College of Engineering

Snapshots recorded from multiple cameras viewing the same dynamic event from different angles can be processed and used for the dynamic tracking of 3D displacements of multiple targets placed on the model. This report describes the first combined use of new high-speed Photron cameras and the TEMA Classic 3D software at the Center for Geotechnical Modeling (CGM) at University of California, Davis (UCD). The cameras and their mounting, as well as the target markers, lighting, camera calibration, and camera triggering are described, followed by a discussion on the software options selected for the analysis of videos recorded for a centrifuge model test conducted on the 9 m-radius centrifuge. The results presented show that this method is effective and reliable in obtaining the positions, displacements, velocities, and accelerations of the targets. Recommendations are made for improvements in future applications.

Obtaining the 3D displacements of targets requires multiple cameras to take snapshots(images) of the target from different view angles and a software to perform the image analysis.The Photron High Speed Camera system available at CGM UCD is equipped with six MH6monochromatic cameras that can record videos up to 10,000 frame per second and with amaximum resolution up to 1920 × 1400 pixels (only applicable for frame rates less than 1000 fps).Multiple trigger methods are available to trigger the cameras to start recording. The ResDAQsoftware at the CGM was modified to synchronize the Photron’s image acquisition systemsynchronized with the DAQ system. Triggers (CAMERATrigger and SNAPSHOTTrigger) weredeployed within the shaker controller to ensure the image acquisition and DAQ coincided with thedynamic shaking experiment. The CAMERATrigger triggers the Photron image acquisition systemto start saving recordings at the beginning of the shaking, i.e., when the motion file is sent to theshaker to control the shaking of the servo-hydraulic shaking table. The SNAPSHOTTrigger systemenables taking snapshots at a variable rate which is especially useful in a dynamic test when imagesare needed to be taken at a fast rate during shaking, and slower rate during reconsolidation. TheTEMA CLASSIC 3D software offers a library of tracking algorithms (correlation, quadrant, virtualpoints, center of gravity, etc.) that can be used to process images and track multiple targetssimultaneously to obtain their 3D position. Depending upon the plane of motion and number ofcameras used, it can obtain the 2D as well as the 3D motion of the object.

Using cameras and image analysis to obtain 3D movements of the model comprises several steps. These steps in order of implementation include: planning the marker locations, preparation of model surface, designing and producing the markers, positioning the markers, mounting the cameras, providing appropriate lightning, recording, and synchronizing videos, calibrating the cameras for lens distortion, determining the camera location and orientation, and finally using image processing to obtain 3D movements. Placing well-designed camera target markers at key locations makes it easier for TEMA to track them in the recorded images. A larger size target marker should be used for distant objects (away from the camera). Target markers should be placed on moving parts of the model (such as soil, pile, model container, and the centrifuge bucket) to enable calculation of their relative motion. The material used to fabricate the markers should not produce glare in the videos taken. Having proper lighting is key, especially at high frame rates. Sufficient light, uniform light, and no reflections are desired. At least two camera views must overlap for each target of interest such that the recorded images can be later processed to obtain its 3D position. The camera pairs should be mounted on a stiff beam, properly positioned, and focused to monitor the important parts of the model surveyed with target markers. When displacements are important in the direction of the view angle of the camera, the cameras should be moved apart to increase the stereo angle. It is further advised to take practice videos using the actual lighting, frame rate, and shutter speed to confirm the image quality and the field of view. This report outlines and describes all the steps in detail through an example implementation on a centrifuge model test featuring a layered liquefiable deposit with three embedded piles (SKS03). Two pairs of high-speed Photron cameras were placed in the model to monitor movements in the north and south section of model. The camera beam, light beams, and camera holder system were designed to be modular to make it easy to position and orient the cameras in any direction within the model. Three strips of LED lights (1000 lumens/foot) produced sufficient lighting to run the cameras at 1600 fps and a 4000 Hz shutter speed. Quadrant target markers and square grid markers were designed and placed throughout the model (on the soil surface, the piles, the container, and the centrifuge bucket). The model was shaken with multiple earthquake motions and videos of the model with target markers were recorded.

The snapshots recorded during and post shaking were processed in TEMA to obtain the 3D dynamic position of target markers. Soil and pile movements were obtained relative to the container by subtracting the average movement of the container top ring from their absolute movements. Movements obtained in the center section of the model independently from the north pair and the south pair cameras were identical. The obtained movements had a precision of 0.15 mm with smaller noise likely due to beam vibration, lighting variability, reflections from reflection from moving targets. Pile settlements obtained from the image analyses matched with the hand measurements taken using a depth gage. It was also possible to differentiate the marker positions obtained from the image analysis to obtain a reasonable estimate of the accelerations of the objects. The natural frequency of the camera beam (60 Hz) was found to be smaller than the applied shaking (in the order of 100 Hz). The vibration of the camera beam introduced noise in the obtained movements and as such installing the cameras on a stiffer beam would have reduced these vibrations. Results obtained on soil and pile movement showed that this method is effective and reliable in obtaining positions, displacements, velocities, and accelerations of the targets, and thus promising for use in future applications.

The use of cameras makes the model instrumentation relatively easier, cleaner (i.e., no LVDTracks and cables running across the model) and provides more model space for performing otherimportant investigations. It offers contactless sensing, which reduces the potential disturbance ofthe model. At the same time, the video recordings offer an immense amount of data which can beprocessed to get 3D displacements at any point within the model. The high-speed Photron camerasand TEMA software are found to be a great addition to the Center for Geotechnical Modeling atthe University of California, Davis, towards simplifying the model instrumentation whileadvancing the sensing capabilities in centrifuge tests, and they overall make an important steptowards the future of contactless model instrumentation and monitoring.