Portable automatic crimping device and automatic gripping method for aerial sputtering

Abstract

This application provides a portable automatic crimping device and automatic gripping method for aircraft connectors. The device includes: an automatic identification plug-in structure for identifying the wire core and the plug-in position, and carrying the wire core to the plug-in position to the connector socket of the connector crimping structure for insertion; and a connector crimping structure for crimping the connector to the end of the wire core placed inside the connector socket. This application provides an automatic crimping device for aircraft connectors including an automatic identification plug-in structure and a connector crimping structure, achieving accurate identification of the wire core and the plug-in position, and precisely inserting the wire core into the connector socket through automated operation, while simultaneously completing a high-quality crimping process. This solves the problem in existing automatic crimping devices for aircraft connectors and cable ends where inaccurate identification of the connector and cable leads to low crimping quality.

Description

Technical Field

[0001] This invention relates to the field of cable connector technology, and more specifically, to a portable automatic crimping device for connectors, an automatic gripping method for connectors, and a computer-readable storage medium. Background Technology

[0002] Aviation connectors (or simply aviation connectors) are a type of connector originating from the military industry, hence their name. As a key electromechanical component used to connect or disconnect circuits in electrical wiring, the electrical parameters of aviation connectors are crucial for their selection. Correct selection and use of aviation connectors not only ensures circuit reliability but also improves the overall performance of equipment. Aviation connectors, also known as plug sockets, are widely used in various fields such as industry and aerospace due to their diverse types and wide range of applications. Improving the reliability of aviation connectors is the primary responsibility of manufacturers.

[0003] In existing technologies, for example, patent 202410613302.X discloses a solderless crimping device and crimping method for aviation connectors. By achieving solderless technology, it avoids problems such as burns to the aviation connector positioning block, short circuits due to welding adhesion, and operational difficulties caused by welding in traditional processes. However, existing automatic crimping devices for aviation connectors and cable end cores suffer from insufficient accuracy in identifying connectors and cables during actual crimping, making it difficult to meet high standards of reliability and quality in the crimping process. This technological limitation significantly affects the widespread application of aviation connectors in high-precision scenarios and urgently requires improvement.

[0004] Currently, there are no other patents related to the automatic crimping technology for substation control cables. This patent utilizes a designed mechanical structure and an automatic identification and control device to automatically crimp control cables to the connectors, solving the standardization process problem of automatically connecting substation control cables to the connector terminals. Summary of the Invention

[0005] The main objective of this application is to provide a portable automatic crimping device for aircraft connectors, an automatic gripping method for aircraft connectors, and a computer-readable storage medium, so as to at least solve the problem that in the prior art, the automatic crimping device for aircraft connectors and cable end cores has inaccurate identification of connectors and cables during crimping, resulting in low quality of crimping process.

[0006] To achieve the above objectives, according to one aspect of this application, a portable automatic crimping device for aircraft connectors is provided. The device includes: an automatic identification crimping structure for identifying the wire core and the crimping position, and carrying the wire core to the crimping position to the position of the connector socket of the connector crimping structure and inserting it; and a connector crimping structure for crimping the connector to the end of the wire core placed in the connector socket.

[0007] Optionally, the automatic identification and insertion structure includes: an identification moving track; a sliding seat that engages with the identification moving track; a connecting rod fixed to the bottom of the sliding seat; a side rod connected to the bottom of the connecting rod; an identifier connected to the bottom of the side rod, the identifier having a camera mounted on it, the camera being used to identify the image and position of the fiber core, the connector socket, and the connector; a mechanical gripper movably connected to the bottom of the side rod; and a protective clamp connected to the bottom of the side rod for gripping the wire core and inserting it into the connector socket.

[0008] Optionally, the connector crimping structure includes: a connector moving track, placed parallel to the identification moving track; and a connector placement seat, slidably mounted inside the connector moving track, wherein the surface of the connector placement seat has a plurality of connector sockets for placing the connector.

[0009] Optionally, when the wire core is inserted into the connector socket, the single fiber core provides a longitudinal insertion force greater than or equal to 10N, an insertion depth of not less than 9mm, and when the fiber core is inserted, the total insertion and separation force is greater than or equal to 100N.

[0010] According to another aspect of this application, an automatic gripping method for an aircraft connector is provided. This automatic gripping method is applied to any of the aforementioned automatic crimping devices. The method includes: initially identifying the cable end core and connector socket, and moving the cable end core to the position of the connector socket; locating the connector based on a target recognition algorithm using accumulated quantized gradient direction features to obtain connector positioning; controlling the automatic identification and insertion structure to reposition the connector socket using an image feature positioning algorithm based on the connector positioning to obtain connector socket positioning; locating the gripping portion of the cable using a multi-view stereo vision positioning algorithm to obtain gripping portion positioning; repositioning the cable end core based on the connector socket positioning and the gripping portion positioning to obtain cable end core positioning; and performing an insertion operation based on the connector socket positioning and the cable end core positioning.

[0011] Optionally, a target recognition algorithm based on accumulated quantized gradient direction features is used to locate the connector, obtaining connector localization, including: recognizing a first target image and solving for the gradient components of each color channel of the first target image in the x and y directions using the Sobel operator to obtain a first gradient component and a second gradient component; calculating the gradient magnitude and gradient direction based on the first gradient component and the second gradient component of each color channel respectively, and determining the gradient direction corresponding to the color channel with the maximum gradient magnitude as the target gradient direction; obtaining a preset correction angle, and when the target gradient direction is negative, summing the preset correction angle and the target gradient direction, and updating the target gradient direction based on the summation result; determining the quantization interval to which the target gradient direction belongs based on the updated target gradient direction to obtain a target quantization interval, wherein the quantization interval is obtained by averaging a preset angle interval, and when any two target gradient directions differ by a first preset angle, the two target gradient directions are determined to be the same target gradient direction; and determining the quantization interval based on the boundary values ​​of the target quantization interval and the updated target gradient direction. The target gradient direction with the smaller difference is updated again, and the number corresponding to the target gradient direction is binary encoded to obtain a first target direction code. Based on the first target direction code, a first gradient direction feature map corresponding to the first target image is obtained. A reference image is acquired, and each pixel of the reference image undergoes multiple random rotation and translation transformations. The second direction code after each random rotation and translation transformation is recorded, and the second direction codes are integrated to obtain multiple second gradient direction feature maps corresponding to the reference image. The multiple second gradient direction feature maps are accumulated to obtain a third gradient direction feature map, and the cumulative number of times the third gradient direction feature map is accumulated is recorded to obtain the target number. The third gradient direction feature map is used as a window, and a sliding traversal is performed on the first gradient direction feature map. The score of the window at each position is calculated based on the objective function to obtain the target score. The position corresponding to the maximum target score is determined as the connector, and the connector is located based on its position in the first target image to obtain the connector location.

[0012] Optionally, controlling the automatic identification and insertion structure to reposition the connector socket based on the connector positioning using an image feature localization algorithm to obtain the connector socket positioning includes: determining the position of the connector in the first target image as the second target image; dividing the second target image into multiple sub-regions, calculating a histogram for the pixel grayscale of each sub-region, and calculating the average pixel grayscale of each sub-region; obtaining a cropping coefficient based on the average pixel grayscale, cropping the portion of the histogram greater than a first threshold based on the preset cropping coefficient, and distributing the cropped portion evenly across the grayscale levels of the histogram; and applying the histogram to the second target image. A third target image is obtained by bilinear interpolation of pixel gray levels in the target image; a fourth target image is obtained by normalizing each pixel row and / or each pixel column of the third target image; the fourth target image is scanned line by line based on horizontal scan lines to construct a position index-gray level histogram for a single row of pixels, where the horizontal axis of the position index-gray level histogram is the index of each pixel in the pixel row, and the vertical axis of the position index-gray level histogram is the gray level; a sliding window of a preset width is constructed and slides along the horizontal axis of the position index-gray level histogram to traverse the peaks in the position index-gray level histogram to obtain a first target set, wherein the width is an odd number and less than or equal to... The width of two adjacent peaks is used as the basis for determining a first target step size. Each peak in the first target set is traversed, and if the difference between the boundary pixels of a peak is greater than a second threshold, the peak is removed from the first target set to obtain a second target set. A preset number is obtained, and if the number of peaks in the second target set is equal to the preset number, the second target set is determined as an edge feature point set. The edge feature point sets corresponding to each pixel row are integrated to obtain a first edge feature set, and the fourth target image is scanned column by column based on a vertical scan line to determine the corresponding second edge feature set. A preset distribution is obtained. The arrangement of the connector holes on the connector is as follows: Based on the first edge feature set and the second edge feature set, they are recombined according to the preset distribution to obtain multiple third edge feature sets. Each third edge feature set corresponds one-to-one with the connector hole, and each third edge feature set includes multiple edges. A weighted average is calculated for each element in each third edge feature set to obtain the center point of each edge. The mean is calculated based on the center points corresponding to the horizontally adjacent edges and the vertically adjacent edges to obtain the first mean and the second mean. The average value is calculated based on the first mean and the second mean to obtain the pixel coordinates corresponding to the center of the connector hole.The spatial coordinates of the center of the connector socket are determined based on the pixel coordinates, thus obtaining the connector socket positioning.

[0013] Optionally, the grasping part of the cable is located using a multi-view stereo vision positioning algorithm to obtain the grasping part location, including: acquiring a fifth target image including the end core of the cable after grasping; counting the number of pixels in each white area of ​​the fifth target image to obtain the target number; deleting the white area if the target number is greater than a third threshold or less than a fourth threshold; cropping the fifth target image with the bounding rectangle of each white area as the boundary to obtain a sixth target image, and scanning it with horizontal scan lines to determine the single-sided boundary point of the white area in each sixth target image as the first sampling point; using the first sampling point as... Multiple rays are constructed from the starting point, wherein the angles of the rays are within a first preset range, and the angle difference between two adjacent rays is a second preset angle. A search is performed along each ray, and the intersection of the ray with the opposite boundary is determined as a second sampling point. The boundary distances between the first sampling points and each of the second sampling points are calculated to obtain multiple first target distances. If the first target distances are within the second preset range, the second sampling points are added to a third target set. A fourth target set is constructed based on all the first sampling points, and a fifth target set is constructed based on all the third target sets. The distances between the fourth and fifth target sets are then determined based on the number of elements in the fourth target set and the fifth target set. The ratio of the number of elements in the target set is calculated. If the ratio is greater than a fifth threshold, the region corresponding to the sixth target image is determined as a cable region, resulting in the first cable region. A seventh target image, including the cable end core, is obtained before capture, and this seventh target image is also marked with a cable region, resulting in the second cable region. Based on the first and second cable regions, the sixth and seventh target images are processed using the Bouguet stereo correction method to obtain the eighth and ninth target images. Gaussian filtering is applied to the eighth and ninth target images, and the filtered eighth target image is... The eighth and ninth target images are converted into single-channel grayscale images, and gradient direction codes are calculated using a target recognition algorithm to obtain corresponding gradient direction histograms. A neighborhood of a preset size is constructed in each gradient direction histogram, centered on the position of each edge point. The gradient direction codes in the neighborhood are statistically analyzed to obtain corresponding target histograms. Based on each target histogram, corresponding edge point sets are determined to obtain a sixth target set and a seventh target set. Feature point matching is performed between the sixth target set and the seventh target set, and successfully matched feature point pairs are stored in the eighth target set.Based on the eighth target set, the coordinates of all edge points in the mechanical gripper's base coordinate system are calculated to obtain the first coordinate value. The depth coordinates of the edge points are then calculated to obtain the second coordinate value. If the deviation between the second coordinate value and its average value is greater than a sixth threshold, the corresponding first coordinate value is deleted. The first coordinate value is then converted into spatial coordinates to obtain the third coordinate value. The direction vector of the cable is determined based on the third coordinate value to obtain the positioning of the gripping part.

[0014] Optionally, based on the connector socket positioning and the gripping part positioning, the cable end core is repositioned to obtain the cable end core positioning, including: when not gripping, controlling the mechanical gripper to move to a preset position, acquiring an image of the mechanical gripper, converting the image to a grayscale image, and determining the pixel region corresponding to the mechanical gripper in the grayscale image as a template image; matching the template image with a tenth target image including the gripped mechanical gripper, determining the position of the cable end core based on the mechanical gripper region in the tenth target image and the positional offset between the mechanical gripper region and the cable end core; correcting the cable end core position using the Bouguet stereo correction method, and extracting the pixel points corresponding to the cable end core and the cable body; filling the pixel points using a morphological opening and closing algorithm, and deleting discrete pixels. The eleventh target image is obtained by dividing the pixels into pixels. The portion of the eleventh target image corresponding to the cable end core is converted into an HSV channel to obtain the twelfth target image. By performing multi-channel threshold segmentation on the twelfth target image, the region corresponding to the cable body is obtained. The outer contour of the cable end core is determined based on the region corresponding to the cable body. Based on the outer contour, the minimum bounding rectangle of the outer contour and the centroid position of the minimum bounding rectangle are calculated using vertex chain code and rotation method. The slope of the line containing the longest side of the minimum bounding rectangle is calculated to obtain the target slope. A straight line passing through the centroid position is constructed using the target slope, and the intersection point of the straight line and the outer contour is determined. The center point and direction vector of the cable end core are determined based at least on the intersection point to obtain the target position and target vector, thus obtaining the positioning of the cable end core.

[0015] According to another aspect of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform any of the methods described.

[0016] Applying the technical solution of this application, the device includes: an automatic identification plug-in structure for identifying the wire core and the plug-in position, and carrying the wire core to the plug-in position to the connector socket of the connector crimping structure and inserting it; and a connector crimping structure for crimping the connector to the end of the wire core placed in the connector socket. This application provides an automatic crimping device for aviation connectors that includes an automatic identification plug-in structure and a connector crimping structure. It achieves accurate identification of the wire core and the plug-in position, and through automated operation, precisely inserts the wire core into the connector socket while completing a high-quality crimping process. This solves the problem in existing automatic crimping devices for aviation connectors and cable ends where inaccurate identification of the connector and cable leads to low crimping quality. Attached Figure Description

[0017] Figure 1 A schematic diagram of a portable automatic crimping device for an aircraft plug provided in an embodiment of this application is shown;

[0018] Figure 2 A schematic diagram of the automatic identification plug-in structure of a portable automatic crimping device for an aircraft plug provided according to an embodiment of this application is shown;

[0019] Figure 3 A schematic diagram of the connector crimping structure of a portable automatic crimping device for aircraft connectors provided according to an embodiment of this application is shown;

[0020] Figure 4 A flowchart illustrating an automatic acquisition method for an aerial plug according to an embodiment of this application is shown.

[0021] The above figures include the following reference numerals:

[0022] 1. Automatic identification of the plug-in structure; 2. Connector crimping structure; 101. Identification moving track; 102. Sliding seat; 103. Connecting rod; 104. Side rod; 105. Identifier; 106. Mechanical claw; 107. Protective clamp; 201. Connector moving track; 202. Connector placement seat; 203. Connector socket. Detailed Implementation

[0023] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0024] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] For ease of description, the following explains some of the nouns or terms used in the embodiments of this application:

[0027] Bouguet stereo rectification is based on Jean-Yves Bouguet's camera calibration toolkit and focuses on the rectification of stereo cameras. The goal of stereo rectification is to transform a pair of misaligned camera views into coplanar, row-aligned views, simplifying stereo matching algorithms and thus accurately estimating depth information. Bouguet stereo rectification transforms the camera's calibration parameters and image points to a common plane, ensuring that corresponding points in the two views are aligned only along the horizontal line. This simplifies stereo matching algorithms, reducing the search from two dimensions to one.

[0028] The Sobel operator is a classic image processing tool used for edge detection. It calculates local gradients in an image by applying filter kernels, thereby detecting regions with significant changes in grayscale values; these regions typically correspond to edges in the image. Due to its simplicity and efficiency, the Sobel operator is widely used in image processing and computer vision.

[0029] HSV channels refer to the three components used to describe color in a color space: Hue, Saturation, and Value, representing the color type, purity, and brightness, respectively. Hue defines the basic attributes of a color, such as red, green, and blue, and is usually expressed as a degree; Saturation describes the vividness of a color, gradually increasing from gray to pure colors; and Value reflects the lightness or darkness of a color, increasing from black to the brightest color. The separability of HSV channels makes them particularly suitable for image processing and visual tasks, facilitating precise color analysis and manipulation.

[0030] As described in the background section, existing automatic crimping devices for aircraft connectors and cable end cores suffer from insufficient accuracy in identifying connectors and cables during the actual crimping process. This results in the reliability and quality of the crimping process failing to meet high standards. To address the problem of inaccurate connector and cable identification in existing automatic crimping devices for aircraft connectors and cable end cores, leading to low crimping quality, embodiments of this application provide a portable automatic crimping device for aircraft connectors, an automatic gripping method for aircraft connectors, and a computer-readable storage medium.

[0031] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0032] This application provides a portable automatic crimping device for aircraft connectors. Figure 1 This is a schematic diagram of the structure of a portable automatic crimping device for aircraft inserts according to an embodiment of this application, as shown below. Figure 1 As shown, the device includes:

[0033] Automatic identification plug-in structure 1 is used to identify the wire core and the plug-in position respectively, and carry the wire core to the plug-in position to the position of the connector socket 203 of the connector crimping structure 2 and insert it;

[0034] Specifically, the automatic identification and insertion structure 1 is one of the core components of this device. It can automatically identify the characteristics (such as diameter, material, length, etc.) of the wire core to be processed and the target insertion position through vision sensors, position sensors, or other identifiers. After grasping the cable, the position of the terminal relative to the tool coordinate system needs to be detected to determine the validity of the grasp and whether insertion is allowed. This allows for more precise control of the wire core movement and insertion position, enabling the wire core to be accurately inserted into the connector. Finally, crimping is performed, resulting in higher crimping efficiency between the connector and the wire core.

[0035] Connector crimping structure 2 is used to crimp the connector to one end of the wire core placed inside the connector socket 203.

[0036] Specifically, the main function of the connector crimping structure 2 is to place the wire core into the connector socket 203 and then firmly connect the wire core and the socket by controlling the crimping action of pressure and position.

[0037] This embodiment of the application, through precise identification of the connector and cable images and positions, combined with cable gripping and end positioning, delivers the cable end to the connector insertion position and further identifies the specific posture of the wire core, thereby ensuring that the wire core is completely positioned in the connector connection position during the crimping process. This method effectively improves the accuracy of the insertion process, ensures reliable contact between the wire core and the connector, and improves the crimping quality. Simultaneously, the automated insertion and crimping process significantly improves operational efficiency, providing technical assurance and comprehensive support for efficient and precise automated operation of aircraft connectors. It solves the problem in existing automatic crimping devices for aircraft connectors and cable ends where inaccurate identification of the connector and cable leads to low crimping quality.

[0038] As one possible implementation method, Figure 2 This is a schematic diagram of the automatic identification and insertion structure 1 of the automatic crimping device according to an embodiment of this application, as shown below. Figure 2 As shown, the automatic identification and insertion structure 11 includes: an identification moving track 101; a sliding seat 10, which engages with the identification moving track 101; a connecting rod 103, fixed to the bottom of the sliding seat 102; a side rod 104, connected to the bottom of the connecting rod 103; an identifier 105, connected to the bottom of the side rod 10, on which a camera is mounted, which is used to identify the image and position of the fiber core, the connector socket 203, and the connector; a mechanical claw 106, movably connected to the bottom of the side rod 104; and a protective clamp 107, connected to the bottom of the side rod 104, used to clamp the wire core and insert it into the connector socket 203.

[0039] Specifically, a sliding seat 102 is slidably engaged on the inner side of the identification moving track 101. A connecting rod 103 is fixed at the bottom of the sliding seat 102. A side rod 104 is installed at the bottom of the connecting rod 103. An identifier 105 is installed at the bottom of the side rod 104. Two mechanical claws 106 are movably connected to the bottom of the side rod 104. A protective clamp 107 is connected to the bottom of the mechanical claws 106.

[0040] Therefore, by setting the identifier 105 to identify the image and position of the connector, the image and position of the connector can be quickly obtained, and the cable core can be identified. The core and connector can be accurately identified and positioned. Then, the mechanical claw 106 and the protective clamp 107 clamp the cable, align it with the position of the connector socket 203 and send it in, so as to realize the preparation of the cable core and the connector for insertion, making the insertion efficiency higher.

[0041] As one possible implementation method, Figure 3This is a schematic diagram of the connector crimping structure 2 of a portable automatic crimping device for aircraft connectors according to an embodiment of this application, as shown below. Figure 3 As shown, the connector crimping structure 2 includes: a connector moving track 201, which is placed parallel to the identification moving track 101; and a connector placement seat 202, which is slidably installed inside the connector moving track 201. The surface of the connector placement seat 202 is provided with a plurality of connector sockets 203 for placing connectors.

[0042] Specifically, the connector moving track 201 is made of high-strength, low-friction track material and is placed parallel to the identification moving track 101 to support the connector placement seat 202 and achieve precise sliding. The parallel placement design of the track optimizes the wire core movement path and the mating angle of the connector, thereby ensuring smooth crimping. The connector placement seat 202 is slidably connected to the inside of the connector moving track 201 via a precision slider, and its surface is designed with multiple connector sockets 203 for fixing the connector to be crimped. The placement seat has a modular design, which can be replaced and adjusted according to different connector specifications, improving the adaptability of the device.

[0043] As one possible implementation, when the fiber core is inserted into the connector socket 203, the single fiber core provides a longitudinal insertion force greater than or equal to 10N, the insertion depth is not less than 9mm, and the total insertion and separation force is greater than or equal to 100N when the fiber core insertion is completed.

[0044] Specifically, during the insertion of the wire core into the connector socket 203, the single-core insertion design requires a longitudinal insertion force of no less than 10N to ensure that contact resistance is overcome and a firm initial contact is achieved during insertion. Simultaneously, the insertion depth is designed to be no less than 9mm to ensure that the core fully enters the conductive area of ​​the connector, achieving optimal conductivity and mechanical stability. After insertion, the total insertion separation force is greater than or equal to 100N, indicating that the connection between the wire core and the connector remains stable even under strong vibration or tension, meeting the requirements of harsh industrial or aerospace environments.

[0045] In the above embodiments, the portable automatic crimping device is designed to address situations where the cable crimping position is not fixed in the application scenario. The portable automatic crimping device is small in size and easy to move according to the requirements.

[0046] This embodiment provides an automatic gripping method for a portable automatic crimping device that operates in any form. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0047] Figure 4 This is a flowchart of an automatic packet capture method according to an embodiment of this application. Figure 4 As shown, the method includes the following steps:

[0048] Step S401: Initially identify the cable end core and connector socket, and move the cable end core to the position of connector socket 203;

[0049] Specifically, the identifier of the automatic identification and insertion structure 1 performs preliminary identification of the cable end core and connector socket. By acquiring images through a camera, the automatic connector identification and insertion algorithm primarily achieves connector target identification and connector socket positioning, as well as cable end detection and spatial pose estimation based on multi-view stereo vision technology. Subsequently, it controls the cable end core to move closer to the connector socket, laying the foundation for subsequent precise positioning and insertion operations.

[0050] Step S402: Based on the target recognition algorithm of accumulated quantized gradient direction features, the connector is located to obtain the connector location;

[0051] Specifically, the connector recognition algorithm employs a target recognition method based on Histogram of Oriented Gradients (HOG) to accurately locate the connector as a whole. By encoding and accumulating the gradient directions of the image, the edge and shape features of the connector are extracted to generate the connector's location information, ensuring accurate identification of the connector even in complex backgrounds. This method is more robust to object recognition in cluttered backgrounds and can achieve real-time detection.

[0052] Step S403: The automatic identification and plug-in structure 1 uses an image feature positioning algorithm to reposition the connector socket based on the connector positioning, thereby obtaining the connector socket positioning.

[0053] Specifically, based on the overall positioning of the connector, image feature localization algorithms (such as template matching and deep learning) are further used to accurately identify the specific location of the connector socket. The algorithm generates high-precision three-dimensional coordinates of the socket by analyzing image feature points in the socket area, such as edges and color differences, and provides them to the robotic gripper for docking operations.

[0054] Step S404: The gripping part of the cable is located by a multi-view stereo vision positioning algorithm to obtain the gripping part location.

[0055] Specifically, by using multi-view stereo vision technology, image data of the cable is acquired from different angles to reconstruct the three-dimensional structure of the cable gripping area, and its spatial position and orientation are accurately calculated. This method effectively solves the gripping difficulties caused by cable bends or complex shapes, ensuring that the robotic gripper can reliably grasp the cable.

[0056] Step S405: Based on the connector socket positioning and the gripping part positioning, the cable end core is repositioned to obtain the cable end core positioning. The insertion operation is then performed based on the connector socket positioning and the cable end core positioning.

[0057] Specifically, based on the positioning of the connector socket and the cable gripping part, the cable end core is precisely positioned again to ensure that the core is perfectly aligned with the socket. By combining the high-precision position data from both methods, the mechanical gripper is controlled to complete the insertion operation, achieving precise insertion of the cable core.

[0058] This embodiment of the application achieves efficient and automated grasping and insertion of cable end cores and connector sockets. Preliminary identification provides rapid initial positioning, while the target recognition algorithm based on accumulated quantized gradient direction features and multi-view stereo vision technology further improves the positioning accuracy of the connector and cable grasping parts. Image feature positioning algorithms ensure precise alignment of the sockets and cores, and the multi-step collaborative process effectively solves problems such as cable bending and complex backgrounds, thereby significantly improving the automation, accuracy, and reliability of the insertion operation, providing an efficient solution for complex cable connection scenarios.

[0059] As one possible implementation, a target recognition algorithm based on accumulated quantized gradient direction features is used to locate the connector, resulting in connector localization, including:

[0060] Step S501: Based on the first target image, perform recognition and solve the gradient components of each color channel of the first target image in the x and y directions based on the Sobel operator to obtain the first gradient component and the second gradient component.

[0061] Specifically, by applying the Sobel operator to the first target image, the gradient components of each color channel in the x and y directions are calculated to obtain the first gradient component and the second gradient component. This step utilizes the edge enhancement properties of the Sobel operator to extract regions with significant gradient changes in the image, thus initially capturing the boundary features of the connector.

[0062] Step S502: Calculate the gradient magnitude and gradient direction based on the first gradient component and the second gradient component of each color channel, and determine the gradient direction corresponding to the color channel with the maximum gradient magnitude as the target gradient direction.

[0063] Specifically, the formulas for gradient magnitude and gradient direction are as follows: Where Mc represents the gradient magnitude of each color channel; Ac represents the gradient direction of each color channel, with an angle range of (-π, π]. Then, the gradient magnitudes of the three channels are compared, and the color channel with the largest magnitude is selected. Its gradient direction will be used as the target gradient direction, i.e.: Where x represents the position in the image; Mc represents one of the three color channels of the original image (RGB); P(x) is the set of image positions corresponding to the maximum gradient values ​​in the three channels; orient(x) represents the gradient direction corresponding to position x; and Ao(x) is the gradient direction feature map corresponding to the maximum gradient value.

[0064] Step S503: Obtain the preset correction angle. When the target gradient direction is negative, sum the preset correction angle and the target gradient direction, and update the target gradient direction based on the summation result.

[0065] Specifically, the negative gradient direction is corrected by introducing a preset correction angle. For example, for A... o Adding 2π to the negative angle in (x) adjusts the angle range to [0, 2π). This ensures that all gradient direction values ​​are within a uniform positive range, providing consistent input for subsequent quantization steps.

[0066] Step S504: Determine the quantization interval to which the target gradient direction belongs based on the updated target gradient direction to obtain the target quantization interval. The quantization interval is obtained by dividing the preset angle interval by an average. If any two target gradient directions differ by a first preset angle, the two target gradient directions are determined to be the same target gradient direction.

[0067] Specifically, the gradient direction is uniformly quantized according to a preset angle range. By dividing and classifying the updated target gradient direction into quantization intervals, the impact of subtle perturbations in the gradient direction on recognition is reduced. For example, the direction angle [0, 2π) interval is divided into 2n smaller intervals. To further reduce the computational load and ensure that features are not lost, the gradient direction angles α and (α ± π) are regarded as the same gradient direction. Finally, the gradient direction can be divided into n quantization intervals.

[0068] Step S505: Based on the boundary value of the target quantization interval that has the smallest difference from the updated target gradient direction, update the target gradient direction again, and encode the number corresponding to the target gradient direction into binary code to obtain the first target direction code;

[0069] Specifically, after calculating the final gradient direction, the gradient direction value is quantized into the number corresponding to the interval boundary value closest to the direction value according to the quantization interval it belongs to. Then, according to the sequential binary encoding rule, the quantized gradient direction is converted into the corresponding n-bit binary direction code, providing the basis for subsequent feature integration.

[0070] Step S506: Integrate based on the first target direction code to obtain the first gradient direction feature map corresponding to the first target image;

[0071] Specifically, after quantization, the gradient direction feature map of the first target image is generated by integrating the direction codes of all pixels, which initially represents the edge direction distribution features in the image and provides input for subsequent matching.

[0072] Step S507: Obtain the reference image, perform multiple random rotation and translation transformations on each pixel of the reference image, record the second direction code after each random rotation and translation transformation, and integrate the second direction codes to obtain multiple second gradient direction feature maps corresponding to the reference image.

[0073] Specifically, the cumulative quantized gradient direction features of the reference image are then extracted. To improve the robustness of recognition and template matching, more noise-resistant feature information needs to be obtained from the reference image. Based on the idea of ​​geometric fuzziness and combined with the method of image affine transformation, each pixel of the reference image undergoes multiple random rotation and translation transformations within a small range. The pixel transformation matrix is ​​as follows: Where θ is the angle of rotation of the pixel around the center; t x t is the pixel translation amount in the x-direction; y The pixel translation amount in the y-direction; select t x and t y The translation range is [0,1] pixels, and the rotation angle θ ranges from [0,π / 9]. The gradient direction is calculated for the image obtained after each rotation and translation transformation. The gradient direction quantization method is then used to quantize the gradient direction features of the template after each transformation, generating a second direction code. Finally, n images with quantized gradient direction features are obtained. Then, a bitwise OR operation of the second direction code is performed on the same position in each feature image, and multiple second gradient direction feature maps are generated, i.e.: In the formula, ori j (x,y) represents the gradient direction code at position (x,y) in the j-th feature image; CO j (x,y) represents the cumulative direction code after adding the j-th feature map; n is the number of images with quantized gradient direction features.

[0074] Step S508: Accumulate multiple second gradient direction feature maps to obtain a third gradient direction feature map, and record the number of times the third gradient direction feature map is accumulated to obtain the target number;

[0075] Specifically, accumulating the quantized gradient direction feature image (second gradient direction feature map) generates the final accumulated quantized direction feature map (third gradient direction feature map). Then, the feature set transformed from the prepared reference image is accumulated, and the number of times the feature is accumulated at each location in the accumulated quantized direction feature map needs to be counted. In the formula, k j (x,y) indicates whether there is a quantization direction code at pixel position (x,y) in each quantization gradient direction feature map. The value is 0 or 1, and the cumulative feature frequency of the reference image can be obtained.

[0076] Step S509: Use the third gradient direction feature map as a window, slide through the first gradient direction feature map, and calculate the score of the window at each position based on the objective function to obtain the target score.

[0077] Specifically, after extracting the quantized gradient direction features of the target image and the cumulative direction features of multiple reference images, reference matching is required to identify the connector and determine its position. The created reference image is used as a sliding window, which is traversed across the acquired connector image. By constructing the corresponding objective function, the score of the sliding window at each position is obtained, i.e.: Where H is the height of the window, i.e., the height of the reference image; W is the width of the window, i.e., the width of the reference image; φ represents the number of accumulated features at the corresponding position when the accumulated quantization gradient direction features of the reference image contain the quantization direction features extracted from the target image. The attribution relationship can be determined by bitwise AND operation, i.e., when ori j ∈ori T When, satisfying ori j ∧ori T >0, therefore we can obtain:

[0078] Step S5010: The position corresponding to the maximum target score is determined as the connector, and the connector is located based on its position in the first target image to obtain the connector location.

[0079] Specifically, by combining the above formulas, a baseline matching result is obtained. The position with the highest score in the result is selected, and the actual position of the connector is determined by the position corresponding to the maximum target score. Finally, connector localization based on gradient direction features is completed, serving as the final identification and coarse localization result. This step provides accurate connector coordinate information for automatic grasping and crimping operations.

[0080] According to the above technical solution, when recognizing the image of the connector socket 203, the target recognition method based on the accumulated quantization gradient direction features is used to identify the connector, and the approximate area of ​​it in the original image and its angular deviation from the horizontal direction of the image are determined. However, for the plugging task, it is still necessary to use the socket image features to perform a positioning algorithm to further accurately locate the position of the socket.

[0081] As one possible implementation, the connector socket 203 positioning algorithm first uses a contrast-limited adaptive histogram equalization method to preprocess the identified connector area to improve the uneven brightness distribution at the socket edge. The principle of this method is mainly to limit the contrast amplitude after local histogram equalization, thereby improving brightness uniformity while suppressing excessive noise amplification. The automatic identification of the plug-in structure 1 then uses an image feature positioning algorithm to reposition the connector socket 203 based on the connector positioning, resulting in the connector socket 203 positioning, including:

[0082] Step S601: Determine the position of the connector in the first target image as the second target image;

[0083] Specifically, the corresponding position of the connector in the first target image is cropped to generate a second target image for further analysis.

[0084] Step S602: Divide the second target image into multiple sub-regions, calculate the histogram of the pixel grayscale of each sub-region, and calculate the average pixel grayscale of each sub-region.

[0085] Specifically, the image is divided into several sub-regions of equal size, which are continuous and non-overlapping. Histogram statistics are performed on the pixel grayscale of each sub-region, and the average grayscale value of each sub-region is calculated to provide a basis for subsequent cropping and adjustment.

[0086] Step S603: Obtain the cropping coefficient based on the average pixel gray level, crop the portion of the histogram that is greater than the first threshold based on the preset cropping coefficient, and distribute the cropped portion evenly to each gray level of the histogram.

[0087] Specifically, the cropping coefficient is determined by the average pixel gray value of the sub-region, the portion of the histogram that is greater than the first threshold is cropped, and the cropped portion is evenly distributed to each gray level of the histogram to smooth the gray level distribution and reduce noise interference.

[0088] Step S604: Perform bilinear interpolation of pixel gray levels on the second target image based on the histogram to obtain the third target image;

[0089] Specifically, bilinear interpolation of gray levels is performed, that is, linear interpolation is performed in the horizontal and vertical directions of the image to overcome the block effect of gray level discontinuity in adjacent areas, generate a smoother third target image, and create conditions for subsequent edge feature extraction.

[0090] By preprocessing the identified connector region using a restricted contrast adaptive histogram equalization method, the uneven brightness distribution at the edge of the connector socket 203 is improved, while the grayscale of the corresponding pixels inside the connector socket 203 remains almost unchanged. The contrast between the edge and the inside of the connector socket 203 in the image is increased, reducing the difficulty of extracting the edge of the connector socket 203 in the subsequent process. After obtaining the connector region and angle deviation through template matching of accumulated directional features, the angle of the original image is corrected, and then the connector region is cropped and subjected to the aforementioned restricted contrast histogram equalization process to obtain the corrected image.

[0091] Next, the edge features of connector socket 203 are extracted. According to the preprocessing results, by limiting the contrast histogram equalization of the local grayscale of the image, the edge of connector socket 203 on the connector mating surface is basically in a bright state. From the whole image, the bright part is distributed in a grid pattern, and the grayscale distribution on both sides of the socket edge is roughly the same. From the perspective of a single row of pixels, high grayscale pixels appear periodically. When mapped to the grayscale histogram composed of a single row of pixels, it is represented by multiple local peaks. Multiple peaks correspond to multiple socket edges on the connector, and the histograms near the peaks belonging to the socket edges are basically symmetrically distributed. The algorithm adopts the strategy of scanning from top to bottom using horizontal scan lines and scanning from left to right using vertical scan lines, and then filtering and extracting edge features through the constructed scan line histogram.

[0092] Step S605: Normalize each pixel row and / or each pixel column of the third target image to obtain the fourth target image;

[0093] Specifically, to further reduce the impact of uneven image brightness and facilitate data manipulation, it is necessary to normalize the data of each pixel row and / or pixel column of the third target image separately to obtain the fourth target image, ensuring consistent data distribution. That is: Where N is the normalized image; I is the original image; α is the set lower limit value of the normalized image; β is the set upper limit value of the normalized image; α is 0 and β is 1, which means that the pixel is transformed from the original gray level 0-255 to the range of 0-1.

[0094] Step S606: Scan the fourth target image line by line based on the horizontal scan line to construct a position number-gray level histogram of a single row of pixels. The horizontal axis of the position number-gray level histogram is the number of each pixel in the pixel row, and the vertical axis of the position number-gray level histogram is the gray level.

[0095] Specifically, the image is scanned line by line from top to bottom using horizontal scan lines to construct a histogram of pixel position indices and gray levels for each row. The horizontal axis of this histogram corresponds to the index of each element in the pixel row, and the vertical axis represents the pixel gray level. Normalization is then performed on this histogram, and finally, peaks are extracted and filtered. Let x be the position of the k-th element in a given pixel row. k Its grayscale value is f(x) k If the extraction and filtering of local peaks in the histogram are as shown in steps S607 to S6010, then the specific steps are as follows.

[0096] Step S607: Construct a sliding window of preset width, slide it in the horizontal direction of the position number-grayscale histogram, traverse the peaks in the position number-grayscale histogram, and obtain the first target set, wherein the width is an odd number and is less than or equal to the width of two adjacent peaks.

[0097] Specifically, a window of width W is constructed and slides along the horizontal axis of the histogram, where the width is an odd number and satisfies 3≤W≤δ, where δ is the width of two adjacent peaks.

[0098] Step S608: Determine the first target step size based on the preset width, traverse each peak in the first target set, and delete the peak from the first target set if the peak's boundary pixel difference is greater than the second threshold, thus obtaining the second target set.

[0099] Specifically, the sliding window moves at x k Centered on the element, slide one pixel at a time to traverse the other pixels within the window. If f(x) k The value is greater than the set threshold (second threshold), and satisfies: Then x k The elements at the given positions are assigned to the first target set P. Considering that the grayscale distribution on both sides of the edge of connector socket 203 is roughly the same, all elements p in P are traversed. Within the sliding window centered on p, if the following conditions are met: Where t is a set threshold; if the condition is met, element p is retained; otherwise, it is removed from the first target set P. For adjacent elements p in P... i and p i+1 The sequence numbers are compared. If the difference is too small, it means the distance between adjacent peaks is too small, and the smaller value needs to be deleted. Therefore, peaks that do not meet the requirements are deleted from the first target set to form the second target set.

[0100] Step S609: Obtain a preset number. If the number of peaks in the second target set is equal to the preset number, determine the second target set as the edge feature point set.

[0101] Specifically, based on the prior information about the number of connector sockets 203, the number of elements in the second target set is counted. If the number of peaks after final screening is the same as the number of edges of connector sockets 203, then the second target set is used as the final set of edge feature points extracted for this horizontal scan line. Otherwise, it is determined to be an invalid set of peaks, and the edge feature extraction of the next horizontal scan line is performed.

[0102] Step S6010: Integrate the edge feature point sets corresponding to each pixel row to obtain the first edge feature set, and scan the fourth target image column by column based on the vertical scan line to determine the corresponding second edge feature set.

[0103] Specifically, by applying the above steps to each horizontal scan line, the set of edge features L = {P} in the horizontal direction of the connector socket 203 can be obtained. i |i=1,…N}, where N is the number of effective peak sets on the horizontal scan line, i.e., the number of holes in the horizontal direction. For the vertical scan line, the same method is used to extract the edge point features V={P} of the hole in the vertical direction. i |i=1,…M}, where M is the number of effective peak sets of the vertical scan line, that is, the number of edges of the connector socket 203 in the vertical direction.

[0104] Step S6011: Obtain the preset distribution, which is the arrangement of connector sockets 203 on the connector;

[0105] Step S6012: Based on the first edge feature set and the second edge feature set, they are recombined according to a preset distribution to obtain multiple third edge feature sets. Each third edge feature set corresponds one-to-one with the connector socket 203, and each third edge feature set includes multiple edges.

[0106] Specifically, the next step is to locate the center position of the connector socket 203. Based on the connector structure and the distribution of the connector socket 203, we can know the number of connector sockets 203 on the connector and the pattern of their grid arrangement. This prior information is helpful in determining the center position of the connector socket 203. By extracting the features of the connector socket 203, we can obtain the horizontal edge feature set L and the vertical edge feature set V. Next, we need to classify and reorganize these feature sets according to the position of the edges.

[0107] Assuming the number of rows of connector sockets 203 on a single connector is M+1 and the number of columns is N+1, then the number of horizontal edges of connector sockets 203 is N, and the number of vertical edges is M. That is, excluding the outermost edge of the hole, the edges extracted by the horizontal scan lines are called horizontal edges, and the edges extracted by the vertical scan lines are called vertical edges. j Let represent the j-th element of each unit li in the horizontal edge feature set L, that is, a point on the j-th horizontal edge, while vi j This represents the j-th element of each cell vi in ​​set V, that is, a point on the j-th vertical edge, such as... Figure 3 As shown, in order to obtain the center position of each connector socket 203, it is necessary to reorganize each element in sets L and V to obtain all feature points of the edge of each connector socket 203. The specific steps are as follows: Step 1: Extract all feature points of the nth horizontal edge from L, and reassemble the set as follows. Similarly, extract all feature points of the m-th vertical edge from V and reassemble the set as follows: Step 2: For Fit a straight line y to all edge points m =f m (x; b1, b2), this straight line can distinguish the edge points of the upper and lower rows of connector sockets, that is... The elements can be divided into upper and lower parts. If a point is above the line, it is assigned to the edge set S of the upper connector socket. N·m+n Otherwise, it is assigned to the lower edge set S. N(m+1)+n Step 3: Reorganize and classify all horizontal and vertical edges using the steps described above, ultimately obtaining the set of edge points {S} belonging to all connector sockets 203. k |k=1,…,MN}.

[0108] Step S6013: Perform a weighted average calculation on each element in each third edge feature set to obtain the center point of each edge. Solve the mean value based on the center points of the horizontally adjacent edges and the vertically adjacent edges respectively to obtain the first mean value and the second mean value. Calculate the average value based on the first mean value and the second mean value to obtain the pixel coordinates corresponding to the center of the connector socket 203.

[0109] Specifically, for each connector socket 203 edge point set S obtained by the above reorganization k Each point s in i To calculate the weighted average, i.e.: Where n is S k The number of elements is calculated; the coordinates of the midpoint of each socket edge are obtained; and finally, the average of the coordinates of two adjacent midpoints in the horizontal direction is calculated to obtain the pixel coordinates of the center of the connector socket 203 in the image.

[0110] Step S6014: Determine the spatial coordinates of the center of the connector socket 203 based on the pixel coordinates to obtain the positioning of the connector socket 203.

[0111] Specifically, the pixel coordinates of the center of the socket are converted into spatial coordinates to obtain the precise positioning result of the connector socket 203.

[0112] Therefore, by acquiring a single image of the connector, utilizing algorithms for identifying and locating connector sockets 203, and combining this with the calibration results of the motion platform, the spatial position of the center of each connector socket 203 on the connector can be obtained. This method achieves high-precision positioning of the connector sockets, effectively filtering noise, correcting grayscale anomalies, extracting edge features, and accurately calculating the spatial coordinates of the socket center. This process combines various image processing techniques such as grayscale histogram processing, sliding window filtering, and feature point integration, resulting in high positioning accuracy and low error. This provides reliable support for subsequent automatic crimping and insertion operations, significantly improving the operational efficiency and stability of automated aerial splicing equipment.

[0113] According to the above technical solution, when the recognizer identifies the wire core image, after obtaining the precise position of the connector socket 203, in order to realize the wire hole assembly, it is also necessary to obtain the pose information of the wire core. Before extracting the cable feature points, it is necessary to obtain the area of ​​the cable in the image. Since the cable diameter is not affected by deformation and the diameter of each position on the cable is almost the same, the projection onto the image is a strip-shaped area with a basically unchanged width. By using the shape features combined with the area size, the part belonging to the cable can be filtered out from all areas.

[0114] As one possible implementation, a multi-view stereo vision positioning algorithm is used to locate the gripping part of the cable, resulting in the gripping part location, including:

[0115] Step S701: Obtain the captured fifth target image including the wire core at the end of the cable; count the number of pixels in each white area of ​​the fifth target image to obtain the target quantity.

[0116] Step S702: If the number of targets is greater than the third threshold or less than the fourth threshold, delete the white area;

[0117] Specifically, the number of pixels in each white region within the fifth target image is calculated, which is the region area A. i According to condition t I ≤A i ≤t u , where t I and t u These are the upper and lower limits for area filtering. If the limits are met, the corresponding image region is retained; otherwise, it is discarded.

[0118] Step S703: The fifth target image is cropped using the bounding rectangle of each white area as the boundary to obtain the sixth target image. The image is then scanned using horizontal scan lines to determine the single-sided boundary point of the white area in each sixth target image as the first sampling point.

[0119] Specifically, firstly, image blocks are extracted using the bounding rectangle of each retained white area as the boundary. For a single image block, a horizontal scan line is used to scan from top to bottom, searching for the left boundary point of the region within the image block. As the sampling point (i is the sampling point number), that is, the first sampling point.

[0120] Step S704: Construct multiple rays starting from the first sampling point, wherein the angle of the rays is within a first preset range, and the angle difference between two adjacent rays is a second preset angle.

[0121] Specifically, with Construct n rays starting from the origin, each ray having an angle θ. j ∈[-λ,λ], where j is an integer in [1,n], λ is the set upper limit of the angle, and λ = π / 4 is taken. The angle difference between two adjacent rays is the set value Δλ.

[0122] Step S705: Search along each ray and determine the intersection of the ray and the opposite boundary as the second sampling point;

[0123] Specifically, the search then proceeds along the ray to find the intersection of the ray and the right boundary of the region. That is, the second sampling point.

[0124] Step S706: Calculate the boundary distance between the first sampling point and each second sampling point to obtain multiple first target distances. If the first target distance is within a second preset range, add the second sampling point to the third target set.

[0125] Specifically, calculation The distance is calculated for each ray using the same method, and the minimum width of the region is found from this. Satisfying s∈[w min w max ], that is, the minimum distance (distance to the first target) is within the set range [w min w max If the sampling point is positive, it is assigned to the point set P (the third target set); otherwise, the next sampling point is tested.

[0126] Step S707: Construct a fourth target set based on all the first sampling points, construct a fifth target set based on all the third target sets, calculate the ratio of the number of elements in the fourth target set to the number of elements in the fifth target set, and if the ratio is greater than the fifth threshold, determine the region corresponding to the sixth target image as the cable region to obtain the first cable region;

[0127] Specifically, to speed up the calculation, select and Left boundary point of the region differing by h rows of pixels As the next sampling point, repeat the above steps to obtain the final point set P. Count the total number of sampling points m and the total number of point sets P m. P Calculate the ratio of the two. like e is a set parameter, here it is set to 0.8, then the area is determined to be a cable area, otherwise the area is discarded;

[0128] Step S708: Obtain the seventh target image, including the wire core at the end of the cable, before grabbing, and mark the cable area on the seventh target image to obtain the second cable area.

[0129] Specifically, the above method is used to process the two images taken before the cable was captured, and the area belonging to the cable in each image can be obtained.

[0130] Step S709: Based on the first cable region and the second cable region, the sixth target image and the seventh target image are processed using the Bouguet stereo correction method to obtain the eighth target image and the ninth target image;

[0131] Specifically, cable grabbing requires knowledge of the cable's spatial location, which can be described by the spatial coordinates of some feature points at the grabbing location. For cables lacking texture, edge points are arguably the most prominent feature points. Therefore, the cable's pose before grabbing can be calculated by determining the 3D coordinates of edge points near the grabbing location. Calculating the 3D coordinates of spatial points requires two or more images. Feature matching is used to find the projection positions of points in different images, and then constraint equations are used to solve the problem. To improve matching speed and reduce mismatches, the Bouguet stereo correction method is used to correct two images captured by the camera at two different locations.

[0132] The Bouguet stereo correction method transforms two non-coplanar images using camera calibration parameters to obtain two coplanar and aligned images. This allows for the conversion of two-dimensional search into one-dimensional search during image matching, greatly improving matching speed. The device moves the camera to two different positions to acquire cable images, which can be regarded as a binocular imaging model. Since the image planes corresponding to the two imaging positions are difficult to guarantee to be coplanar, the Bouguet stereo correction method is used to correct the images before feature extraction and matching.

[0133] According to the Bouguet stereo correction principle, it is necessary to obtain the transformation matrix of the camera coordinate system at the second imaging position relative to the first imaging position. From this, we can conclude that; in, This is the transformation matrix from the tool coordinate system to the camera coordinate system; This represents the pose matrix for two positions in the tool coordinate system; it can be directly obtained by reading data from the controller. During stereo correction, to reduce reprojection distortion, the image planes at each of the two imaging positions are rotated by half, thus reducing the matrix... rotational component Decomposed into two rotation matrices r l and r r And satisfy At this point, the two rotated image planes are parallel but their rows are misaligned. Therefore, further rotation correction is needed to align the epipolar lines of the two image planes. Let the correction rotation matrix be: Among them, e1, e2, and e3 are mutually orthogonal, and vector e1 is selected as the translation vector of the camera coordinate system between the two positions. After normalization, it becomes: in, We choose vector e2 as the normal vector of the plane formed by e1 and the optical axis, that is: Therefore: e3 = e1 × e2, thus, the correction matrix of the image plane at the two positions can be finally obtained: According to the Bouguet stereo correction algorithm, the relationship between the camera coordinate system and the pixel coordinate system at the first corrected position can be obtained as follows: Where d is the disparity, which can be calculated by matching points in the two images after correcting the images obtained at the first and second positions, and Q is the reprojection matrix, satisfying: Among them, T x c is the baseline distance. x and c yHere, x and y are the x and y coordinates of the principal point of the image acquired at the first location, respectively, and cx is the x-coordinate of the principal point of the image acquired at the second location. Applying the above stereo correction method to the original cable image and the binary image obtained from cable region segmentation and filtering, the corrected images are aligned, which significantly reduces the search range during matching.

[0134] For the binary image of the cable region after stereo correction, edge extraction is performed based on the contour of the region to obtain the edge point set of all cable bodies. Next, it is necessary to construct the feature of each edge point of each cable to provide a basis for subsequent point matching.

[0135] To obtain stable and reliable edge point features from the texture-deficient cable surface, it is necessary to fully utilize the pixel distribution information near the edge points and combine it with robust gradient direction features. The local gradient direction histogram formed by pixels in the neighborhood of the edge point is used as the edge point feature. Based on the extracted pixel coordinates of the edge points, the corresponding positions are found in the acquired original image for feature construction and extraction. The specific steps are as follows:

[0136] Step S7010: Gaussian filtering is applied to the eighth target image and the ninth target image, and the filtered eighth target image and the ninth target image are converted into single-channel grayscale images. The gradient direction code is calculated on the converted eighth target image and the ninth target image using a target recognition algorithm to obtain the corresponding gradient direction histogram.

[0137] Specifically, the original cable image is first subjected to Gaussian filtering to reduce noise interference. Then, the 3-channel RGB image is converted into a single-channel grayscale image, and the gradients in the x and y directions of the entire grayscale image are calculated using the Sobel operator to obtain the gradient direction histogram S. x and S y Through S x and S y Calculate the gradient direction at each location in the image, add 2π to the negative angles in the gradient direction, and adjust the angle interval to [0, 2π). The gradient direction feature map is then obtained as A(x, y). Next, quantize each gradient direction of A(x, y). Divide the possible gradient direction angle interval [0, 2π) into k equal subintervals, where k is an integer greater than or equal to 2. Then, based on the subinterval where the gradient direction A(x, y) lies, quantize it into the index corresponding to the lower limit of the interval. This can be achieved by rounding down.

[0138] Step S7011: Construct a neighborhood of a preset size with the position of each edge point as the center in each gradient direction histogram, count the gradient direction codes in the neighborhood to obtain the corresponding target histogram, and determine the corresponding edge point set based on each target histogram to obtain the sixth target set and the seventh target set.

[0139] Specifically, a local gradient orientation histogram is constructed: For the image acquired by the camera at the first position, each edge point can be extracted after region filtering and its pixel coordinates are obtained. In the quantized gradient orientation histogram, the position (x, y) of each edge point is used as the coordinates. j ,y j Using a point as the center, construct a neighborhood of size s*s. Then, statistically analyze the gradient direction values ​​within the neighborhood to obtain the feature set of all edge points. By applying the same neighborhood size to the edge points extracted from the second image and constructing a local orientation histogram using the method described above, a feature set can be obtained: Where n1 and n2 represent the number of edge points extracted from the first and second images, respectively. These represent the local gradient direction histograms of feature points in two images, each an array containing k elements, with each element counting the frequency of the corresponding gradient direction.

[0140] Through the above steps, a corresponding local gradient orientation histogram feature can be constructed for each edge point of the cable in each image, thereby forming the feature sets of the two images respectively.

[0141] Step S7012: Perform feature point matching based on the sixth target set and the seventh target set, and store the successfully matched feature point pairs into the eighth target set;

[0142] Specifically, image feature point matching and segment spatial pose calculation are performed. By extracting the gradient direction histogram of edge points, feature sets F for each of the two images are obtained. I and F II Using this information, and in conjunction with the results of stereo correction, it is possible to quickly match edge points between two images.

[0143] First, it is necessary to define the search interval for matching. Observing the stereo-corrected image, we can see that the two images are already aligned. Let F... I Any feature element in is F II Any feature element in is When starting feature matching, for an edge point feature in the first image... Only need to F II Find with Having the same row That is, satisfy Then the corresponding features Incorporated into In the set of points to be matched, each feature obtained from the first image only needs to be matched with a few features obtained from the second image, greatly reducing the amount of computation; F II The local gradient direction histogram of all features within the search interval, and... The corresponding local orientation histogram features are matched one by one to calculate the matching cost function: Where s is the neighborhood size and k is the number of intervals in the gradient direction, calculated by... With F II The cost function of all features within the search interval is compared to obtain the minimum cost and the matching feature. The edge point corresponding to this feature is then compared with... The corresponding edge points are used as the final matching point pairs; combined with the above formula, the coordinate values ​​of the cable points in the robot's base coordinate system can be obtained.

[0144] Step S7013: Calculate the coordinate values ​​of all edge points in the coordinate system of the robotic claw 106 based on the eighth target set to obtain the first coordinate value, and calculate the depth coordinate value of the edge point to obtain the second coordinate value. If the deviation between the second coordinate value and the average value of the second coordinate value is greater than the sixth threshold, delete the corresponding first coordinate value.

[0145] Specifically, for each calculated cable edge point in the space, a screening process is performed by comparing its depth coordinate value z one by one. Δd is the average depth of all spatial points at the cable edge. max To set a threshold, edge points that do not meet the conditions will be removed;

[0146] The algorithm described above achieves the filtering and extraction of cable regions, obtains the pixel coordinates of the left boundary sampling points of the cable regions in the image, and calculates their spatial coordinates through matching.

[0147] Step S7014: Based on the first coordinate value, convert it into spatial coordinates to obtain the third coordinate value, and determine the direction vector of the cable based on the third coordinate value to obtain the positioning of the gripping part.

[0148] Specifically, the spatial pose of the cable gripping segment is further calculated, and the pixel coordinates of the boundary sampling points are selected based on the region of the cable gripping segment. The corresponding three-dimensional spatial coordinates are Since the cable gripping segment is only a small part of the entire cable, it can be regarded as an undeformed cylinder. This can be achieved by sampling all points on this cable segment. By performing line fitting, a straight line in space can be obtained. Among them, the straight line direction vector s is (a,b,c), which is the direction vector of this cable segment, and finally determines the location of the gripping part.

[0149] After obtaining the direction vector of the cable grab segment, the next step is to detect the pose of the cable terminals. This requires detection in two aspects to ensure the final insertion of the cable terminals:

[0150] The first aspect: determining whether the crawling is valid;

[0151] The second aspect: Calculate the terminal's pose relative to the tool coordinate system and determine whether this pose is within the adjustable range of the inlet guide mechanism. The effective gripping of the gripper structure composed of the mechanical claw 106 and the protective clamp 107 means that the gripping position of the cable and the deviation of the cable terminal's orientation from the gripper are within the expected range. Both of these aspects can be determined by calculating the deviation of the cable terminal's direction vector from the coordinate axis of the tool coordinate system on the gripper. Ignoring the terminal's own radial rotation angle, the gripper is pre-taught to two fixed positions relative to the fixed camera. During testing, if... Figure 1 As shown, the gripper grabs the cable, and the fixed camera takes a picture each time it moves to one of these two positions.

[0152] Based on the set conditions, it can be known that the area of ​​the cable terminal in the image can be indirectly identified through the positioning gripper area. The terminal has a complex metal structure and lacks texture and obvious feature points on the surface. Therefore, starting from its edge point features, the target edge point is found based on its contour features. Then, combined with stereo correction, matching point pairs can be searched to calculate the spatial position of the point on the front surface of the terminal and the direction vector of the terminal.

[0153] Since the gripped cable is located on the gripper, the cable terminal area can be identified by recognizing the area of ​​the gripper. Since a fixed camera is used to detect the terminal, and the fixed camera captures images of the cable being gripped by the gripper at the teaching point each time, the area of ​​the gripper can be located by template matching.

[0154] As one possible implementation, based on the positioning of the connector socket 203 and the gripping part, the cable end core is repositioned to obtain the cable end core positioning, including:

[0155] Step S801: When not gripping, control the mechanical gripper 106 to move to a preset position, acquire an image of the mechanical gripper 106, convert the image to a grayscale image, and determine the pixel area corresponding to the mechanical gripper 106 in the grayscale image as a template image.

[0156] Specifically, the template is first created. When the gripper is not gripping the cable, the device moves to the teaching point and captures an image of the gripper on the cable terminal under no-load conditions using a fixed camera. The RGB image is then converted to a grayscale image, and the pixel area of ​​the gripper portion in the image at this time is extracted as the template image.

[0157] Step S802: Match the tenth target image, including the grasped mechanical claw 106, based on the template image, the mechanical claw 106 region in the tenth target image, and determine the position of the cable end core based on the positional offset between the mechanical claw 106 region and the cable end core.

[0158] Specifically, the Normalized Cross-Correlation (NCC) matching method is used for template matching. Assuming TTT is the template image and III is the image to be matched, the formula for calculating the normalized cross-correlation coefficient is: Where T(x,y) and I(x,y) represent the pixel values ​​at position (x,y) in the template and the image to be matched, respectively. T and μ I Let T and I represent the mean values ​​of the template and the image to be matched, respectively. The resulting NCC(T,I) ranges from -1 to 1, where a value closer to 1 indicates a more similar matching region. Therefore, the matched gripper image region can be obtained, and the cable terminal region can be obtained by offsetting the position of the region.

[0159] Step S803: Based on the position of the wire core at the end of the cable, the Bouguet stereo correction method is used to correct the position, and the corresponding pixel points of the wire core at the end of the cable and the cable body are extracted.

[0160] Specifically, the next step is to calculate the pose of the cable terminal. The two images from the fixed camera can be considered as a binocular imaging system in the tool coordinate system. Therefore, the Bouguet stereo correction method can be used, and a multi-threshold segmentation method can be applied to the two images. Here, two selected thresholds t1 and t2 are used for segmentation, which satisfies the following: The pixel blocks of the cable terminals and part of the cable body can be separated from the background.

[0161] Step S804: Fill the pixels with a morphological opening and closing algorithm and delete discrete pixels to obtain the eleventh target image;

[0162] Specifically, after filling the holes and removing discrete pixel blocks through morphological opening and closing operations, the next step is to remove the portion of the cable body from the pixel block.

[0163] Step S805: Convert the portion of the cable end core in the eleventh target image into an HSV channel to obtain the twelfth target image. By performing multi-channel threshold segmentation on the twelfth target image, obtain the region corresponding to the cable body. Determine the outer contour of the cable end core based on the region corresponding to the cable body.

[0164] Specifically, based on the cable color characteristics, the terminal area image in the original RGB image is extracted separately and converted into an HSV channel. After multi-channel threshold segmentation, the pixel blocks of the cable body in the image are obtained, and finally the outer contour of the terminal is extracted.

[0165] Step S806: Based on the outer contour, using vertex chain code and rotation method, calculate the minimum bounding rectangle of the outer contour and the centroid position of the minimum bounding rectangle.

[0166] Specifically, the outlines of the two terminals can be obtained through the above operations. For the first image, the minimum circumscribed rectangle of the outline and its centroid position are calculated according to the vertex chain code and rotation method.

[0167] Step S807: Calculate the slope of the line containing the longest side of the smallest bounding rectangle to obtain the target slope, construct a line passing through the centroid position with the target slope, and determine the intersection point of the line and the outer contour.

[0168] Specifically, considering that the cable terminal is long and narrow, the slope k of the straight line containing the long side of the smallest circumscribed rectangle is calculated, and a straight line passing through the centroid of the rectangle is constructed with k as the slope. This line is the symmetry line in the direction of the long side of the smallest circumscribed rectangle. Then, all intersection points of the straight line with the contour of the first image are solved.

[0169] Step S808: Determine the center point and direction vector of the cable end core based at least on the intersection point to obtain the target position and target vector, and thus obtain the cable end core positioning.

[0170] Specifically, by using coordinate position relationships, two edge points p on both sides of the image are selected. a p b Using the target edge points as references, and combining the row alignment constraints of stereo correction, a matching point p between the two target edge points can be found in the contour of the second image. a '、p b Then, by using the epipolar constraint combined with the formula, the spatial coordinates P of the two target edge points can be calculated. a and P b These two points correspond to two points on the outer wall of the terminal. Combining this with prior information about the cable terminal dimensions, we can obtain the center point P of the front face of the cable terminal. c The direction vector s of the terminal is used to accurately position the wire core at the end of the cable.

[0171] Furthermore, the tool coordinate system is established on the gripper, with its positive x-axis pointing in the insertion direction. Representing the terminal direction vector s in the tool coordinate system, and taking the unit vector of the positive x-axis of the tool coordinate system, we can obtain the angle θ between it and the direction vector s. Define a region with radius r starting from the origin of the tool coordinate system. Then, determine if the distance from the center point Pc of the terminal front face to the origin of the tool coordinate system and the angle θ between the direction vector and the unit vector of the xE axis satisfy the following conditions: This is used to determine the degree to which the cable terminal deviates from the gripper. If the condition is met, the gripping is deemed reliable and effective, and it is within the adjustable range of the insertion hole guide mechanism. The final cable hole assembly is then performed in this position. If the condition is not met, the gripping needs to be repeated.

[0172] After grabbing the cable, the position of the terminal relative to the tool coordinate system needs to be detected to determine the validity of the grab and whether the insertion is allowed.

[0173] This invention provides a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device containing the computer-readable storage medium to perform the automatic capture method for the flight plug.

[0174] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0175] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0176] 1) This application discloses a portable automatic crimping device for aircraft connectors. The device includes: an automatic identification mating structure for recognizing the wire core and mating position, and carrying the wire core to the mating position to the connector socket of the connector crimping structure for insertion; and a connector crimping structure for crimping the connector to the end of the wire core placed within the connector socket. This application, through precise recognition of the connector and cable images and positions, combined with cable gripping and end positioning, delivers the cable end to the connector mating position and further identifies the specific posture of the wire core, thereby ensuring that the wire core is completely positioned in the connector's connection position during crimping. This method effectively improves the accuracy of the mating process, ensures reliable contact between the wire core and the connector, and improves crimping quality. Simultaneously, the automated mating and crimping process significantly improves operational efficiency, providing technical assurance and comprehensive support for efficient and precise automated operation of aircraft connectors. It solves the problem in existing automatic crimping devices for aircraft connectors and cable ends where inaccurate connector and cable identification leads to low crimping quality.

[0177] 2) The automatic gripping method for cable connectors in this application achieves efficient and automated gripping and insertion of cable end cores and connector sockets. Preliminary identification provides rapid initial positioning, while the target recognition algorithm based on accumulated quantization gradient direction features and multi-view stereo vision technology further improves the positioning accuracy of the connector and cable gripping areas. The image feature positioning algorithm ensures precise alignment of the socket and core. The multi-step collaborative process effectively solves problems such as cable bending and complex backgrounds, thereby significantly improving the automation, accuracy, and reliability of the insertion operation, providing an efficient solution for complex cable connection scenarios.

[0178] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A portable automatic crimping device for aircraft connectors, characterized in that, The device includes: Automatic identification of the plug-in structure is used to identify the wire core and the plug-in position respectively, and to carry the wire core to the plug-in position to the position of the connector socket of the connector crimping structure and insert it; A connector crimping structure, wherein the connector crimping structure is used to crimp the connector to one end of the wire core located within the connector socket; the automatic identification mating structure includes: Identify the moving track; The sliding seat engages with the identified moving track; The connecting rod is fixed to the bottom of the sliding seat; The side rod is connected to the bottom of the connecting rod; A sensor is connected to the bottom of the side rod, and a camera is mounted on the sensor. The camera is used to identify the image and position of the fiber core, the connector socket, and the connector. The mechanical gripper is movably connected to the bottom of the side rod; A protective clip, connected to the bottom of the side rod, is used to clamp the wire core and insert it into the connector socket; The connector crimping structure includes: The connector moving track is placed parallel to the identification moving track; A connector placement base is slidably mounted on the inner side of the connector moving track, and the surface of the connector placement base has multiple connector sockets for placing the connector.

2. The apparatus according to claim 1, characterized in that, When the wire core is inserted into the connector socket, the single fiber core provides a longitudinal insertion force greater than or equal to 10N and an insertion depth of not less than 9mm. When the fiber core is inserted, the total insertion and separation force is greater than or equal to 100N.

3. An automatic method for acquiring aerial inserts, characterized in that, The automatic gripping method is applied to the portable automatic crimping device according to claim 1 or 2, and the method includes: The cable end wire core and connector socket are initially identified, and the cable end wire core is moved to the position of the connector socket; A target recognition algorithm based on accumulated quantized gradient direction features is used to locate the connector, thus obtaining the connector location. The automatic identification and plugging structure is controlled to reposition the connector socket based on the connector positioning using an image feature positioning algorithm, thereby obtaining the connector socket positioning; The gripping part of the cable is located by using a multi-view stereo vision positioning algorithm, and the gripping part location is obtained. Based on the connector socket positioning and the gripping part positioning, the cable end core is repositioned to obtain the cable end core positioning, and the insertion operation is performed based on the connector socket positioning and the cable end core positioning.

4. The method according to claim 3, characterized in that, A target recognition algorithm based on accumulated quantized gradient direction features is used to locate the connector, resulting in connector localization, including: Based on the first target image, identification is performed, and the gradient components of each color channel of the first target image in the x and y directions are solved based on the Sobel operator to obtain the first gradient component and the second gradient component. The gradient magnitude and gradient direction are calculated based on the first gradient component and the second gradient component of each color channel, and the gradient direction corresponding to the color channel with the maximum gradient magnitude is determined as the target gradient direction. Obtain a preset correction angle. If the target gradient direction is negative, sum the preset correction angle and the target gradient direction, and update the target gradient direction based on the summation result. Based on the updated target gradient direction, the quantization interval to which the target gradient direction belongs is determined to obtain the target quantization interval. The quantization interval is obtained by dividing a preset angle interval by an average. If any two target gradient directions differ by a first preset angle, the two target gradient directions are determined to be the same target gradient direction. Based on the boundary value of the target quantization interval that has the smallest difference from the updated target gradient direction, the target gradient direction is updated again, and the number corresponding to the target gradient direction is binary encoded to obtain the first target direction code; Based on the first target direction code, a first gradient direction feature map corresponding to the first target image is obtained by integration. A reference image is acquired, and each pixel of the reference image is subjected to multiple random rotation and translation transformations. The second direction code after each random rotation and translation transformation is recorded, and the second direction code is integrated to obtain multiple second gradient direction feature maps corresponding to the reference image. Multiple second gradient direction feature maps are accumulated to obtain a third gradient direction feature map, and the number of times the third gradient direction feature map is accumulated is recorded to obtain the target number; Using the third gradient direction feature map as a window, a sliding traversal is performed on the first gradient direction feature map, and the score of the window at each position is calculated based on the objective function to obtain the target score; The location corresponding to the maximum target score is determined as the connector, and the connector is located based on its position in the first target image to obtain the connector location.

5. The method according to claim 4, characterized in that, The automatic identification and mating structure is controlled to reposition the connector socket based on the connector positioning using an image feature positioning algorithm, thereby obtaining the connector socket positioning, including: The position of the connector in the first target image is determined as the second target image; The second target image is divided into multiple sub-regions. Histogram calculation is performed on the pixel grayscale of each sub-region, and the average pixel grayscale of each sub-region is calculated. Obtain a cropping coefficient based on the average pixel grayscale, crop the portion of the histogram that is greater than a first threshold based on the preset cropping coefficient, and distribute the cropped portion evenly to each grayscale level of the histogram; Based on the histogram, the second target image is subjected to bilinear interpolation of pixel gray levels to obtain the third target image; Normalize each pixel row and / or each pixel column of the third target image to obtain the fourth target image; The fourth target image is scanned line by line based on the horizontal scan lines to construct a position number-gray level histogram of a single row of pixels. The horizontal axis of the position number-gray level histogram is the position number of each pixel in the pixel row, and the vertical axis of the position number-gray level histogram is the gray level. Construct a sliding window of a preset width, slide it along the horizontal axis of the position number-grayscale histogram, traverse the peaks in the position number-grayscale histogram to obtain a first target set, wherein the width is an odd number and is less than or equal to the width of two adjacent peaks; Based on the preset width, a first target step size is determined. Each peak in the first target set is traversed. If the peak satisfies that the difference between the boundary pixels of the peak is greater than a second threshold, the peak is deleted from the first target set to obtain a second target set. Obtain a preset number, and if the number of peaks in the second target set is equal to the preset number, determine the second target set as the edge feature point set; The edge feature point sets corresponding to each pixel row are integrated to obtain the first edge feature set, and the fourth target image is scanned column by column based on the vertical scan line to determine the corresponding second edge feature set; Obtain a preset distribution, wherein the preset distribution is the arrangement of the connector sockets on the connector; Based on the first edge feature set and the second edge feature set, they are recombined according to the preset distribution to obtain multiple third edge feature sets. Each third edge feature set corresponds to a connector socket, and each third edge feature set includes multiple edges. A weighted average is calculated for each element in each of the third edge feature sets to obtain the center point of each edge. The mean is calculated based on the center points of the horizontally adjacent edges and the vertically adjacent edges to obtain the first mean and the second mean. The average value is calculated based on the first mean and the second mean to obtain the pixel coordinates corresponding to the center of the connector socket. The spatial coordinates of the center of the connector socket are determined based on the pixel coordinates to obtain the positioning of the connector socket.

6. The method according to claim 4, characterized in that, The gripping area of ​​the cable is located using a multi-view stereo vision positioning algorithm, resulting in the following positioning details: Obtain the captured fifth target image including the wire core at the end of the cable, and count the number of pixels in each white area of ​​the fifth target image to obtain the target quantity; If the number of targets is greater than the third threshold or less than the fourth threshold, the white area is deleted. The fifth target image is cropped using the bounding rectangle of each white region as the boundary to obtain the sixth target image. The image is then scanned using a horizontal scan line, and the single-sided boundary point of the white region in each sixth target image is determined as the first sampling point. Multiple rays are constructed starting from the first sampling point, wherein the angle of the rays is within a first preset range, and the angle difference between two adjacent rays is a second preset angle. The search is performed along each of the aforementioned rays, and the intersection of the ray with the opposite boundary is determined as the second sampling point; Calculate the boundary distance between the first sampling point and each of the second sampling points to obtain multiple first target distances. If the first target distance is within a second preset range, add the second sampling point to the third target set. A fourth target set is constructed based on all the first sampling points, a fifth target set is constructed based on all the third target sets, and a ratio is calculated based on the number of elements in the fourth target set and the number of elements in the fifth target set. If the ratio is greater than a fifth threshold, the region corresponding to the sixth target image is determined as a cable region, thus obtaining the first cable region. Obtain a seventh target image including the wire core at the end of the cable before grabbing, and mark the cable area on the seventh target image to obtain a second cable area; Based on the first cable region and the second cable region, the sixth target image and the seventh target image are processed using the Bouguet stereo correction method to obtain the eighth target image and the ninth target image; Gaussian filtering is applied to the eighth target image and the ninth target image, and the filtered eighth target image and the ninth target image are converted into single-channel grayscale images. Gradient direction codes are calculated on the converted eighth target image and the ninth target image using a target recognition algorithm to obtain the corresponding gradient direction histogram. In each of the gradient direction histograms, a neighborhood of a preset size is constructed with the position of each edge point as the center. The gradient direction codes in the neighborhood are statistically analyzed to obtain the corresponding target histogram. Based on each of the target histograms, the corresponding set of edge points is determined to obtain the sixth target set and the seventh target set. Feature point matching is performed based on the sixth target set and the seventh target set, and the successfully matched feature point pairs are stored in the eighth target set. Based on the eighth target set, calculate the coordinate values ​​of all edge points in the mechanical claw base coordinate system to obtain the first coordinate value, and calculate the depth coordinate value of the edge point to obtain the second coordinate value. If the deviation between the second coordinate value and the average value of the second coordinate value is greater than the sixth threshold, delete the corresponding first coordinate value. Based on the first coordinate value, a third coordinate value is obtained by converting it into spatial coordinates, and the direction vector of the cable is determined based on the third coordinate value to obtain the positioning of the gripping part.

7. The method according to claim 4, characterized in that, Based on the connector socket positioning and the gripping part positioning, the cable end core is repositioned to obtain the cable end core positioning, including: Without grasping, the mechanical gripper is controlled to move to a preset position, and an image of the mechanical gripper is acquired. The image is then converted into a grayscale image, and the pixel region corresponding to the mechanical gripper in the grayscale image is determined as a template image. The template image is used to match the tenth target image, which includes the grasped mechanical claw, the mechanical claw region in the tenth target image, and the position of the cable end core is determined based on the positional offset between the mechanical claw region and the cable end core. The cable end core position is corrected using the Bouguet stereo correction method, and the corresponding pixel points of the cable end core and the cable body are extracted. The pixels are filled using a morphological opening and closing algorithm, and discrete pixels are deleted to obtain the eleventh target image; The portion of the eleventh target image corresponding to the cable end core is converted into an HSV channel to obtain the twelfth target image. By performing multi-channel threshold segmentation on the twelfth target image, the region corresponding to the cable body is obtained. The outer contour of the cable end core is determined based on the region corresponding to the cable body. Based on the outer contour, the minimum bounding rectangle of the outer contour and the centroid position of the minimum bounding rectangle are calculated using vertex chaining and rotation methods. The slope is calculated using the line containing the longest side of the minimum bounding rectangle to obtain the target slope. A line passing through the centroid is constructed using the target slope, and the intersection point of the line and the outer contour is determined. The center point and direction vector of the cable end core are determined based on at least the intersection point to obtain the target position and target vector, thereby obtaining the positioning of the cable end core.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method of any one of claims 3 to 7.

Images