A method for rejecting targets for a sorting system and use thereof

Abstract

The application discloses a kind of reject target object method for sorting system and its application, the steps of the method include:1.extract target object feature point image in material RGB image;2.extract background image in material RGB image;3.utilize the sliding window of dimension M×N to the background image erosion and carry out inflation, obtain the background binary image after pre-processing and feature point binary image;4.connectivity domain marks target object, statistics feature point number, mark out rejection object;5.calculate the falling time L of material in sorting system from start to completely pass through nozzle;6.calculate the falling motion time of the n th target object from discharge port to nozzle in sorting system;7.calculate the delay of nozzle to the n th target object to send rejection action, and after delay, nozzle produces blowing action in a certain range, to reject the n th target object.The application can accurately reject different shapes of material, so as to improve sorting effect.

Description

Technical Field

[0001] This invention belongs to the field of color sorting and sorting technology, specifically a method for removing target objects in a sorting system and its application. Background Technology

[0002] Existing sorting methods for rejecting target objects after identification involve using algorithms to identify material feature points and recording their positions. Since the material's viewpoint is approximately 4-10 cm away from the nozzle during the rejection action, a fixed delay is applied after the algorithm identifies the target object before the nozzle performs the blowing action to reject the identified material. Because the rejection process focuses on the location of the material's feature points, materials with feature points far from the material's center may not be rejected. Furthermore, for slender materials, a fixed delay before rejection may not reach the center, resulting in the material not being blown away. Summary of the Invention

[0003] The present invention addresses the shortcomings of the prior art by proposing a method for removing target objects in a sorting system and its application, aiming to accurately remove materials of different shapes, thereby improving the sorting effect and ensuring sorting accuracy.

[0004] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0005] The present invention provides a method for removing target objects in a sorting system, characterized by comprising the following steps:

[0006] Step 1: Use a CCD linear array sensor to acquire RGB images of materials in the sorting system and extract target feature point images from the RGB images of the materials.

[0007] Step 2: Extract the background image from the RGB image of the material;

[0008] Step 3: Use a sliding window of dimension M×N to erode and dilate the background-removed image to obtain the preprocessed image, thereby obtaining the feature point image;

[0009] Step 4: Mark the N target objects in the background removal image using four connected components, and mark the coordinates of the four vertices of the bounding rectangle of each target object. Count the number of corresponding feature points for each target object in the feature point image. If the number of feature points is greater than the set threshold for the number of feature points, the corresponding target object is removed; otherwise, the corresponding target object is retained.

[0010] Step 5: Calculate the falling time L of the nth target object in the sorting system from the moment it passes through the nozzle until it has completely passed through the nozzle. n ;

[0011] Step 6: Calculate the falling motion time T of the nth target object in the sorting system from the discharge opening to the nozzle. n ;

[0012] Step 7: Calculate the delay t for the nozzle to perform a rejection action on the nth target object. n , and after the delay t n , the nozzle generates a blowing action within the range of to reject the nth target object; where represents the abscissa of the left vertex of the circumscribed rectangle of the nth target object, represents the abscissa of the right vertex of the circumscribed rectangle of the nth target object;

[0013] When L n < TH1, it means that 1 rejection action is generated for the nth target object, and the delay t n = T n ;

[0014] When TH1 ≤ L n < TH2, it means that 2 rejection actions are generated for the nth target object, and the delay t n is obtained using Equation (1):

[0015]

[0016] When TH(m - 1) ≤ L n < THm, it means that m rejection actions are generated to reject the nth target object, and the delay t n is obtained using Equation (2):

[0017]

[0018] In Equations (1) - (2), TH1, TH2,..., THm are m thresholds respectively, and TH1 < TH2 <.... < THm.

[0019] The feature of a method for rejecting target objects in a sorting system according to the present invention also lies in that in Step 1, the feature point image of the target object in the RGB image of the material is extracted using Equation (3):

[0020]

[0021] In equation (3), g(x,y) represents any pixel in the RGB image of the material mapped to the color model; g1 and g2 represent the lower and upper thresholds of the target object feature points in the color model, respectively; h(x,y) represents the pixel in the target object feature point image corresponding to g(x,y); when h(x,y) = 1, it means that the pixel corresponding to g(x,y) is a feature point; when h(x,y) = 0, it means that the pixel corresponding to g(x,y) is a non-feature point.

[0022] In step 2, the background image is extracted from the RGB image of the material using equation (4):

[0023]

[0024] In equation (4), i(x,y) represents the pixel in the background image corresponding to g(x,y); g3 and g4 represent the upper and lower limits of the pixel threshold in the color model, respectively; when i(x,y) = 1, it means that the pixel corresponding to g(x,y) is a foreground pixel, and when i(x,y) = 0, it means that the pixel corresponding to g(x,y) is a background pixel.

[0025] In step 3, the feature point image is obtained using equation (5):

[0026] k(x,y)=j(x,y)&h(x,y) (5)

[0027] In equation (5), j(x,y) represents the pixel in the preprocessed image corresponding to g(x,y), k(x,y) represents the pixel in the feature point image corresponding to g(x,y), and & represents the logical AND operation.

[0028] In step 5, the falling time L is calculated using equation (6). n :

[0029]

[0030] In equation (6), Let be the ordinate of the upper vertex of the bounding rectangle of any nth target object. Let y be the ordinate of the lower vertex of the bounding rectangle of the nth target object; n∈[1,N].

[0031] In step 6, the falling motion time T is calculated using equation (7). n :

[0032]

[0033] In equation (7), h is the falling height of the material, and a nLet T be the acceleration of the nth target object, and T0 be the scan cycle of each row in the background image.

[0034] The present invention provides an electronic device, including a memory and a processor, characterized in that the memory is used to store a program supporting the processor in executing the target removal method, and the processor is configured to execute the program stored in the memory.

[0035] The present invention provides a computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, performs the steps of the target removal method.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] 1. This invention employs a method of multiple blowing actions from nozzles corresponding to the entire target object. After identifying the target object to be removed, a dynamic delay removal time is calculated, and the target object is removed multiple times. The removal position corresponds to the nozzle position of the entire target object, thereby solving the problem that slender materials cannot be removed due to a fixed delay time, and also solving the problem that some materials with uneven feature points cannot be removed because the removal position is the position of the target object's feature points.

[0038] 2. Before the connected component labeling in this invention, the binary image after background removal is subjected to an erosion and then dilation method, which filters out high-frequency noise and reduces the impact of noise in the image on the recognition results.

[0039] 3. In the process of labeling connected components, this invention does not label the background-removed binary image and the feature point binary image separately. Instead, it simultaneously counts the number of feature points in the connected component when labeling the background-removed binary image, thereby reducing the amount of computation and saving processor logic resources.

[0040] 4. This invention proposes a normalized calculation method for calculating the dynamic delay time of the target object, thereby simplifying the calculation of the dynamic delay time and making it easier to implement on the production line.

[0041] 5. The method of the present invention can be implemented using a low-end FPGA chip. Since FPGA uses parallel computing, it has a faster processing speed, fixed processing time, and significantly reduced cost compared to traditional processors. Attached Figure Description

[0042] Figure 1 This is a flowchart of the method of the present invention;

[0043] Figure 2 This is a schematic diagram of the target object falling into the rejection action of the present invention. Detailed Implementation

[0044] In this embodiment, as Figure 1 As shown, a method for removing target objects in a sorting system involves acquiring material falling images at a viewpoint using a linear array sensor, and then calculating background and target feature point images based on algorithms such as RGB components, HSV model, and color difference algorithm. Materials with a feature point count exceeding a set threshold are identified as target objects to be removed. The bounding rectangle of the target object is marked using connected components, the nozzle position for the removal action is calculated, and the removal delay time and number of removals are calculated longitudinally to achieve precise removal of target objects. Specifically, the method includes the following steps:

[0045] Step 1: Use a CCD linear array sensor to acquire RGB images of the materials in the sorting system, and use formula (1) to extract the target object feature point image from the RGB images of the materials:

[0046]

[0047] In equation (1), g(x,y) represents any pixel in the RGB image of the material mapped to the color model; g1 and g2 represent the lower and upper thresholds of the target object feature points in the color model, respectively; h(x,y) represents the pixel in the target object feature point image corresponding to g(x,y); when h(x,y) = 1, it means that the pixel corresponding to g(x,y) is a feature point; when h(x,y) = 0, it means that the pixel corresponding to g(x,y) is a non-feature point; the color model g(x,y) combines the commonly used HSV model and color difference algorithm, making feature point detection more accurate.

[0048] Step 2: Extract the background image from the RGB image of the material using formula (2):

[0049]

[0050] In equation (2), i(x,y) represents the pixel in the background image corresponding to g(x,y); g3 and g4 represent the upper and lower limits of the pixel threshold in the color model, respectively; when i(x,y) = 1, it means that the pixel corresponding to g(x,y) is a foreground pixel, and when i(x,y) = 0, it means that the pixel corresponding to g(x,y) is a background pixel.

[0051] Step 3: Use an M×N sliding window to erode and dilate the background-removed image to obtain the preprocessed image. The size of the M×N sliding window depends mainly on the size of the noise points in the background-removed image; the smallest sliding window size that can filter out the noise is preferred. The feature point image is obtained using equation (3):

[0052] k(x,y)=j(x,y)&h(x,y) (3)

[0053] In equation (3), j(x,y) represents the pixel in the preprocessed image corresponding to g(x,y), k(x,y) represents the pixel in the feature point image corresponding to g(x,y), and & represents the logical AND operation;

[0054] Step 4: Mark the N target objects in the background removal image using four connected components, and mark the coordinates of the four vertices of the bounding rectangle of each target object. Count the number of corresponding feature points for each target object in the feature point image. If the number of feature points is greater than the set threshold for the number of feature points, the corresponding target object is removed; otherwise, the corresponding target object is retained.

[0055] Step 5: Because FPGA uses parallel computing, the target object algorithm has a very short processing time. When the target object is identified, it has not yet fallen to the nozzle position. Therefore, the processor needs to delay for a certain period of time before sending the blowing signal. This delay determines where the air blown from the nozzle hits the object, thus determining whether the hit is accurate. Figure 2 As shown, the time unit obtained from the subsequent formula calculations is the row scanning cycle of the background image. Equation (4) is used to calculate the falling time L of the nth target object in the sorting system from the moment it passes through the nozzle until it has completely passed through the nozzle. n :

[0056]

[0057] In equation (4), Let be the ordinate of the upper vertex of the bounding rectangle of any nth target object. Let L be the ordinate of the lower vertex of the bounding rectangle of the nth target object; n∈[1,N]; L n In fact, it represents how many rows of scan cycles the target object occupies in the background image.

[0058] Step 6: Calculate the falling motion time T of the nth target object in the sorting system from the feed port to the nozzle using equation (5). n :

[0059]

[0060] In equation (5), h is the falling height of the material, and a n Let T0 be the acceleration of the nth target object, and T0 be the scan cycle of each row in the background image.

[0061] Step 7: Calculate the delay t of the nozzle issuing the rejection action to the nth target object. n And during the delay t n Afterwards, the nozzle... Generate a blowing action once within the range to remove the nth target object; where represents the abscissa of the left vertex of the circumscribed rectangle of the nth target object, represents the abscissa of the right vertex of the circumscribed rectangle of the nth target object;

[0062] When L n < TH1, it means generating 1 removal action for the nth target object and obtaining a delay t n = T n ;

[0063] When TH1 ≤ L n < TH2, it means generating 2 removal actions to remove the nth target object and obtaining the delay t using Equation (6) n :

[0064]

[0065] When TH(m - 1) ≤ L n < THm, it means generating m removal actions to remove the nth target object and obtaining the delay t using Equation (7) n :

[0066]

[0067] In Equations (6) - (7), TH1, TH2,..., THm are m thresholds respectively, and the sorting of the value ranges is TH1 < TH2 <.... < THm. As Figure 2 shown is a schematic diagram of three removal actions for any target object. At the delay a first blowing and removing action is issued, at the delay a second blowing and removing action is issued, and at the delay T a third blowing and removing action is issued.

[0068] In this embodiment, an electronic device includes a memory and a processor. Among them, the memory is used to store a program that supports the processor to execute the above method for removing target objects, and the processor is configured to execute the program stored in the memory.

[0069] In this embodiment, a computer-readable storage medium stores a computer program, and when the computer program is run by a processor, the steps of the above method for removing target objects are performed.

Claims

1. A method for removing target objects in a sorting system, characterized in that, Includes the following steps: Step 1: Use a CCD linear array sensor to acquire RGB images of materials in the sorting system and extract target feature point images from the RGB images of the materials. Step 2: Extract the background image from the RGB image of the material; Step 3: Use a sliding window of dimension M×N to erode and dilate the background-removed image to obtain the preprocessed image, thereby obtaining the feature point image; Step 4: Mark the N target objects in the background-removed image using four connected components, and mark the coordinates of the four vertices of the bounding rectangle of each target object. Count the number of corresponding feature points for each target object in the feature point image. If the number of feature points exceeds the set threshold for the number of feature points, the corresponding target object will be removed; otherwise, the corresponding target object will be retained. Step 5: Calculate the falling time of the nth target object in the sorting system from the moment it passes through the nozzle until it has completely passed through the nozzle. ; Step 6: Calculate the falling time of the nth target object in the sorting system from the feed inlet to the nozzle. ; Step 7: Calculate the delay of the nozzle issuing the rejection action for the nth target object. and during the delay Afterwards, the nozzle... A blowing action is generated within the range to eliminate the nth target object; where, This represents the x-coordinate of the left vertex of the bounding rectangle of the nth target object. The x-coordinate of the right vertex of the bounding rectangle of the nth target object; When L n <TH1, it means that one rejection action is performed on the nth target object, and the delay is obtained = ; When TH1 ≤ L n <When TH2, it means that two rejection actions are generated for the nth target object, and the delay is obtained using Equation (1) :[[]END]] (1) When TH(m - 1) ≤ L n <When THm, it means that the nth target object is subjected to m rejection actions, and the delay is obtained using Equation (2) : = (2) In equations (1)-(2), TH1, TH2, ..., THm are m thresholds, and TH1 <TH2<....<THm。 2. The method for removing target objects in a sorting system according to claim 1, characterized in that, In step 1, the target object feature point image in the RGB image of the material is extracted using formula (3): (3) In equation (3), This represents any pixel in the RGB image of the material mapped to the pixel in the color model. and These represent the lower and upper limits of the pixel threshold for the feature points of the target object in the color model, respectively. Indicates the corresponding feature points in the target object image pixels; when When, it indicates that it corresponds to The pixels are feature points, when When, it indicates that it corresponds to The pixels are non-feature points.

3. The method for removing target objects in a sorting system according to claim 2, characterized in that, In step 2, the background image is extracted from the RGB image of the material using equation (4): (4) In equation (4), Indicates the background image corresponding to The pixels; and These represent the upper and lower limits of the pixel threshold for the background point in the color model, respectively; when When, it indicates that it corresponds to The pixel is the foreground pixel, when When, it indicates that it corresponds to The pixels are background pixels.

4. The method for removing target objects in a sorting system according to claim 3, characterized in that, In step 3, the feature point image is obtained using equation (5): = & (5) In equation (5), This indicates that the preprocessed image corresponds to pixels, Indicating that the feature point image corresponds to The pixels, & represents the logical AND operation.

5. A method for removing target objects in a sorting system according to claim 4, characterized in that, In step 5, the falling time is calculated using equation (6). : (6) In equation (6), Let be the ordinate of the upper vertex of the bounding rectangle of any nth target object. Let y be the ordinate of the lower vertex of the bounding rectangle of the nth target object; ; This indicates how many rows of scan cycles the target object occupies in the background image.

6. A method for removing target objects in a sorting system according to claim 5, characterized in that, In step 6, the falling motion time is calculated using equation (7). : (7) In equation (7), h is the falling height of the material. Let n be the acceleration of the nth target object. This is the scan cycle for each row in the background image.

7. An electronic device, comprising a memory and a processor, characterized in that, The memory is used to store a program that supports the processor in executing the target removal method of claim 1, the processor being configured to execute the program stored in the memory.

8. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the steps of the target removal method of claim 1.

Images