Cable data set generation method based on chroma keying technology

Abstract

The embodiment of the invention relates to the technical field of computer vision and image processing and the field of automatic wiring, in particular to a cable data set generation method based on a chroma keying technology. According to the method, a cable image is collected under a specific background color, and a cable foreground is accurately extracted by using a chroma keying algorithm, so that high-quality segmentation annotation data is automatically generated. The method comprises the following steps: setting a standard background and a uniform illumination environment, and collecting a multi-view and multi-attitude cable image; removing a background by adopting a chromaticity matting algorithm to obtain an accurate mask of the cable; and forming a segmentation data set by the original image and the mask. Compared with a traditional manual labeling mode, the method has the advantages of being high in efficiency, low in cost, capable of achieving large-scale automatic processing and the like, and is particularly suitable for training cable detection, recognition and segmentation models. The method can be widely applied to cable identification tasks in industrial automation, intelligent assembly, machine vision and other scenes.

Description

Technical Field

[0001] The present invention relates to the fields of computer vision and image processing technology and automated wiring, and specifically to a cable data set generation method based on chroma keying technology. Background Art

[0002] With the advancement of artificial intelligence and computer vision technologies, deep learning-based cable detection and segmentation have broad application prospects in intelligent manufacturing, automated wiring, and other fields. To train robust visual models, building large-scale, accurately annotated cable image segmentation datasets is a key prerequisite. Currently, cable image datasets primarily rely on manual image capture and annotation, which is inefficient and costly, susceptible to subjective factors, and results in inconsistent annotation quality. Manual annotation is particularly challenging when dealing with large numbers of curved and overlapping cable images. Furthermore, the complex and ever-changing backgrounds and diverse cable shapes in real-world scenarios place higher demands on the model's generalization capabilities. Traditional image acquisition methods struggle to obtain high-quality and diverse training samples in a short period of time, limiting model performance. In recent years, chroma keying technology has been widely used in film and television. It exploits the chroma difference between background and foreground colors to extract foreground images, offering the advantages of high precision and automation. Incorporating this technology into cable dataset construction could significantly improve the efficiency and accuracy of image segmentation label generation and reduce data production costs. Therefore, there is an urgent need for a cable dataset generation method based on chroma keying technology, which can efficiently obtain multi-perspective and multi-morphological cable images, and construct a diversified, high-quality structured dataset through automatic matting and background replacement to meet the needs of visual model training. Summary of the Invention

[0003] The purpose of the present invention is to provide a cable dataset generation method based on chroma keying technology, which automatically generates high-quality segmentation and annotation data by collecting cable images under a specific background color and accurately extracting the cable foreground using a chroma keying algorithm.

[0004] To achieve the above objectives, the present invention provides the following solutions:

[0005] The present invention provides a cable data set generation method based on chroma keying technology, comprising the following steps:

[0006] Create a standard shooting environment, including setting up a single background color screen and uniform lighting conditions to enhance the color difference between the foreground and background;

[0007] Collect multi-angle and multi-modal image data containing cables under the shooting environment;

[0008] The captured image is input into a chroma keying algorithm, which converts the original image from RGB to HSV color space. A background color threshold is set to automatically isolate the foreground area of ​​the cable. The foreground area is then binarized to generate a mask image. The mask image can be labeled and manually corrected using image annotation software (Labelme) to ensure accuracy.

[0009] The extracted cable foreground image is combined with a diverse set of domain-randomized background images. Background images can include industrial scenes, desktop environments, and mechanical structures. The synthesis process supports image fusion enhancement processing, such as brightness matching, edge feathering, resizing, and random perturbation. The result is a structured dataset with a one-to-one correspondence between the foreground mask and the synthesized image, which can be exported to common segmentation model training formats such as YOLO, COCO, or VOC.

[0010] Furthermore, the background color curtain preferably adopts a green or blue background, and the background color has high saturation and high color separation in the HSV space to enhance the accuracy of foreground extraction.

[0011] Furthermore, the shooting system includes a multi-angle turntable and bracket to enable the collection of cables at different angles and postures in three-dimensional space, thereby improving the diversity and coverage of image data.

[0012] Furthermore, after the foreground extraction, image processing operations such as hole filling, edge smoothing and small area denoising are performed on the cable mask to improve the structural integrity and edge accuracy of the mask.

[0013] Furthermore, the background image is derived from images collected from real industrial scenes, public image datasets, and synthesized through an image generation algorithm (GAN).

[0014] Furthermore, when synthesizing foreground and background images, image fusion algorithms such as edge feathering and Gaussian blending are used to enhance the realism and naturalness of the synthesized image.

[0015] Furthermore, in order to avoid unreasonable occlusion or floating phenomena in the synthesized image, depth map estimation or scene geometry constraints are introduced to assist in judging the rationality and spatial consistency of foreground placement.

[0016] Furthermore, the synthesized images are automatically paired with the corresponding mask images one by one and exported into standard structured dataset formats, such as YOLO, VOC, and COCO formats, to facilitate downstream visual model training and evaluation.

[0017] It can be seen from the above technical solution that compared with the prior art, the beneficial effects of the present invention are:

[0018] By introducing chroma keying technology, the cable foreground area is automatically extracted, reducing the reliance on manual pixel-by-pixel annotation and significantly shortening the data production cycle;

[0019] Standardized shooting environment and uniform background color settings significantly enhance the difference in foreground and background colors. Combined with HSV color space segmentation and post-processing algorithms, high-quality mask images with complete structure and clear edges can be obtained.

[0020] By automatically fusing with domain-randomized background images, synthetic images covering various scenarios are constructed, effectively simulating the complex background environments in actual industrial applications and improving the robustness of the model in real-world scenarios.

[0021] The present invention has a high degree of automation capability, can process images and generate their labels in batches, and can be easily integrated into the image data construction pipeline to achieve rapid expansion and iterative update of the data set;

[0022] The generated dataset supports common formats such as YOLO-Seg, COCO, and VOC, and can be directly used for training and evaluation of various semantic segmentation and instance segmentation networks, with good practicality and versatility.

[0023] With reference to the following description and drawings, specific implementations of the embodiments of the present application are disclosed in detail, indicating how the principles of the embodiments of the present application can be employed. It should be understood that the embodiments of the present application are not limited in scope. Within the spirit and scope of the appended claims, the embodiments of the present application include many variations, modifications, and equivalents. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of the generation method of the present invention

[0025] Figure 2 Schematic diagram of the generation method of the present invention

[0026] Figure 3 These are some images of the dataset of the present invention. DETAILED DESCRIPTION

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] The purpose of the present invention is to provide a cable data set generation method based on chroma keying technology, such as Figure 1 shown.

[0029] A green curtain was placed as the background in the shooting area. A highly saturated green with a distinct chromaticity (RGB value approximately (0, 255, 0)) was chosen as the background color. Ring-shaped soft lights were placed on the top and sides to ensure even, shadow-free lighting in the shooting area. The shooting platform was constructed of high-contrast, anti-reflective material to prevent ambient light from interfering with image quality.

[0030] Place the cable to be imaged in front of a background cloth. Manually change the cable's shape (e.g., bend, twist, wrap, etc.) and posture, and use the automated rotating platform to capture images at different angles (15° rotations). Capture at least 50 images of each cable to ensure coverage of various shapes and viewing angles.

[0031] Import the collected RGB image into the computer, use the preset chroma keying script to convert the image into the HSV color space, and set the background color range (such as H∈[60, 80], S≥100, V≥100). Pixels that meet this range are identified as the background, and the non-background part is the cable foreground.

[0032] The extracted foreground is binarized to generate a cable mask image. Labelme is used to review and label the mask image, with 1 representing the cable and 0 representing the background. If necessary, inaccurate edges are corrected to ensure labeling quality.

[0033] The extracted cable foreground image is saved as a PNG transparent image in the form of an alpha channel, and background images (including real backgrounds or simulated images such as workshop floors, work surfaces, and electronic cabinets) are randomly selected for synthesis.

[0034] Finally, the synthesized image and its corresponding mask image are exported in the YOLO Segmentation format, automatically generating image, mask, and annotation files for model training. It supports mainstream image segmentation models such as YOLO, DeepLab, and Mask R-CNN.

[0035] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The present application embodiment relates to a method for generating a cable dataset based on chroma keying technology, the method comprising: (1) Construct a standard shooting environment, including setting up a single background color screen and uniform lighting conditions to enhance the color difference between the foreground and background; (2) collecting multi-angle and multi-modal image data containing cables under the shooting environment; (3) applying a chroma keying algorithm to the image, separating the cable foreground and background according to a preset background color threshold, and generating a binary mask image of the cable through annotation software; (4) The processed cable foreground image is combined with the domain randomized background image to construct a structured cable image segmentation dataset.

2. The cable data set generation method based on chroma keying technology according to claim 1, characterized in that: In (1), the standard background color is a pure color to reduce environmental interference and maximize the contrast with the cable color.

3. The cable dataset generation method based on chroma keying technology according to claim 1, characterized in that: In (2), cable samples are placed manually or automatically to cover different bending degrees, cross shapes and distribution states to enhance the diversity and robustness of the data, and a rotatable platform or multi-angle bracket is configured in the shooting system to achieve image acquisition of the cable at different viewing angles such as top view, side view, and tilt angle.

4. The cable data set generation method based on chroma keying technology according to claim 1, characterized in that: In (3), the chroma keying algorithm is based on the HSV color space, converting the image from the RGB space to the chroma-sensitive space to enhance the distinction between the background color and the cable color, and accurately extracting the non-background area as the cable foreground by setting the hue and saturation range threshold of the background color; The separated foreground area is imported into the annotation software for manual annotation and correction to form segmentation label data that can be directly used for training.

5. The cable data set generation method based on chroma keying technology according to claim 1, characterized in that: Said (4), the domain randomized background image comes from a variety of real and synthetic scenes, and is used to simulate the visual performance of cables under different backgrounds; in the process of synthesizing the foreground and background, image fusion techniques such as edge feathering, color matching, brightness adjustment and shadow synthesis are applied to enhance the natural fusion of the foreground and background, and the synthetic images are paired with the corresponding masks one by one to form a data set format with a standard structure.

Images