Shoulder and Knee Abnormality Examination Based on Artificial Landmark Estimation

Anthropometric detection tasks play a crucial role in medical and military recruitment processes as they help identify abnormalities that could otherwise be missed. Presently, these measurements are carried out manually using markers, which is a time-consuming process and prone to errors. This paper presents a computer vision-based system for detecting shoulder and knee abnormalities by automatically measuring shoulder tilt and knee distance to observe the knock-knees and bowlegs condition. The proposed system employs deep learning and BlazePose landmark estimation to accurately identify anomalies in the shoulders and legs. The Atan and Dist theoretical basis is applied for shoulder tilt and knee distance measurements, respectively. The proposed system can measure shoulder tilt and knee distance with an error rate of less than 10%. The automation of these measurements reduces the time required for examination and eliminates subjectivity and potential errors associated with manual measurements. Therefore, the proposed system has the potential to revolutionize shoulder and knee abnormality examinations by offering more accurate and efficient diagnoses.