Study on Classification of Metal Objects by Fluxgate Magnetometer Cube Structure

Abstract In order to classification the kinds of metal objects, we propose a recognition method based on ResNet-18. First, we built a magnetic field feature collector that we called FMCAS (Fluxgate magnetometer cube arrangement structure), using 8 fluxgate magnetometer sensor array structures to ensure a distance of 400mm between each sensitive unit. We use FMCAS to collect the magnetic field data of a survey line along the east-west direction on the north side of the measured target. Next, we change the location and type of the target and build a database of magnetic target models, which enriches the diversity of the training dataset. Construct the magnetic flux density tensor matrix to create the experimental dataset. Finally, we use the improved ResNet-18 to train the data to get the recognition classification recognizer. The recognition accuracy of this approach is 84.1%, according to the test results from 107 groups of validation sets. The target with larger magnetic moment has the best recognition effect, with a recognition accuracy rate of 96.3%, a recall rate of 96.4%, and a precision rate of 96.4%. Experimental results show that our improved RestNet-18 network can effectively handle the classification of metal objects.