy greater than that of the second X-ray images; determining a frequency scatter plot envelope based on the first initial image set; constructing an ore statistical model based on the frequency scatter plot envelope; collecting a second initial image set, the second initial image set comprising first X-ray images, second X-ray images and visible light images of a material to be measured; determining a single-target feature sequence based on the second initial image set; and sorting the material to be measured based on the single-target feature sequence and the ore statistical model, thereby improving the efficiency and accuracy of ore sorting.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ore sorting, and particularly relates to an ore sorting method and device based on X-ray image and visible light image fusion and a storage medium. BACKGROUND

[0002] X-ray transmission (XRT) technology is the most widely used technology in the current ore sorting field, and its core content is to determine ore and waste by analyzing the different attenuation rates of high and low energy X-rays after transmitting the ore. Different properties of substances have different attenuation rates. However, the attenuation rates of some target ore bodies and non-target surrounding rocks are not obviously different, so additional features need to be introduced to improve the recognition accuracy. Visible light color selection is a reliable feature compensation method, which can directly determine ore and waste by extracting the surface features such as color, texture and crystal shape of the ore. At present, it has become a trend to combine XRT technology and visible light color selection technology for ore sorting.

[0003] Chinese patent application with publication number CN117299581A discloses an intelligent ore sorting method based on visible light and X-ray image fusion, which includes the following steps: S1, when the ore on the conveying belt passes through the CCD color line scan camera and the X-ray line array scan camera, the CCD color line scan camera and the X-ray line array scan camera respectively collect visible light image data and X-ray image data of the ore, and send them to the recognition host, i.e. industrial computer; S2, the recognition host, i.e. industrial computer, receives the two kinds of image data, and performs image registration and affine transformation on them through the image processing module to align them; then, edge detection is performed on the visible light image data to extract edge information and determine the frame to select the visible light image data and X-ray image of the target ore; then, the two kinds of images after alignment and registration are input into the encoder in sequence to extract features, and the generated feature map is fused using or addition strategy, and the fused feature map is reconstructed by the decoder to obtain a fused image; S3, the fused image is input into the subsequent neural network classification model to extract features and classify and distinguish, and the class and confidence of each ore are output, and the ore is divided into high grade, medium grade and low grade categories according to the pre-set classification standard and threshold, and then the image processing module sends the classification results and control signals to the control module, and this step is realized by the deep learning module in the recognition host. However, this method needs to be continuously trained and optimized to realize image fusion and ore sorting, which is computationally complex, resulting in low efficiency and accuracy of ore sorting.

[0004] Therefore, how to improve the efficiency and accuracy of ore sorting based on X-ray image and visible light image fusion has become a problem to be solved. SUMMARY

[0005] The application provides a mineral sorting method and device based on fusion of X-ray images and visible light images and a storage medium, aiming to improve the efficiency and accuracy of mineral sorting based on fusion of X-ray images and visible light images.

[0006] In a first aspect, the embodiments of the application provide a mineral sorting method based on fusion of X-ray images and visible light images, which comprises:

[0007] Collecting a first initial image set, the first initial image set comprising first X-ray images, second X-ray images and visible light images of each mineral sample in a mineral sample set, the first X-ray images corresponding to X-ray energy greater than that of the second X-ray images;

[0008] Determining a frequency scatter plot envelope based on the first initial image set;

[0009] Constructing a mineral statistical model based on the frequency scatter plot envelope;

[0010] Collecting a second initial image set, the second initial image set comprising first X-ray images, second X-ray images and visible light images of a material to be tested;

[0011] Determining a single-target feature sequence based on the second initial image set;

[0012] Sorting the material to be tested based on the single-target feature sequence and the mineral statistical model.

[0013] In a possible implementation, the determining of the frequency scatter plot envelope based on the first initial image set comprises:

[0014] Data registration and data fusion are performed on the first X-ray images, the second X-ray images and the visible light images corresponding to each mineral sample in the first initial image set to obtain a fusion image corresponding to each mineral sample;

[0015] A first multi-target feature sequence is determined based on the fusion image corresponding to each mineral sample;

[0016] The frequency scatter plot envelope is determined based on the first multi-target feature sequence.

[0017] In a possible implementation, the constructing of the mineral statistical model based on the frequency scatter plot envelope comprises:

[0018] A first target window is used to perform a sliding window search on the frequency scatter plot envelope to determine a first target region, the first target region being a region with the highest frequency in the frequency scatter plot envelope;

[0019] searching along normal directions of the frequency scatter point envelope from a spherical center point of the first target region to determine a first target coordinate point set;

[0020] constructing the ore statistical model based on the first target coordinate point set.

[0021] In a possible implementation, the first initial image set further includes first X-ray images, second X-ray images and visible light images of each waste rock sample in the waste rock sample set; and the frequency scatter point envelope includes an ore sample frequency scatter point envelope and a waste rock sample frequency scatter point envelope.

[0022] The determining, based on the first initial image set, of the frequency scatter point envelope includes:

[0023] performing data registration and data fusion on the first X-ray images, the second X-ray images and the visible light images corresponding to each ore sample and each waste rock sample in the first initial image set to obtain a fusion image corresponding to each ore sample and each waste rock sample;

[0024] determining a second multi-target feature sequence based on the fusion image corresponding to the ore sample, and determining the ore sample frequency scatter point envelope based on the second multi-target feature sequence;

[0025] determining a third multi-target feature sequence based on the fusion image corresponding to each waste rock sample, and determining the waste rock sample frequency scatter point envelope based on the third multi-target feature sequence.

[0026] In a possible implementation, the constructing, based on the frequency scatter point envelope, of the ore statistical model includes:

[0027] performing a sliding window search on the ore sample frequency scatter point envelope and the waste rock sample frequency scatter point envelope based on a second target window to determine a second target region and a third target region, the second target region being a highest frequency region of the ore sample frequency scatter point envelope, and the third target region being a highest frequency region of the waste rock sample frequency scatter point envelope;

[0028] searching along normal directions of the ore sample frequency scatter point envelope and the waste rock sample frequency scatter point envelope from a spherical center point of the second target region and the third target region to determine a waste-ore boundary surface and a candidate coordinate point set;

[0029] determining a second target coordinate point set based on the waste-ore boundary surface and the candidate coordinate point set;

[0030] constructing the ore statistical model based on the second target coordinate point set.

[0031] In a possible implementation, the determining of the single-target feature sequence based on the second initial image set comprises:

[0032] performing data registration on each image in the second initial image set to obtain a target image set;

[0033] performing data fusion on each image in the target image set to obtain a target fusion image;

[0034] determining a single-target feature sequence based on the target fusion image, the single-target feature sequence comprising three-dimensional coordinates corresponding to all pixel point values of each channel in the target fusion image.

[0035] In a possible implementation, the performing of data registration on each image in the second initial image set to obtain a target image set comprises:

[0036] aligning starting positions of each image in the second initial image set based on acquisition times of the images;

[0037] dividing each image after alignment of the starting positions into a plurality of intervals according to a preset distance, and setting a standard pixel value of the intervals;

[0038] counting actual pixel values of each interval in each image after alignment of the starting positions;

[0039] when an actual pixel value of a target interval is greater than the standard pixel value, determining a pixel position to be removed of the target interval based on a difference between the actual pixel value and the standard pixel value, the target interval being any interval;

[0040] performing horizontal distortion correction on each image after alignment of the starting positions based on the pixel position to be removed of each interval to obtain the target image set.

[0041] In a possible implementation, the sorting of the material to be measured based on the single-target feature sequence and the ore statistical model comprises:

[0042] counting feature points in the single-target feature sequence that exist in the ore statistical model as target feature points;

[0043] calculating a ratio of a number of the target feature points to a number of all feature points in the single-target feature sequence;

[0044] when the ratio is greater than a preset threshold, determining that the material to be measured is ore; and when the ratio is less than or equal to the preset threshold, determining that the material to be measured is waste rock.

[0045] In a second aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method according to the first aspect or any one of the implementation forms thereof when executing the computer program.

[0046] In a third aspect, a computer-readable storage medium is provided, which stores a computer program, and the computer program is executable by a processor to implement the method according to the first aspect or any one of the implementation forms thereof.

[0047] In a fourth aspect, a computer program product is provided, which, when executed on an electronic device, causes the electronic device to perform the method according to the first aspect or any one of the implementation forms thereof.

[0048] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: based on the first initial image set containing the X-ray high-energy image, the X-ray low-energy image and the visible light image of each ore sample in the ore sample set, the frequency scatter plot envelope is determined, and based on the frequency scatter plot envelope, the ore statistical model is constructed, so that the ore statistical model can reflect the pixel distribution of the ore, the more the number of ore samples in the ore sample set, the more accurate and the wider the coverage of the final obtained ore statistical model; based on the second initial image set containing the X-ray high-energy image, the X-ray low-energy image and the visible light image of the to-be-measured material, the single-target feature sequence corresponding to the to-be-measured material is determined; based on the single-target feature sequence and the ore statistical model, the to-be-measured material is sorted, without the need for constant model training and optimization, the calculation complexity is reduced, and the efficiency and accuracy of ore sorting are improved.

[0049] It can be understood that the electronic device, the computer-readable storage medium and the computer program product provided by the embodiments of the present application have the same beneficial effects as the ore sorting method based on fusion of X-ray images and visible light images as described above, and will not be described here. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Figure 1 A flowchart of a method of ore sorting based on fusion of X-ray images and visible light images provided by an embodiment of the present application is shown in the figure.

[0052] Figure 2 A schematic diagram of an image acquisition device according to an embodiment of the present application is provided.

[0053] Figure 3 A flowchart of an implementation of S150 of a method of ore sorting based on fusion of X-ray images and visible light images according to an embodiment of the present application is provided.

[0054] Figure 4 A schematic diagram of a transverse distortion correction scale according to an embodiment of the present application is provided.

[0055] Figure 5 A schematic diagram of a comparison of RGB and HSV models according to an embodiment of the present application is provided.

[0056] Figure 6 A schematic diagram of an image fusion approach according to an embodiment of the present application is provided. DETAILED DESCRIPTION

[0057] In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular sequences of steps, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, circuits, and

[0058] It will be understood that the term "includes," "comprises," "includes," or "has" when used in the specification and throughout the claims, indicates the presence of the described features, integers, steps, operations, elements, and / or components but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0059] It will be understood that the term "and / or," when used in the specification and throughout the claims, means any one and / or all possible combinations of the associated listed items and includes all possible combinations.

[0060] As used in the specification and throughout the claims, the term "if" can be interpreted as meaning "when" or "upon" or "in response to a determination" or "in response to a detection" depending on the context. Similarly, the phrase "if determined" or "if detected [the described condition or event]" can be interpreted as meaning "upon a determination" or "in response to a determination" or "upon a detection of [the described condition or event]" or "in response to a detection of [the described condition or event]" depending on the context.

[0061] In addition, in the description of the present application and the appended claims, the terms "first", "second", "third", etc. are used only to distinguish descriptions and cannot be understood as indicating or implying relative importance.

[0062] Reference to "one embodiment" or "some embodiments" or "one implementation" or "some implementations" described in the present application means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearance of the phrases "in one embodiment", "in some embodiments", "in other some embodiments", "in yet some embodiments" or the like in various places in the specification is not necessarily all referring to the same embodiment, but means "one or more but not all embodiments", unless otherwise specifically stated. The terms "include", "contain", "have" and their variants mean "including but not limited to", unless otherwise specifically stated.

[0063] For ease of understanding, the technical solutions of the present application will be described in detail below with reference to the accompanying drawings.

[0064] Figure 1 A flowchart of a mineral sorting method based on fusion of X-ray images and visible light images provided by an embodiment of the present application is shown. As shown in Figure 1 for ease of description, only parts related to the present embodiment are shown, the method provided by the present embodiment includes the following steps:

[0065] S110, a first initial image set is collected, the first initial image set including first X-ray images, second X-ray images and visible light images of each mineral sample in a mineral sample set, the X-ray energy corresponding to the first X-ray images being greater than the X-ray energy corresponding to the second X-ray images.

[0066] Specifically, a sufficient amount of mineral samples is selected as the mineral sample set, a charge-coupled device (CCD) color line scanning camera is used to collect the visible light images of each mineral sample, and an X-ray transmission device is used to collect the first X-ray images and the second X-ray images of each mineral sample, wherein the X-ray transmission device includes an X-ray emitter and an X-ray line array scanning camera and the like. The first X-ray image is an X-ray high-energy image, and the second X-ray image is an X-ray low-energy image.

[0067] As an example, an image collection device as shown in Figure 2 is used to collect the first initial image set, as shown in Figure 2As shown, the CCD color line scanning camera and the X-ray transmission device are arranged above the conveying belt, and are spaced apart from each other, each ore sample in the ore sample set is sequentially conveyed onto the conveying belt, the CCD color line scanning camera scans the ore surface along the movement direction of the conveying belt to obtain a visible light image of each ore sample, and the X-ray emitter in the X-ray transmission device can simultaneously or quickly switch two X-ray beams of different energies to scan the ore sample, and the X-ray linear array scanning camera scans the ore sample surface along the movement direction of the conveying belt, the X-ray image acquired under the high-energy X-ray beam is a first X-ray image, and the X-ray image acquired under the low-energy X-ray beam is a second X-ray image.

[0068] In S120, a frequency scatter point envelope is determined based on the first initial image set.

[0069] In a possible implementation, S120 can optionally include the following steps:

[0070] The first X-ray image, the second X-ray image and the visible light image corresponding to each ore sample in the first initial image set are subjected to data registration and data fusion to obtain a fusion image corresponding to each ore sample.

[0071] A first multi-target feature sequence is determined based on the fusion image corresponding to each ore sample.

[0072] The frequency scatter point envelope is determined based on the first multi-target feature sequence.

[0073] Specifically, the fusion image corresponding to each ore sample is a three-channel image, and the three channels of the fusion image correspond to the first X-ray image, the second X-ray image and the visible light image respectively.

[0074] In a specific implementation, the fusion image corresponding to each ore sample is subjected to target segmentation. Since the fusion image has been subjected to data registration, segmentation of a single channel can complete segmentation of the entire fusion image. Moreover, X-ray transmission imaging is not affected by light and other environments, and the imaging is clear and reliable, and the segmentation effect is better than that of a visible light image. Therefore, the channel corresponding to the second X-ray image is preferably selected as a target channel for target segmentation to obtain a first multi-target image sequence corresponding to the ore sample set. The pixel point values of all three channels in the first multi-target image sequence are extracted respectively, and the three values are projected into a three-dimensional space, i.e., the three-channel pixel values in the first multi-target image sequence are mapped into X, Y and Z three-dimensional axis coordinates to obtain a first multi-target feature sequence. The points with the same coordinates in the first multi-target feature sequence are combined in frequency to form a frequency scatter point envelope in the three-dimensional space.

[0075] In another possible implementation, the first initial image set further includes first X-ray images, second X-ray images and visible light images of each waste rock sample in the waste rock sample set; the frequency scatter plot envelope includes an ore sample frequency scatter plot envelope and a waste rock sample frequency scatter plot envelope; correspondingly, S120 can optionally include the following steps:

[0076] Data registration and data fusion are performed on the first X-ray images, the second X-ray images and the visible light images corresponding to each ore sample and each waste rock sample in the first initial image set, to obtain a fusion image corresponding to each ore sample and each waste rock sample.

[0077] Based on the fusion image corresponding to the ore sample, a second multi-target feature sequence is determined, and based on the second multi-target feature sequence, the ore sample frequency scatter plot envelope is determined.

[0078] Based on the fusion image corresponding to each waste rock sample, a third multi-target feature sequence is determined, and based on the third multi-target feature sequence, the waste rock sample frequency scatter plot envelope is determined.

[0079] In a specific implementation, the second X-ray image corresponding to the fusion image corresponding to each ore sample is taken as a target channel for target segmentation, to obtain a second multi-target image sequence corresponding to the ore sample set. Pixel point values of all pixels in three channels in the second multi-target image sequence are extracted respectively, and the three values are projected into a three-dimensional space to obtain a second multi-target feature sequence. Points with the same coordinates in the second multi-target feature sequence are combined in frequency, to form an ore sample frequency scatter plot envelope in the three-dimensional space. Similarly, the second X-ray image corresponding to the fusion image corresponding to each waste rock sample is taken as a target channel for target segmentation, to obtain a third multi-target image sequence corresponding to the waste rock sample set. Pixel point values of all pixels in three channels in the third multi-target image sequence are extracted respectively, and the three values are projected into a three-dimensional space to obtain a third multi-target feature sequence. Points with the same coordinates in the third multi-target feature sequence are combined in frequency, to form a waste rock sample frequency scatter plot envelope in the three-dimensional space.

[0080] It should be noted that the steps of data registration and data fusion performed on the first X-ray images, the second X-ray images and the visible light images corresponding to each ore sample and each waste rock sample in the step can refer to the contents of S1501 and S1502 described below or the steps of image data registration and data fusion in the prior art, and will not be described here in detail.

[0081] S130, based on the frequency scatter plot envelope, constructing an ore statistical model.

[0082] In a possible implementation, S130 can optionally include the following steps:

[0083] Based on the first target window, a sliding window search is performed on the frequency scatter point envelope diagram to determine a first target region, and the first target region is a highest frequency region of the frequency scatter point envelope diagram;

[0084] From the spherical center point of the first target region, a search is performed along each normal direction of the frequency scatter point envelope diagram to determine a first target coordinate point set;

[0085] Based on the first target coordinate point set, a mineral statistical model is constructed.

[0086] In a specific implementation, a first round of median filtering is performed on the frequency scatter point envelope diagram, with a window size of 5*5*5, to smooth the local maximum frequency points and compensate for the local abnormal frequency 0 points. Then, a spherical surface with a radius of a preset size is set as the first target window, and a sliding window search is performed on the frequency scatter point envelope diagram in the three-dimensional space. The frequency values of all coordinate points in the first target window are counted to locate the highest frequency region. When there are multiple highest frequency regions, the radius of the spherical surface corresponding to the first target window is appropriately expanded by the preset size, and the search is performed again until a unique first target region is located. From the spherical center point of the first target region, a search is first performed outward along any normal direction of the spherical surface of the frequency scatter point envelope diagram, with a preferred search step of 1. All coordinate points that are searched and have a frequency greater than 0 are taken as target coordinate points. When the frequency of the current coordinate point searched is 1 and the frequency of the next coordinate point along the current normal direction is 0, the search in this round is ended. The next round of search is started along another normal direction of the spherical surface of the frequency scatter point envelope diagram. The target coordinate points are recorded until all points in the frequency scatter point envelope diagram are searched. All target coordinate points recorded in the search process are taken as the first target coordinate point set, and the first target coordinate point set is taken as the mineral statistical model.

[0087] However, because there is a gray zone of intersection between the pixel points corresponding to the ore and the pixel points corresponding to the waste rock, the above implementation only constructs the mineral statistical model by searching the frequency scatter point envelope diagram of the ore sample, and records the pixel points intersecting with the waste rock sample in the mineral statistical model, resulting in low accuracy of the constructed mineral statistical model. Based on this, another possible implementation is proposed to solve the technical problems in the above implementation and improve the accuracy of the mineral statistical model.

[0088] In another possible implementation, S130 can optionally include the following steps:

[0089] Based on the second target window, a sliding window search is performed on the frequency scatter point envelope diagram of the ore sample and the frequency scatter point envelope diagram of the waste rock sample respectively to determine a second target region and a third target region. The second target region is the highest frequency region of the frequency scatter point envelope diagram of the ore sample, and the third target region is the highest frequency region of the frequency scatter point envelope diagram of the waste rock sample.

[0090] Starting from the center point of the second and third target regions respectively, the search is carried out along the normal directions of the frequency scatter envelope of ore samples and the frequency scatter envelope of waste rock samples to determine the boundary surface between ore and waste rock and the set of candidate coordinate points.

[0091] Based on the boundary surface of the mine waste and the set of candidate coordinate points, determine the set of coordinate points for the second target.

[0092] A statistical model for ore is constructed based on the set of coordinate points of the second target.

[0093] In practice, the scatter plots of frequency distributions for both ore and waste rock samples are first subjected to a median filter with a window size of approximately 5×5×5. A sphere with a preset radius is then used as the second target window. A sliding window search is performed on both the ore and waste rock sample scatter plots in three-dimensional space. By statistically analyzing the frequency values ​​of all points within the second target window, the region with the highest frequency is located. If multiple regions with the highest frequency exist, the preset radius of the sphere corresponding to the second target window is appropriately increased, and the search is repeated until... Locate the unique second and third target regions corresponding to the scatter plots of ore sample frequencies and waste rock sample frequencies, respectively. Starting from the center point of each target region, search outwards along any normal direction of the scatter plots of ore sample frequencies and waste rock sample frequencies, with a preferred search step size of 1. Use all searched coordinate points representing ore samples with frequencies greater than 0 as candidate coordinate points. When two search paths intersect, or when they do not intersect, select the target region corresponding to the scatter plot of ore sample frequencies. When searching in the network diagram, if the frequency of the current coordinate point is 1 and the frequency of the next coordinate point along the current normal direction is 0, the search for that round of the scatter envelope diagram of the ore sample frequency is stopped. Similarly, when searching in the scatter envelope diagram of the waste rock sample frequency, if the frequency of the current coordinate point is 1 and the frequency of the next coordinate point along the current normal direction is 0, the search for that round of the scatter envelope diagram of the waste rock sample frequency is stopped. The spherical searches of the ore sample frequency scatter envelope diagram and the waste rock sample frequency scatter envelope diagram are performed simultaneously, that is, each completes one round of search before proceeding along the ore sample frequency scatter envelope diagram. The next round of search is performed outward along the other normal direction of the point envelope map and the scatter envelope map of waste rock sample frequency. The candidate coordinate points and the intersection points of the two search paths are recorded until all points in the scatter envelope map of ore sample frequency and the scatter envelope map of waste rock sample frequency are searched. The intersection points recorded during the search process are used to construct the mine-waste boundary surface. The coordinate points of the candidate coordinate points located in the region formed by the mine-waste boundary surface and the scatter envelope map of ore sample frequency are used to construct the second target coordinate point set. The second target coordinate point set is used as the ore statistical model.

[0094] In the possible implementation, the pixel points that are coincident with the ore sample frequency scatter point envelope graph are screened out from the ore sample frequency scatter point envelope graph by the determined ore waste boundary surface, and the remaining pixel points of the ore sample frequency scatter point envelope graph after removing the coincident pixel points are constructed as the ore statistical model, thereby improving the accuracy of the ore statistical model.

[0095] S140, a second initial image set is collected, the second initial image set including a first X-ray image, a second X-ray image and a visible light image of the material to be measured.

[0096] In a specific implementation, the second initial image set is collected by using an image collection device as shown in Figure 2 The material to be measured is transported onto a conveying belt, a CCD color line scanning camera scans the surface of the material to be measured along the movement direction of the conveying belt to obtain the visible light image of the material to be measured, and an X-ray line array scanning camera in an X-ray transmission device scans the surface of the material to be measured along the movement direction of the conveying belt, collects the first X-ray image of the material to be measured when a high-energy X-ray beam is emitted by an X-ray emitter, and collects the second X-ray image of the material to be measured when a low-energy X-ray beam is emitted by the X-ray emitter.

[0097] S150, a single-target feature sequence is determined based on the second initial image set.

[0098] In a possible implementation, as shown in Figure 3 S150, the following steps can be further included:

[0099] S1501, data registration is performed on each image in the second initial image set to obtain a target image set.

[0100] As an example, step S1501 can include the following steps:

[0101] The starting positions of each image are aligned based on the collection time of each image in the second initial image set;

[0102] Each image after the starting position alignment is divided into a plurality of intervals according to a preset distance, and a standard pixel value of the interval is set;

[0103] Actual pixel values of each interval in each image after the starting position alignment are counted;

[0104] When the actual pixel value of a target interval is greater than the standard pixel value, a pixel position to be removed of the target interval is determined based on the difference between the actual pixel value and the standard pixel value, and the target interval is any interval;

[0105] Each image after the starting position alignment corresponding to the pixel position to be removed of each interval is subjected to horizontal distortion correction to obtain the target image set.

[0106] In a specific implementation, the first X-ray image, the second X-ray image and the visible light image of the to-be-tested material are set to have the same resolution, so that the pixel sizes of the images of the target object are consistent; preferably, the pixel size of each image is set to 0.8 dots per inch (dpi), that is, the real size of each pixel is 0.8 millimeters (mm); the time offset generated by each image is obtained according to the acquisition time of the first X-ray image, the second X-ray image and the visible light image of the to-be-tested material, and then, in combination with the distance between the CCD color line scanning camera and the X-ray transmission device and the rotation speed of the transmission belt on which the to-be-tested material is placed, a time stamp is set for each image, so that the starting positions of the images are aligned.

[0107] Further, the distortion of the industrial camera is generated by the lens, and each image in the present application is collected by the line scanning camera, so that only the horizontal correction distortion needs to be performed, and one preferred example of the horizontal distortion correction is as follows:

[0108] 1) A scale with intervals of 8 mm is made, and the scale pattern is as shown in the following figure: Figure 4 Based on the scale, the images after the starting positions are aligned are divided into intervals with a preset distance of 8 mm; since the resolution of the first X-ray image, the second X-ray image and the visible light image of the to-be-tested material is set to 0.8 dpi, the standard pixel value corresponding to each interval in the scale is 10 pixels.

[0109] 2) The central part of the scale is placed directly below each image and the scale image is taken.

[0110] 3) The middle part of the image usually has no distortion, so the interval between the two scale lines near the middle of each image is 10 pixels, and the interval becomes larger as it approaches the two sides; starting from the vertical middle line of each image, the actual pixel values between the scale lines are counted in sequence to the two sides, for example, when the actual pixel value of the target interval is 11, one column at the middle position of the two scale lines corresponding to the target interval is taken as the to-be-removed pixel position. Similarly, when the actual pixel value of the target interval is n (n>10), the number of columns corresponding to the to-be-removed pixel position is n-10, and the to-be-removed pixel position is i / n of the adjacent two scale lines, where i∈(0, n-10). The positions of the to-be-removed pixel positions in the original scale image are marked, and the position sequence is denoted as Q.

[0111] 4) The columns corresponding to the position sequence Q are removed, and the remaining columns after the removal are filled to the middle of each image to obtain a target image set. For example, when the actual pixel value of the target interval is 11, one column at the middle position of the two scale lines corresponding to the target interval is taken as the to-be-removed pixel position and is removed, and the remaining columns after the removal are filled to the middle of the image.

[0112] S1502, data fusion is performed on each image in the target image set to obtain a target fused image.

[0113] Specifically, the visible light image of the material to be measured collected by the CCD color line scanning camera is in RGB mode, and the visible light image needs to be converted from RGB mode to HSV mode because HSV mode has intuitive color and is more suitable for computer processing mode. The difference between RGB model and HSV model is shown in Figure 5 wherein 5(a) is the RGB model and 5(b) is the HSV model. The present application focuses on the color information of the target object, so only the chroma component representing the color information is extracted in the HSV mode, which is denoted as H_color, while the RGB model needs to use three components to represent the color information, which has poor intuitiveness of feature quantization and is easily affected by light intensity. Moreover, the use of single channel reduces the data volume by 2 / 3, reduces the calculation amount, and improves the real-time performance of the system.

[0114] In a specific implementation, the first X-ray image and the second X-ray image in the target image set after data registration and lateral distortion correction are both compressed to 8 bits, and are denoted as H_XRT and L_XRT respectively, which further compresses the calculation amount while ensuring the same number of bits as the chroma component H_color, facilitating the visualization operation of feature fusion and subsequent feature normalization operation. As shown in Figure 6 The first X-ray image, the second X-ray image and the chroma component are merged into a new three-channel image according to the fusion mode (H_XRT, L_XRT, H_color), that is, a target fusion image.

[0115] S1503, based on the target fusion image, determining a single target feature sequence, the single target feature sequence including three-dimensional coordinates corresponding to all pixel point values in each channel of the target fusion image.

[0116] In a specific implementation, the channel corresponding to the second X-ray image is preferably used as the channel for target segmentation. Since the target fusion image has realized data registration, segmentation of a single channel can complete the segmentation of the entire three-channel image. X-ray transmission imaging is not affected by light and other environments, and the imaging is clear and reliable, so the segmentation effect is better than that of visible light image. After target segmentation, a single target image sequence is obtained, all pixel point values in the three channels of the single target image sequence are extracted, and the three values are projected into a three-dimensional space, that is, the three-channel pixel values in the single target image sequence are mapped into X, Y and Z three-dimensional axis coordinates, to determine the single target feature sequence.

[0117] S160, based on the single target feature sequence and the ore statistical model, sorting the material to be measured.

[0118] In one possible implementation, S160 can optionally include the following steps:

[0119] The feature points existing in the ore statistical model in the single-target feature sequence are counted as target feature points.

[0120] A ratio of the number of the target feature points to the number of all feature points in the single-target feature sequence is calculated.

[0121] When the ratio is greater than a preset threshold, the material to be tested is determined to be ore; and when the ratio is less than or equal to the preset threshold, the material to be tested is determined to be waste rock.

[0122] It should be noted that the preset threshold can be self-defined according to actual conditions, and the present application does not limit this.

[0123] The technical scheme provided in the embodiment is based on a first initial image set containing X-ray high-energy images, X-ray low-energy images and visible light images of each ore sample in an ore sample set, determines a frequency scatter plot envelope, and based on the frequency scatter plot envelope, constructs an ore statistical model, so that the ore statistical model can reflect the pixel distribution of the ore, the more the number of ore samples in the ore sample set, the more accurate and the wider the coverage of the final obtained ore statistical model; and based on a second initial image set containing X-ray high-energy images, X-ray low-energy images and visible light images of the material to be tested, a single-target feature sequence corresponding to the material to be tested is determined; and based on the single-target feature sequence and the ore statistical model, the material to be tested is sorted, without the need for constant model training and optimization, thereby reducing the calculation complexity and improving the efficiency and accuracy of ore sorting.

[0124] The present application combines dual-energy XRT technology with visible light color selection technology, solves the problem of target recovery metal accompanied by other low-value metal interference, and through position matching and data fusion of XRT images and CCD images, more beneficial features are obtained to distinguish different properties of substances. It can be applied in the fields of ore sorting, construction waste recycling, production line product incoming sorting, etc.

[0125] On the other hand, the present application also provides a computer storage medium storing executable program codes; the executable program codes are used to execute any of the above ore sorting methods based on fusion of X-ray images and visible light images.

[0126] On the other hand, the present application also provides an electronic device including a memory and a processor; the memory stores program codes executable by the processor; the program codes are used to execute any of the above ore sorting methods based on fusion of X-ray images and visible light images.

[0127] The program code can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the program code in the electronic device.

[0128] The electronic device can be a desktop computer, a notebook computer, a palm computer, a cloud server, and the like. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the electronic device can also include an input / output device, a network access device, a bus, and the like.

[0129] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and the like. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0130] The memory can be an internal storage unit of the electronic device, such as a hard disk or a memory. The memory can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Further, the memory can include both the internal storage unit and the external storage device of the electronic device. The memory is used to store the program code and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or will be output.

[0131] The above computer storage medium and electronic device are created based on the above-mentioned emotion analysis method, and the technical effects and advantages thereof will not be repeated here. The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above-mentioned embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, it should be considered as the scope of the present application.

[0132] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A method for ore sorting based on fusion of X-ray images and visible light images, characterized by, The method comprises: collecting a first initial image set, the first initial image set comprising first X-ray images, second X-ray images and visible light images of each ore sample in an ore sample set, the first X-ray images corresponding to X-ray energy greater than the X-ray energy corresponding to the second X-ray images; based on the first initial image set, determining a frequency scatter plot envelope; comprising: data registration and data fusion of the first X-ray images, the second X-ray images and the visible light images corresponding to each ore sample in the first initial image set to obtain a fusion image corresponding to each ore sample; based on the fusion image corresponding to each ore sample, determining a first multi-target feature sequence; based on the first multi-target feature sequence, determining the frequency scatter plot envelope; the fusion image corresponding to each ore sample is a three-channel image, corresponding to the first X-ray image, the second X-ray image and the visible light image respectively; taking the channel corresponding to the second X-ray image as the target channel for target segmentation to obtain a first multi-target image sequence corresponding to the ore sample set, extracting all pixel point values of the three channels in the first multi-target image sequence respectively, and projecting the three values into a three-dimensional space to obtain a first multi-target feature sequence; and then merging the points with the same coordinates in the first multi-target feature sequence to form a frequency scatter plot envelope in the three-dimensional space; based on the frequency scatter plot envelope, constructing an ore statistical model; comprising: based on a first target window, performing sliding window search on the frequency scatter plot envelope to determine a first target region, the first target region being the highest frequency region of the frequency scatter plot envelope; starting from the spherical center point of the first target region, searching along each normal direction of the frequency scatter plot envelope to determine a first target coordinate point set; based on the first target coordinate point set, constructing the ore statistical model; collecting a second initial image set, the second initial image set comprising first X-ray images, second X-ray images and visible light images of the material to be measured; based on the second initial image set, determining a single-target feature sequence; based on the single-target feature sequence and the ore statistical model, sorting the material to be measured.

2. The method of claim 1, wherein, The first initial image set further comprises first X-ray images, second X-ray images and visible light images of each waste rock sample in a waste rock sample set; and the frequency scatter plot envelope comprises an ore sample frequency scatter plot envelope and a waste rock sample frequency scatter plot envelope. The method comprises: data registration and data fusion of the first X-ray images, the second X-ray images and the visible light images corresponding to each ore sample and each waste rock sample in the first initial image set to obtain a fusion image corresponding to each ore sample and each waste rock sample; based on the fusion image corresponding to each ore sample, determining a second multi-target feature sequence, and based on the second multi-target feature sequence, determining the ore sample frequency scatter plot envelope; Determine a third multi-target feature sequence based on the corresponding fusion images of the waste rock samples, and determine a waste rock sample frequency scatter plot envelope based on the third multi-target feature sequence.

3. The method of claim 2, wherein, Based on the frequency scatter plot envelope, construct an ore statistical model, including: Based on the second target window, perform sliding window search on the ore sample frequency scatter plot envelope and the waste rock sample frequency scatter plot envelope respectively to determine a second target region and a third target region, the second target region being the highest frequency region of the ore sample frequency scatter plot envelope, and the third target region being the highest frequency region of the waste rock sample frequency scatter plot envelope; Start searching from the spherical center points of the second target region and the third target region respectively along the normal directions of the ore sample frequency scatter plot envelope and the waste rock sample frequency scatter plot envelope to determine a waste rock-ore boundary surface and a candidate coordinate point set; Determine a second target coordinate point set based on the waste rock-ore boundary surface and the candidate coordinate point set; Construct the ore statistical model based on the second target coordinate point set.

4. The method of claim 1, wherein, The method comprises the following steps: Perform data registration on each image in the second initial image set to obtain a target image set; Perform data fusion on each image in the target image set to obtain a target fusion image; Determine a single-target feature sequence based on the target fusion image, the single-target feature sequence including three-dimensional coordinates corresponding to all pixel point values in each channel of the target fusion image.

5. The method of claim 4, wherein, The method comprises the following steps: Align the starting positions of each image in the second initial image set based on the acquisition times of the images; Divide each image after starting position alignment into intervals according to a preset distance and set a standard pixel value for each interval; Statistically analyze the actual pixel values of each interval in each image after starting position alignment; When the actual pixel value of a target interval is greater than the standard pixel value, determine the pixel position to be removed in the target interval based on the difference between the actual pixel value and the standard pixel value, the target interval being any interval; Perform horizontal distortion correction on each image after starting position alignment based on the pixel position to be removed in each interval to obtain the target image set.

6. The method according to any one of claims 1 to 5, characterized in that, The method comprises the following steps: Statistically analyze the feature points in the single-target feature sequence that exist in the ore statistical model as target feature points; Calculate the ratio of the number of target feature points to the number of all feature points in the single-target feature sequence; When the ratio is greater than a preset threshold, determine that the material to be tested is ore; when the ratio is less than or equal to the preset threshold, determine that the material to be tested is waste rock.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1 to 6.

8. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 7. The computer program is executed by the processor to implement the method of any one of claims 1 to 6.

Images