Machine-Learned Yarn Cone Status Detection for Knitting Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing yarn detection systems for knitting machines require optical elements that are space-constrained and position-dependent, making it difficult to detect yarn cone positions and connections to tensioning devices accurately.
Innovation Solution
An image processing device using machine-learning models to detect the status of yarn cones, including residual amounts and connections to tensioning devices, without the need for optical elements, allowing for flexible placement and improved versatility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical elements such as photo-receiving sensors and reflective mirrors are provided in the vicinity of yarn cones, then residual amounts of yarns can be detected, but the space for optical elements is restricted and adjustment becomes difficult when yarn cones are arranged densely
Solution Approach 1:
The patent extracts the detection function from physical optical elements and implements it through image processing algorithms. By capturing images with a camera and analyzing them computationally, the system eliminates the need for complex optical element arrangements near yarn cones, thereby resolving the space restriction issue while maintaining detection accuracy
Solution Approach 2:
The patent creates a digital copy of the yarn cone and its surroundings through image capture. This digital replica allows for virtual analysis and measurement without requiring physical optical elements in the detection path, thus avoiding the spatial constraints and adjustment difficulties associated with dense yarn cone arrangements
2Stability of the object's composition
If the relative positions of yarn cones to optical elements are uniquely predetermined, then optical alignment can be fixed, but yarn cones cannot be detected when arranged out of the predetermined positions
Solution Approach 1:
The patent transforms the static, fixed optical alignment system into a dynamic one where the camera captures images from a fixed position but the system can detect and analyze yarn cones regardless of their specific positions. The image processing algorithms dynamically adapt to locate and measure yarn cones in any position within the camera's field of view, providing both stability and flexibility
Solution Approach 2:
The patent creates a universal detection system that can handle various yarn cone positions and configurations. By using image processing rather than fixed optical elements, the system becomes adaptable to different arrangements of yarn cones, tensioning devices, and other components, making it versatile for different machine configurations
3Measurement precision
If optical elements are used for detecting yarn cones, then residual amounts can be detected, but the connection between yarn cones and tensioning devices cannot be detected
Solution Approach 1:
The patent transitions from point-based optical detection to area-based image capture. By capturing two-dimensional images that include the entire workspace, the system can simultaneously detect both the yarn cones and their connections to tensioning devices, as well as other spatial relationships, providing comprehensive information in a single measurement
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing device and so on is provided such that it can detect simply the statuses of an arbitrary number of yarn cones, arranged on the upper part of a knitting machine or at its periphery with high workability and high versatility. The image processing device (4A) comprises an acquisition unit (40) and a yarn cone status detection unit (41) for detecting the statuses of the arbitrary number of yarn cones included in the captured image acquired by the acquisition unit (40) by applying image processing to the captured image. The yarn cone status detection unit (41) input the captured image acquired by the acquisition unit (40) to a learned model (5A) which has machine-learned the correlation between the captured image and the amounts of residual yarns of the arbitrary number of yarn cones in the captured image.