Automatic identification method for assisting in clamping aerospace cable

Abstract

The invention discloses an automatic identification method for assisting in clamping an aerospace cable. The method comprises the following steps: acquiring a cable image; using the cable image to extract and obtain a binary cable area image; in the binarized cable area image, a cable skeleton line is extracted and obtained; pruning processing is carried out on the cable skeleton line; and scanning the pruned cable skeleton line to obtain the coordinates of the skeleton line arranged in sequence in different directions, and calculating to obtain the coordinates of the cable clamping center point. According to the invention, an operator does not need to tighten the tested cable, and the accurate clamping position of the cable is automatically obtained through one-time photographing of the camera in a state that the cable is not shielded and is bent and coiled on a desktop.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic production of aerospace cables, and in particular to an automatic identification method for assisting in clamping aerospace cables. BACKGROUND

[0002] With the increasing complexity of spacecraft functions, the scale of cables is increasing, and the production mode mainly relying on manual operation and manual inspection significantly limits the development progress. Due to the flexible physical characteristics of aerospace cable products, they will randomly exhibit different forms under the action of external force, such as Figure 1 As shown in the figure, it is therefore difficult to realize the automatic processing of cable harnesses. Using computer vision technology to guide the mechanical arm and mechanical claw to grasp the product is a common practice in the field of industrial automation, and it can be successfully implemented for products of fixed size and size.

[0003] When grasping or clamping the cable, how to efficiently identify the flat range of the cable and accurately clamp the effective fixing point of the cable is an important action for realizing the automatic production of the cable. At present, when existing computer vision technology guides the identification of the target, it usually adopts the way of image preprocessing, making a training sample library, picture feature labeling, and neural network training to realize stable identification of a specific target. For aerospace cables, the application of existing methods has the following problems:

[0004] (1) Random change of cable form during transfer processing

[0005] The cable exhibits different forms during transfer processing due to the influence of external force; in addition, the cable harness usually contains multiple branch points, making the placement shape extremely uncertain. For this feature-uncertain target detection, it is impossible to apply the mature industrial vision detection process to solve the geometric size feature detection of aerospace cables; in addition, due to the characteristics of multiple varieties and variable batches of cables, it is difficult to bear the high sample standard training and hardware cost.

[0006] (2) Lack of comprehensive consideration of cable physical and mechanical characteristics

[0007] For the clamping scene in the automatic system such as robot and mechanical arm, it is necessary to guide the clamping mechanism in the visual stage to accurately clamp the center point of the cable, so as to ensure that the cable does not easily slip, drag, interfere and other problems in the subsequent automatic transfer and processing stage. The existing technology can only clamp a single wire, and there is no mature scheme reference for the whole visual guidance and center point clamping of the cable. SUMMARY

[0008] The technical problem solved by the present application is to overcome the shortcomings of the prior art and provide an automatic identification method for assisting in clamping aerospace cables. According to the characteristics of aerospace cables, the accurate skeleton line of complex cable form is obtained, and the accurate clamping position of the cable is obtained.

[0009] The technical solution of the present application is: 1. An automatic identification method for assisting in clamping spaceflight cables, comprising the following steps:

[0010] Collecting a cable image;

[0011] Extracting and processing the cable image to obtain a binary cable region image;

[0012] Extracting a cable skeleton line from the binary cable region image;

[0013] Pruning the cable skeleton line;

[0014] Scanning the pruned cable skeleton line to obtain sequentially arranged coordinates in different directions of the skeleton line, and calculating cable clamping center point coordinates.

[0015] Further, the camera used to collect the cable image is calibrated using Zhang Zhengyou's calibration method.

[0016] Further, the step of extracting and processing the cable image includes image filtering, HSV threshold segmentation, and image preprocessing.

[0017] Further, the image filtering method is: in the spatial domain, using neighborhood averaging method for denoising; in the frequency domain, using low-pass filter for denoising.

[0018] Further, the method of extracting the cable skeleton line is: from the outer edge of the cable to the center of the cable, using a preset size pixel window centered on the pixel to be processed to continuously erode and thin the cable until it is eroded to a single pixel width, obtaining the cable skeleton line.

[0019] Further, the method of pruning the cable skeleton line is:

[0020] For each pixel point on the cable skeleton line, form an 8-neighborhood intersection point set for each pixel point, denoted as p8;

[0021] Find a point set that meets the m-adjacency in the p8 set, denoted as pm, and find an isolated intersection point set in the p8 set, denoted as pl;

[0022] Calculate the number of pixel points between all points in the pm set and the pl set and the nearest end point;

[0023] Record the point set in the pm set and the pl set whose pixel point number < threshold as ps, and delete all points in the ps set and the pixel points on the path from the nearest end point to the points in the ps set.

[0024] Furthermore, the cable skeleton lines after pruning are scanned to obtain the coordinates of the skeleton lines arranged sequentially in different directions, and the coordinates of the cable clamp center point are calculated. The specific method is as follows:

[0025] 1) Traverse the pruned cable skeleton lines to obtain the first non-zero pixel, denoted as P, and store the coordinates of pixel P into array line1.

[0026] 2) Perform an 8-neighborhood scan on point P to obtain the first non-zero pixel P11. Store the coordinates of pixel P11 in array line1 and set the pixel value of point P to 0.

[0027] 3) Continue scanning the 8-neighborhood of point P11 to obtain the first non-zero pixel P12. Store the coordinates of pixel P12 in array line1. Set the pixel value of point P12 to 0.

[0028] 4) Continue scanning the 8-neighborhood of point P12, and so on, until P1... n When performing an 8-neighborhood scan, it is impossible to obtain a non-zero pixel, i.e., P1. n If there are no other pixels around point P, it means that the first direction of scanning of point P is complete.

[0029] 5) For P, P11, P12...P1 stored in the line1 array n Calculate the Euclidean geometric distance between all coordinates of the pixel, denoted as D1;

[0030] 6) Return to point P and scan in the other direction of point P. That is, when performing the 8-neighborhood scan of point P in step 2, find the next non-zero pixel P21 and store its coordinates in array line2. Continue the 8-neighborhood scan on point P21 to obtain the first non-zero pixel P22 and store its coordinates in array line2. Set the pixel value of point P12 to 0. Continue this process until the values ​​of P, P21, P12...P2 stored in array line2 are obtained. m Calculate the Euclidean geometric distance between all coordinates of the pixel, denoted as D2;

[0031] 7) By analogy, after completing the traversal of the cable skeleton lines, all Euclidean geometric distances D1 to D2 are obtained. i The maximum value D in max Based on the coordinate array line corresponding to the maximum value max Output the array line max The coordinates of the 1 / 2 position point are the coordinates of the center point of the cable clamp.

[0032] The present invention also relates to a computer program product that, when executed by a processor, implements the steps of the method as described above.

[0033] The present invention also relates to a computer-readable storage medium, characterized in that: when a computer program within the medium is executed by a processor, it implements the steps of the method as described above.

[0034] The advantages of this invention compared to the prior art are:

[0035] This invention focuses on the characteristics of aerospace cables and proposes a method for rapidly extracting cable skeleton lines and pruning. By taking a single photo, the accurate clamping position of the cable can be automatically determined. Compared with traditional technical solutions, this invention does not require operators to tighten the cable under test. With no obstructions and the cable bend reel placed on a table, the optimal cable clamping position is output through a skeleton line calculation method optimized by taking a photo. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of aerospace cables;

[0037] Figure 2 This is a schematic diagram of the method flow of the present invention;

[0038] Figure 3 This is a comparison image of the extracted cable skeleton wires of the present invention;

[0039] Figure 4 This is a partial enlarged view of the skeleton lines and redundant branches of the present invention;

[0040] Figure 5 This is an exemplary diagram of scenario 1 of the present invention, which uses an 8-neighborhood search intersection point and a pruned skeleton line.

[0041] Figure 6 This is an exemplary diagram of scenario 2 of the present invention, which uses an 8-neighborhood search intersection point and a pruned skeleton line.

[0042] Figure 7 This is an exemplary diagram of the intersection point and the skeleton line after optimized pruning in scenario 2 of the present invention;

[0043] Figure 8 This is a schematic diagram of the 8-neighborhood search for tumors in scenario 1 of the present invention;

[0044] Figure 9 This is an exemplary diagram of the m-neighborhood search intersection point and the pruned skeleton line in scenario 1 of the present invention;

[0045] Figure 10 This is an exemplary diagram of scenario 3 of the present invention, which uses m-neighborhood search for intersection points.

[0046] Figure 11 This is a schematic diagram of the cable after the branch reduction according to the present invention. Detailed Implementation

[0047] To better understand the technical solution of the present invention, the specific embodiments of the present invention are described below.

[0048] This method proposes an automatic identification method for assisting in clamping aerospace cables. Based on the characteristics of aerospace cables, the core idea is to remove redundant information from the cable image, obtain its precise skeleton line for complex cable shapes, and then determine the accurate clamping position of the cable based on the characteristics of the skeleton line. This method refers to... Figure 2 Specifically, it includes the following steps:

[0049] (1) Acquire cable images

[0050] This invention preferably uses an imaging device consisting of a backlight and an industrial camera for image acquisition. After setting up the imaging device, the industrial camera needs to be calibrated, preferably using the Zhang Zhengyou calibration method. After camera calibration, the mapping between the actual cable size and the pixel size can be obtained, facilitating the conversion from pixel size to actual size during final clamping.

[0051] (2) Extraction of binary cable region image

[0052] After obtaining the cable image, a binarized cable region image that can be used to extract the cable skeleton line is needed. The main steps include image filtering, HSV thresholding, and image preprocessing.

[0053] Image filtering can employ commonly used methods in the industry, mainly including spatial domain methods and frequency domain methods. Preferably, in the spatial domain, this invention uses the neighborhood averaging method to calculate the average value of the pixels covered by the template instead of the current pixel value to reduce noise; in the frequency domain, since the energy of the cable image is mainly concentrated in the low-frequency part, while edge energy and noise are mainly concentrated in the high-frequency part, a low-pass filter is preferred to filter out noise.

[0054] HSV thresholding involves converting the cable image to the HSV color space after image filtering and then using an adaptive thresholding method to segment the cable's HSV regions. This is because different cables have different thresholds due to their own factors and the external environment. Therefore, the method based on adaptive thresholding is used to consider the differences between the cable's outline and the background.

[0055] Image preprocessing, after HSV thresholding, includes steps such as image grayscale conversion and morphological operations, which are common methods in the industry and will not be described here.

[0056] (3) Extraction of cable skeleton and pruning treatment

[0057] After obtaining the binarized cable region image, this invention proposes a fast method for extracting the cable skeleton lines; specifically:

[0058] Starting from the outer edge of the cable and moving towards the center, a preset-sized (3*3) pixel window centered on the pixel to be processed is used to continuously erode and refine the cable until it can no longer be eroded, i.e., the width of a single pixel, thus obtaining the skeleton line of the cable image. Figure 3 (a) shows a binarized image of the cable. Figure 3 As shown in (b), the image binarization representation with a single pixel width can be obtained by extracting the skeleton line. The branches of the skeleton line are not shown in this figure.

[0059] After skeleton line extraction, the resulting skeleton lines actually have redundant branches, such as... Figure 4 As shown in (a), the skeleton lines extracted for a certain cable are shown. Figure 4 (b) is a magnified view of the redundant branches of the skeleton line. It can be seen that this part contains many redundant branches. These skeleton branches will make the skeleton structure complex and the skeleton information redundant, which will make the skeleton measurement and skeleton information extraction process difficult.

[0060] To ensure the accuracy of the clamping points, a cable skeleton with single connectivity is required, thus pruning the original skeleton line is necessary. This invention proposes a method for searching intersection points based on a combination of 8-neighborhood and m-neighborhood. First, intersection points are identified, then endpoints are determined. Skeleton pruning is completed by deleting line segments whose length from the endpoint to the intersection point is less than a certain value. An intersection point is defined as a point with ≥3 pixels within its 8-neighborhood, and an endpoint is defined as a non-intersection point with ≤1 adjacent pixels in its 8-neighborhood set. The derivation process of this method is as follows:

[0061] First, using an 8-neighborhood search, the skeleton line in case 1 contains redundant branch regions and intersection points, such as... Figure 5 (a) shows the gray dots. It can be seen that in case 1, although multiple intersections occur based on the 8-neighborhood decision, since only the length at the first intersection is determined, the pruning effect is sufficient from this local example alone. Figure 5 As shown in (b). However, upon examining other parts of the pruning image, the following issues were found with other prunings.

[0062] For scenario 2, Figure 6 As shown in (a), the gray point is identified as an intersection point because it has three pixels in its 8-neighborhood. After pruning, although a redundant branch point is removed, the intersection point is located on the main trunk, resulting in the complete skeleton line being cut into two parts after pruning. Figure 6 As shown in (b), breaks in the skeleton line will affect subsequent length measurements. Therefore, to avoid skeleton line breaks, consider retaining intersections during deletion and only deleting points other than intersections, such as... Figure 7 As shown.

[0063] according to Figure 7 At this point, the intersection point located on the main trunk is preserved, so the skeletal line will not break. However, if the intersection point is complex, multiple intersection point determinations may occur. In this case, preserving the intersection point can lead to the appearance of lumps, such as... Figure 8 As shown, this is not conducive to further processing.

[0064] As can be seen from the above process, relying solely on 8-neighborhood search to determine intersection points is insufficient for practical processing needs. In such cases, using m-adjacency instead of 8-adjacency can better determine intersection points. Figure 9 (b) shows the gray dots. In this case, using m-neighborhood is clearly more scientific, as it can better eliminate lumps and find more precise intersection points. However, due to the diversity of intersection point types, simply using m-neighborhood cannot complete the pruning of the skeleton line, for example... Figure 10 As shown. Figure 10 The gray dot shown should be the intersection point we are looking for. However, according to the m-neighborhood determination rule, this point does not belong to the intersection point at this time, which means that the corresponding branch will not be deleted during pruning. If determined by the 8-neighborhood, this point obviously belongs to the branch point. Therefore, a complete skeleton line pruning should be a combination of 8-neighborhood determination and m-neighborhood determination. The complete method is described below:

[0065] 1) For each pixel on the cable skeleton line, form an 8-neighborhood intersection set for each pixel, denoted as p8;

[0066] 2) Find the set of points in set p8 that satisfy the m-adjacency property, denoted as pm, and find the set of isolated intersection points in set p8, denoted as pl; pm and pl together constitute the set of all intersection points.

[0067] 3) Calculate the number of pixels between all points in the pm and pl sets and their nearest neighboring endpoint;

[0068] 4) Denote the set of points with less than 20 pixels in both the pm and pl sets as ps. Delete all points in the ps set and all pixels on the path between a point in the ps set and its nearest endpoint.

[0069] After the above processing, a complete skeleton line pruning diagram can be obtained as follows: Figure 11 As shown, a skeleton line with a single connectivity was obtained.

[0070] (4) Calculate the coordinates of the center point of the cable clamp.

[0071] In the previous section, we completed the extraction and pruning of the skeleton lines. This section will explain how to scan a smooth skeleton line to obtain the coordinates of the skeleton line arranged sequentially in different directions, and calculate the appropriate clamping points. The method is as follows:

[0072] 1) Traverse the pruned cable skeleton line from left to right and from top to bottom, obtain the first non-zero pixel point, denoted as P, and store the coordinates of pixel point P into array line1.

[0073] 2) Perform an 8-neighborhood scan on point P to obtain the first non-zero pixel P11. Store the coordinates of pixel P11 in array line1 and set the pixel value of point P to 0.

[0074] 3) Continue scanning the 8-neighborhood of point P11 to obtain the first non-zero pixel P12. Store the coordinates of pixel P12 in array line1. Set the pixel value of point P12 to 0.

[0075] 4) Continue scanning the 8-neighborhood of point P12, and so on, until P1... n When performing an 8-neighborhood scan, it is impossible to obtain a non-zero pixel, i.e., P1. n If there are no other pixels around point P, it means that the first direction of scanning of point P is complete.

[0076] 5) For P, P11, P12...P1 stored in the line1 array n Calculate the Euclidean geometric distance between all coordinates of the pixel, denoted as D1;

[0077] 6) Return to point P and scan in the other direction of point P. That is, when performing the 8-neighborhood scan of point P in step 2, find the next non-zero pixel P21 and store its coordinates in array line2. Continue the 8-neighborhood scan on point P21 to obtain the first non-zero pixel P22 and store its coordinates in array line2. Set the pixel value of point P12 to 0. Continue this process until the values ​​of P, P21, P12...P2 stored in array line2 are obtained. m Calculate the Euclidean geometric distance between all coordinates of the pixel, denoted as D2;

[0078] 7) By analogy, after completing the traversal of the cable skeleton lines, all Euclidean geometric distances D1 to D2 are obtained. i The maximum value D in max Based on the coordinate array line corresponding to the maximum value max Output the array line max The coordinates of the 1 / 2 position point are the coordinates of the center point of the cable clamp.

[0079] It is understood that this invention has been described through embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of this invention. Furthermore, under the teachings of this invention, these features and embodiments can be modified to adapt to specific circumstances without departing from the spirit and scope of this invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are protected by this invention.

[0080] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. An automated method of identifying a secondary clamp for holding a space cable, the method comprising: It comprises the following steps: Collecting cable image; Extracting and processing the cable image to obtain a binary cable region image; Extracting the cable skeleton line in the binary cable region image; Pruning the cable skeleton line; Scanning the pruned cable skeleton line to obtain the coordinates of the skeleton line in different directions, and calculating the cable clamping center point coordinates.

2. The method of claim 1, wherein: The camera used to collect the cable image is calibrated by Zhang Zhengyou calibration method.

3. The method of claim 1, wherein: The step of extracting and processing the cable image includes image filtering, HSV threshold segmentation and image preprocessing.

4. The method of claim 2, wherein: The image filtering method is: in the spatial domain, using neighborhood averaging method for denoising; in the frequency domain, using low-pass filter for denoising.

5. The method of claim 1, wherein: The method of extracting the cable skeleton line is: from the outer edge of the cable to the center direction of the cable, using a preset size pixel window centered on the pixel to be processed to continuously erode and thin the cable until it is eroded to a single pixel width, and the cable skeleton line is obtained.

6. The method of claim 1, wherein: The method of pruning the cable skeleton line is: For each pixel point on the cable skeleton line, an 8-neighborhood intersection point set of each pixel point is formed, denoted as p8; Find the point set that meets the m-adjacency in the p8 set, denoted as pm, and find the isolated intersection point set in the p8 set, denoted as pl; Calculate the number of pixel points between all points in the pm set and pl set and the nearest end point; The point set with pixel point number < threshold in the pm set and pl set is denoted as ps, and all points in the ps set and the pixel points on the path from the nearest end point to the points in the ps set are deleted.

7. The method of claim 1, wherein: The method of scanning the pruned cable skeleton line to obtain the coordinates of the skeleton line in different directions is: 1) Traverse the pruned cable skeleton line to obtain the first non-zero pixel point, denoted as P, and store the coordinates of the pixel point P in the array line1; 2) Perform 8-neighborhood scanning on the P point to obtain the first non-zero pixel point P11, continue to store the coordinates of the pixel point P11 in the array line1, and set the pixel value of the P point to 0; 3) Continue to perform 8-neighborhood scanning on the P11 point to obtain the first non-zero pixel point P12, continue to store the coordinates of the pixel point P12 in the array line1, and set the pixel value of the P12 point to 0. 4) continue 8-neighbor scanning for P12, and so on; until P1 n No pixel point other than 0 can be obtained when 8-neighbor scanning is performed for P1 n No pixel point other than 0 can be obtained when 8-neighbor scanning is performed for P1 5) P, P11, P12...P1n stored in line1 array are calculated as follows: n All coordinates of the pixel points are calculated, and the Euclidean distance is recorded as D1. 6) return to P point, scan another direction of P point, that is, when step 2 obtains the next pixel point P21 which is not 0 in the 8-neighborhood scanning of P point, store the coordinates of pixel point P21 in array line2; continue to perform 8-neighborhood scanning on P21 point to obtain the first pixel point P22 which is not 0, continue to store the coordinates of pixel point P22 in array line2; set the pixel value of P12 point to 0; in sequence, ultimately calculate P, P21, P12...P2 m All coordinates of the pixel points, calculate the Euclidean distance, and record as D2; 7) In turn, after completing the cable skeleton line traversal, get all the Euclidean distances D1~D i The maximum value D max Among them, according to the coordinate array line max Corresponding to the maximum value, output the 1 / 2 position coordinate point of the array line max , that is, the cable clamping center point coordinate.

8. A computer program product, characterized by: The computer program product is executed by the processor to realize the steps of the method of any one of claims 1-7.

9. A computer-readable storage medium, characterized in that: The computer program in the medium is executed by the processor to realize the steps of the method of any one of claims 1-7.

Images