Edge Segmentation Applications to Design a Non-Destructive Recognition Algorithm for Continuous Actions in Martial Arts Videos

• Published in: 2024 International Conference on Smart Systems for Electrical, Electronics, Communication and Computer Engineering (ICSSEECC) pp. 707 - 713
• Main Authors: Hussam, Ragheed, Abed, Ansam Mohammed, Ahmed, Ibrahem, Abdulhussain, Zahraa N., Farhan, Khaled, Hasson, Ahmed Rasol, Al-Fatlawy, Ramy Riad
• Format: Conference Proceeding
• Language: English
• Published: IEEE 28-06-2024
• Subjects: Image color analysis; Image edge detection; Image segmentation; martial arts; neural network; Real-time systems; recognition; segmentation; Solid modeling; Virtual reality
• DOI: 10.1109/ICSSEECC61126.2024.10649445

As the difficulty level in competitive kung fu increases, martial artists must produce accurate, stable, excellent, and challenging actions to attain remarkable performance. Many diverse martial arts are interested in the real-time separation of martial arts movements. The main emphasis of this research is on the real-time collection and modelling of virtual reality picture targets in the demanding movement technology of martial arts routines. The pre-processing of existing images is examined in this research while taking the intricate details of martial arts moves into account. This article proposes an edge segmentation algorithm for the martial arts recognition (ESA-MAR) system. The attributes of the acquired photographs are enhanced using contrast enhancement, background subtraction, and conventional median approaches. This method can improve the aesthetic impact of the raw image, and the image size can aid in segmentation in the future. An innovative image classification approach is suggested for the image's colour model. The Hue Saturation Value (HSV) form of the colour image is converted, and the H element-which denotes chromaticity properties-is retrieved based on the HSV model H component. For H elements, the concept of a graph applies. The scaling target in the image is identified based on the distribution of the H element, which also establishes the classification threshold. Virtual reality (VR) technology is used because the modelling area is colour-sensitive to determine the segmentation target rapidly. The automated extraction of items in the image is finished when the above split methods are coupled. The technique of picture extraction employing VR technology is highly precise, with a relative error of 2.82. According to the findings, the approach produces good classification performance and is suited for picture classification and automated image extraction in difficult backgrounds.