System and method for robotic inspection
An inspection system includes a ground penetrating radar (GPR) sensor configured to provide at least one probe signal to a portion of a structure and to receive at least one return signal resulting from the corresponding at least one probe signal. A processor is programmed to: receive the at least one return signal as scan data from the GPR sensor; generate maximum amplitude trace normalized data based on the scan data; generate temporal signal gain data based on the scan data; generate a power spectral density image for each of the scan data, the maximum amplitude trace normalized data, and the temporal signal gain data; and analyze, using a machine learning processor, the power spectral density images to identify an anomaly of interest, wherein the machine learning processor is trained to detect an anomaly of interest based on a training set of power spectral density images.
1 . An inspection system, comprising:
a ground penetrating radar (GPR) sensor configured to provide at least one probe signal to a portion of a structure and to receive at least one return signal resulting from the corresponding at least one probe signal; and
a processor in electronic communication with the GPR sensor, wherein the processor is programmed to:
receive the at least one return signal and combine two or more return signals as scan data from the GPR sensor;
generate maximum amplitude trace normalized data based on the scan data;
generate temporal signal gain data based on the scan data;
generate a power spectral density image for each of the scan data, the maximum amplitude trace normalized data, and the temporal signal gain data; and
analyze, using a machine learning processor, the power spectral density images to identify an anomaly of interest, wherein the machine learning processor is trained to detect an anomaly of interest based on a training set of power spectral density images; and
generate a map based on the identified anomaly of interest.
2 . The inspection system of claim 1 , wherein the anomaly of interest is moisture.
3 . The inspection system of claim 2 , wherein the processor is further programmed to label moisture in a map generated by a mapping circuit based on the identified moisture.
4 . The inspection system of claim 1 , wherein the machine learning processor is an artificial neural network (ANN).
5 . The inspection system of claim 4 , wherein the ANN is a convolutional neural network.
6 . The inspection system of claim 1 , wherein the at least one probe signal is a pulsed signal.
7 . The inspection system of claim 1 , wherein the at least one probe signal is a continuous signal.
8 . The inspection system of claim 1 , further comprising a localization circuit, a mapping circuit, or a simultaneous localization and mapping (SLAM) sensor.
9 . The inspection system of claim 8 , wherein the processor is further configured to receive localization information, mapping information, or both, corresponding to the scan data.
10 . The inspection system of claim 8 , further comprising a robotic platform to which the GPR sensor is affixed, the robotic platform configured to move over a surface.
11 . The inspection system of claim 1 , wherein the processor is further programmed to concatenate the power spectral density images into a 2-dimensional image before analysis using the machine learning processor.
12 . The inspection system of claim 1 , wherein the processor is further programmed to register the two or more return signals into a same coordinate frame.
13 . A method of inspecting a structure, the method comprising:
receiving scan data of a structure from a GPR sensor;
generating maximum amplitude trace normalized data based on the scan data;
generating temporal signal gain data based on the scan data;
generating a power spectral density image for each of the scan data, the maximum amplitude trace normalized data, and the temporal signal gain data; and
analyzing, using a machine learning processor, the power spectral density images to identify an anomaly of interest, wherein the machine learning processor is trained to detect an anomaly of interest based on a training set of power spectral density images; and
generating a map based on the identified anomaly of interest.
14 . The method of claim 13 , wherein the anomaly of interest is moisture.
15 . The method of claim 14 , further comprising labeling moisture in a map of the structure based on the identified moisture.
16 . The method of claim 13 , wherein the machine learning processor is an artificial neural network (ANN).
17 . The method of claim 16 , wherein the ANN is a convolutional neural network.
18 . The method of claim 13 , further comprising receiving localization information, mapping information, or both, corresponding to the scan data.
19 . The method of claim 13 , further comprising concatenating the power spectral density images before analysis using the machine learning processor.