Human Posture Estimation: In Aspect of the Agriculture Industry

Abstract

AbstractPose estimation is an artificial intelligence and computer vision approach. Human Posture Estimate is a more advanced version of pose estimation technology that graphically depicts the position and orientation of a human body. It's one of the most appealing fields of research, and it's gaining popularity thanks to its practicality and versatility—utilized in a range of industries, including gaming, healthcare, agriculture, augmented reality, and sports. This research project intends to establish a deep learning-based human posture identification system that can be used to identify diverse agricultural operations, with the intention of introducing the concept of automation into the agriculture field. A proprietary dataset of farmer postures is used to run this system. The picture from the dataset is pre-processed before a deep neural network is used to detect body points in the image, and OpenCV creates a graphical representation of the points. The angle between body components is crucial in determining posture, which is derived from various calculations. Finally, the result is compared to a threshold value before being processed. Our model could accurately measure a farmer's or human's posture in three major categories: sitting, bending, and standing with a test accuracy of about 77%.KeywordsHuman pose estimationGaussian blurAgriculture posture estimation