Magnetic Localization Algorithm of Capsule Robot Based on BP Neural Network

To explore the positioning tracking problem of capsule endoscopy, the advantages of magnetic positioning technology as a solution are highlighted. Considering that the capsule robot is small and cannot have enough built-in driving and positioning tracking devices, new technical means are needed to solve this problem. Therefore, we propose a prediction method based on the BP neural network model to locate the position of the robot in the magnetic field. To evaluate the accuracy of a BP neural network model with three hidden layers, the predicted results of the model, the results of the nonlinear algorithm, and the actual coordinate points were compared to determine whether it could accurately predict the actual coordinate points. We set the criterion for correct localization as the Euclidean distance between the predicted coordinate point and the actual coordinate point being less than 0.01 mm, and obtained the localization rates of the nonlinear localization algorithm and the neural network prediction method as 46.7% and 95.2%, respectively. By comparing the results, it is found that the positioning accuracy predicted by the BP neural network is higher and has higher accuracy. This shows that the prediction accuracy of the algorithm is more optimized, which can meet the real-time and positioning accuracy requirements of the capsule endoscope.

I. Introduction

As people’s lifestyles change, they are increasingly seeking medical services that are comfortable and efficient. Traditionally, cable endoscopes were used as the primary method for diagnosing and treating gastrointestinal diseases. However, this method could cause patients discomfort and scratch their gastrointestinal tract during insertion. To address these issues, researchers both domestically and internationally have been focusing on developing capsule endoscopy in recent years.