Recognizing subsurface target responses in ground penetrating radar data using convolutional neural networks

Abstract

Improved performance in the discrimination of buried threats using Ground Penetrating Radar (GPR) data has recently been achieved using features developed for applications in computer vision. These features, designed to characterize local shape information in images, have been utilized to recognize patches that contain a target signature in two-dimensional slices of GPR data. While these adapted features perform very well in this GPR application, they were not designed to specifically differentiate between target responses and background GPR data. One option for developing a feature specifically designed for target differentiation is to manually design a feature extractor based on the physics of GPR image formation. However, as seen in the historical progression of computer vision features, this is not a trivial task. Instead, this research evaluates the use of convolutional neural networks (CNNs) applied to two-dimensional GPR data. The benefit of using a CNN is that features extracted from the data are a learned parameter of the system. This has allowed CNN implementations to achieve state of the art performance across a variety of data types, including visual images, without the need for expert designed features. However, the implementation of a CNN must be done carefully for each application as network parameters can cause performance to vary widely. This paper presents results from using CNNs for object detection in GPR data and discusses proper parameter settings and other considerations.

Artificial Neural Networks and Machine Learning techniques applied to Ground Penetrating Radar: A review

“…[46] Polynomial approximation (the feature vector) are ANN inputs. [47,48] The benefit of using a Convolutional NN (CNNs) is that features extracted from the data are a learned parameter of the system. [49] Regularized deconvolution is utilized to increase range resolution.…”

On Choosing Training and Testing Data for Supervised Algorithms in Ground-Penetrating Radar Data for Buried Threat Detection

“…Once the spatial location is obtained, the temporal location can be estimated. By far, the most common approach for temporal estimation relies on extracting keypoints at locations of high energy (e.g., local maxima) in the GPR A-scans [6], [8], [20], [29], [35]- [39]. These energy-based methods often yield multiple keypoints at each spatial location.…”

“…In this scenario, the classifier is applied at regular intervals along the A-scan and keypoints at the locations with the largest classifier decision statistics are utilized [6], [7], [11], [28]. The strategies in [8], [10], [20], [35], [36], [39], [45] set = , so that the same number of keypoints are utilized in training and testing.…”

“…In this work we trained and tested each classifier using fourfold cross-validation. This is a common approach for evaluating the performance of machine learning algorithms, and has been employed previously for BTD with GPR data [6], [20], [28]. Additional care had to be taken in our experiments because the GPR data is collected over the same lane multiple times.…”

“…Although it is possible to manually identify buried threats in GPR data, a great deal of published research has focused on automating this process with computer algorithms that provide a confidence of buried threat presence at each spatial location [4], [7], [8], [11]- [16]. Proposed algorithms have employed a variety of techniques from statistics [17], [18], computer vision [6], [19], [20], and machine learning [7], [21], [22]. The most successful approach to date involves the use of supervised learning techniques [12], [23]- [27].…”