Development Of A Compact, Real-Time, Optical System For 3-D Mapping Of The Ocean Floor.

Transcription

1 Development Of A Compact, Real-Time, Optical System For 3-D Mapping Of The Ocean Floor. Eric Kaltenbacher, Jim Patten, David English, David K. Costello and Kendall L. Carder College of Marine Science University of South Florida 140 Seventh Avenue South Saint Petersburg, FL eak@marine.usf.edu Abstract- We have developed an economical, compact sensor for real-time 3-D mapping of the ocean floor. Our real-time ocean bottom optical topographer (ROBOT) is engineered to use off-theshelf components while maintaining a small physical footprint. The optical imager of ROBOT allows for high-resolution mapping of the seabed. Resolution of the instrument is 2 x 2 x 2 cm with a field of view of 3m x 3m, at an altitude of 3m. A key element of ROBOT is a smart camera which provides real-time processing and data reduction. Processed data can be downloaded and quickly assembled into three-dimensional images. Applications for ROBOT include target detection and recognition as well as small-scale contour plotting. Although designed for operation on an autonomous underwater vehicle (AUV), other moving or towed platforms are feasible vehicles for ROBOT. Image data are coupled with navigational information for accurate location of detected objects. In this paper, we describe the construction of the instrument, discuss possible applications, and present field data. We also present planned enhancements to the prototype to increase the sensor s field of view to 3m x 5m at an altitude of at least 5m. I. Introduction The ability to provide 3-D images of the bottom for freshwater and marine environments has important implications for a wide variety of applications. A few examples include object detection in military applications, crash-site analyses, contour plotting and obstacle avoidance. Both acoustic and optical systems have been developed to provide 3-D images of the ocean floor. Sonar systems have proven effective because they can scan a large area relatively quickly. Water clarity does not have a major impact on the image quality of an acoustic system making it useful in turbid environments. However, sonar systems are limited in resolution and are easily detected by listening devices. For many applications, optical based systems may provide a good alternative to acoustics. Among the benefits offered by optical systems are high spatial resolution and the ability to operate quietly. These benefits, however, come at the expense of reduced image swath width (compared to acoustics) that is a result of the higher attenuation by water of light versus acoustic energy. In the ideal scenario, an optical based imager is used to provide detailed imagery of objects preliminarily identified by an acoustic system. This way, use of the slower optical system is limited to investigations of specific areas of interest. Unfortunately, turbid water presents a serious obstacle to successful imaging in an optical system. Range gating, streak-tube imaging and LIDAR systems have been developed to reduce the sensitivity to turbid imaging conditions. These approaches are effective, but they are expensive and complicated. Another method is to separate the source and detector so that backscattering effects in turbid waters are minimized. The sensor described in this paper is an optical instrument with separated source and detector. This bistatic arrangement is relatively simple and robust. Using off-the shelf components, we were able to build the system economically. Discussions in the following sections review the design of the instrument and present field data from several deployments. II. System Description ROBOT measures ocean floor topography by triangulation. An advantage with this technique is its relative simplicity. A laser is used to project a line of

2 light normal to the ocean floor (Figure 1). A camera positioned to view the laser line at a known angle captures images as the sensor moves through the water. Contour changes cause the image of the laser line on the bottom to change position on the camera s detector. The relationship between the vertical change in position of the line in the image is p FOV w R = S tan φ, (1) PT 2 where R is the vertical distance between the camera and the object, S is the separation between the laser and camera, φ is the camera s viewing angle, FOV w is the half-angle field of view for the imaging optics when submerged, P T is the total number of pixels along the vertical direction, and p is the current pixel and ranges from PT PT p. (2) 2 2 R MAX R MIN LASER FOV w S φ FOV w SEA FLOOR SMART CAMERA Figure 1: Geometrical setup for the optical system in ROBOT. An advantage of this arrangement is that the principle parameters of interest (i.e. R MIN and R MAX ) can be quickly changed. The separation between the laser and the camera or the camera s central viewing angle can be changed to meet current imaging conditions. These are simple mechanical adjustments that are easily made in the field. R MIN and R MAX are also controlled by the FOV of the lens on the smart camera. However, changing this lens in the field is not desirable because it requires disassembly of the pressure vessel. In addition, the R MIN parameter can be adjusted to minimize backscatter effects. Our system is normally operated with R MIN at about 1.5m. The strong backscatter in this region, which would normally corrupt our images, is outside of our field of view. The ROBOT system in its current configuration consists of four basic components: a 532 nm 65 milliwatt laser; a smart camera; a pentium-based single board computer; and a custom control interface. A diode-pumped solid state (DPSS) laser with an output power of 65 milli-watts was chosen as the light source. Emission is at 532nm, which is near the peak of the transmission curve for water. This continuous wave (CW) laser is reasonably priced. Its small size and compatibility with a battery power source also make it well suited for this application. Scanner-less optics are used to generate a line of light from the laser beam. This approach is much simpler and more robust than traditional mechanical methods for generating a line of light. The optics are designed so that the intensity along the length of the line is fairly uniform. The intensity does not have a Gaussian profile normally associated with conventional optics. The heart of ROBOT is an integrated imager and processor or smart camera[1]. Inside the smart camera is a 256 x 256 pixel CMOS image sensor coupled with 256 processor units one for each column of pixels. Simple algorithms can be programmed and executed at high-speed with this sensor because the 256 processors operate in parallel. Since the camera was specifically designed for triangulation measurements, algorithms for locating a line in the sensor s images are already programmed into the sensor. Also, onboard processing data greatly reduces data storage/transmission rates. Instead of sending 256 x 256 bytes of image data, the smart camera only sends 256 bytes indicating the location of the line in each column of the detector. The camera can operate at speeds up to 200 frames/sec in this mode, thus allowing real-time, high-resolution inspection of the ocean floor. Another desirable feature of the smart camera is the ability to record height and albedo information

3 simultaneously. Intensity of the laser line is recorded during the processing of the line s location on the sensor and can be used as an indication of the bottom reflectance. Using combined range and intensity data facilitates object identification and classification. A small-format single-board computer is used to control operation of ROBOT. This Pentium based computer board allows use of standard peripherals such as hard disks, display adapters, and an Ethernet interface. Line-location data from the camera are received through a PCI slot and locally stored on a hard disk. The Ethernet connection is used after a mission to download data In addition to the Ethernet connection, a LONWorks interface is used for communication and control. LONWorks is a commercial, 2-wire, networking system designed for distributed control of equipment. In our application, we use LONWorks to communicate with the propulsion section of the AUV. Complete navigation information is exchanged via this connection and stored. This information can be combined with image data during post processing. The LONWorks interface is also used to control data collection. Variables such as vehicle speed, altitude or position can be monitored and used to trigger the start/stop of a data collection mission. The shipboard control application configures the data collection parameters for a ROBOT mission. These criteria consist of both mission-related parameters (e.g., altitude) and camera-related parameters (e.g., frame rate). Once the data collection parameters are configured, the mission can be started from the shipboard application. section. While OEX has a 24 diameter housing, ROBOT can be configured to fit in an 8 housing and thus is compatible with the next-generation mini-auvs (e.g. FAU Morpheus). A list of specifications for ROBOT can be found in Table 1. Resolution of the instrument is set for 2.0 cm x 2.0 cm x 2.0 cm with a field-of-view of 3m x 3m. This resolution is achieved at a vehicle speed of 2 kts. Faster vehicle speeds (up to 4.5 kts with FAU drive section) can be used with a decrease in along-track resolution. ROBOT is mounted in a 24 diameter nosecone to be compatible with the FAU s OEX tail section. Dry weight of the sensor/nosecone combination is roughly 200 lbs, however it is neutrally buoyant. At 20 frames/second data rate for the instrument is only 5.1 kbytes/second. Two 12V battery packs are used to supply power to the instrument and will last 4 hours with a full instrument load. Resolution Vehicle Speed 2.0 x 2.0 x 2.0 cm at 2 kts. Nominally 2 kts. Size 24 inches diameter by approximately 12 feet with drive section. Weight In air roughly 200 lbs (nosecone and ROBOT), neutrally buoyant in water Data 256 bytes every 50 ms. (Based on current frame rate) Endurance Navigation Table 1: Performance Data 4 hours GPS fixes and acoustic local tracking III. Applications and Data Figure 2: ROBOT nosecone payload mated with the OEX propulsion unit. The forward Stinger houses the source laser. The single board computer, camera system, and laser are mounted in a payload nosecone that couples with the Ocean Explorer (OEX) [2] propulsion unit developed by Florida Atlantic University. Figure 2 shows the ROBOT nosecone mated with the OEX tail ROBOT s 3-D imaging capability can be used in a wide variety of applications. Structure of coral reefs and stromatolite fields can be quantified using ROBOT. Military uses for the system may include mine and bomb detection and classification. Debris at crash sites can be imaged with ROBOT. Topographic information of the ocean floor can also be measured. In shallow waters, this mapping can be compared with satellite imagery as a method for ground-truthing remote sensing. Further possibilities exist for studying the surface of ships hulls or imaging the underside surfaces of icebergs. Finally, if the system is used in a forward-looking mode, it might be employed for obstacle avoidance in small manned or unmanned vehicles.

4 We have successfully deployed ROBOT several times during Coastal Benthic Optical Properties (CoBOP) field initiatives at Lee Stocking Island, Great Exuma, Bahamas and off the coast of Fort Lauderdale. The initial deployment of ROBOT was associated with missions conducted during CoBOP. The system was towed at two to three knots over mixed sand and shell mounds at a depth of two to three meters for the first missions. The nosecone was then fitted to the OEX and the AUV was flown over stromatolites in the Adderly Cut channel at a depth of seven meters and a speed of two to three knots. An image of the stromatolites is shown in Figure 3. Figure 4: Sand profile imaged by ROBOT during CoBOP. Figure 3: ROBOT image of stromatolites near Lee Stocking Island. Later missions at CoBOP were conducted over sand banks, mixed shell, grass and sand bottoms (Figure 4), and mine-like objects that were placed for these tests. All missions were conducted at a depth of five to seven meters at a speed of two and a half knots. The duration of these missions was from two to five minutes. A total of nine missions were completed. ROBOT successfully produced topographic images of the seabed and provided a 3-D image of a mine-like object. A cross-sectional view of the mine-like object is shown in Figure 5. Figure 6 shows corresponding albedo data. Figure 5: Cross sections of a mine-like object. The channel through the middle is a loss of signal on the black chain securing the object to the sea floor. Figure 6: Intensity (albedo) data for the mine-like object. The chain used for securing the mine is running diagonally across the object.

5 Missions at Ft. Lauderdale were conducted at depths of 10 meters and at a speed of three and a half knots. Mission duration was thirty minutes for each of the four missions. The missions were conducted over a test area containing varied types of structure, cables and sand. Figure 6 shows an image of a large (2m high) object recorded during a mission at Fort Lauderdale and demonstrates the vertical dynamic range capability of ROBOT. conditions, however, present a substantial obstacle to underwater range imaging. To better quantify the performance of ROBOT, a series of tests were run in the USF Center for Ocean Technology (COT) flume under controlled conditions. Targets with known reflectances were imaged using ROBOT. The maximum range for each target was determined by monitoring the image of the laser line on the target, while moving the target away from the camera. The maximum range is the distance at which the line image was no longer discernible by the smart camera s algorithms. Maximum range values were determined in varying degrees of turbidity. These range data and data indicating laser line spreading as a function of turbidity were combined to estimate the working conditions for the ROBOT prototype. These estimations are shown in Figure 9. Figure 7: Object in sand off Ft. Lauderdale Figure 9: Estimated working ranges for the prototype ROBOT system as a function of turbidity for different reflectance targets. Current developments are expected to increase the working range of the sensor. IV. FUTURE WORK AND CONCLUSIONS Figure 8: Intensity data for a bomb-like object. Coastal-cleanup efforts might also benefit from a sensor such as ROBOT. Figure 8 shows an intensity image of a bomb-like object. Imagery such as this might be used to identify old ordinance remaining in coastal areas from previous military excercises. Images shown above represent successful ROBOT deployments in relatively clear water conditions (c<0.2/m at 532 nm). Turbid water The prototype ROBOT system has performed well but we recognize a few areas where improvements can be made. Range estimation data (Figure 9) indicate an image intensifier might help extend the operating ranges for our sensor, especially on low reflectance targets. Since the algorithms in the smart camera were designed for use in air, they are not optimized for the underwater imaging condition with a large amount of backscatter. We are pursuing custom algorithms that might more effectively locate the laser line in a noisy image. New high-sensitivity CCD detectors have just recently become available and we are investigating this new detector technology for use in the next version of ROBOT. Combining this

6 detector and a stronger laser should provide the ability to operate at higher altitudes and higher vehicle speeds. Our goal is to maintain 2 x 2 x 2 cm resolution at an altitude of 5m and a vehicle speed of 5kts. Our ROBOT prototype has successfully recorded 3-D images of a variety of objects on the ocean floor. It has proven to be a reliable instrument requiring very little maintenance. While the system remains a clearwater instrument, we are endeavoring to extend its performance in mildly turbid waters. V. References 1. M. Johannesson, Sheet-of-light Range Imaging, Likoeping Studies in Science and Technology, Dissertations No S.M. Smith, K Ganesan, T. Flanigan, and L. Larquis, The intelligent distributed control system architecture in the Ocean Voyager II and Ocean Explorer vehicles, presented at 9 th International Symposium on Untethered Submersible Technology, Durham, NH K.L. Carder, D.K. Costello, and Zhongping Lee, The use of unmanned vehicles to acquire environmental data in support of mine-counter-measure operations, presented at Fourth Intenational Symposium of Technology and the Mine Problem, May K.L. Carder, D.K. Costello, Lawrence C. Langebrake, Weilin Hou, James T. Patten and Eric Kaltenbacher, Ocean-science mission needs: Real-time AUV data for command, control, and model inputs, in press.