A modular, multi-channel imaging system for marine mammal, target detection
Electro-optical/IR (EO/IR) sensor packages on unmanned aircraft systems (UASs) have been used for whale detection, general situational awareness, and high-contrast target detection and tracking. Although recent work has incorporated automatic target-detection algorithms, performance has been held back by the limited video data quality as dictated by the bandwidth of the UASs' data links. Low-contrast targets of interest in the littoral environment, such as near-surface explosives and obstacles, are not detectable with standard-quality video, automatically or otherwise.
We have developed and tested a series of modular EO/IR sensors and demonstrated them in a number of applications. One of the applications, marine mammal detection and tracking, has been performed over the last few years in Hawaii, California, and over the eastern coastline of Canada.1 The detection of submerged targets, such as whales, ultimately can require just two narrow spectral bands that are used to subtract off the light (clutter) reflecting from the sea surface. The result is not only clear images of subsurface targets, but also contrast-enhanced targets that can be detected automatically (see Figure 1).
Historically, airborne surveys for marine mammals were conducted to monitor populations, identify individuals, and observe animal behaviors. These surveys were done routinely from aircraft flying at altitudes much lower than 1000 feet with scientists on board taking pictures and counting animals. These flights can be risky due to the low altitudes, as well as costly. If the observers (and even the pilot) can be removed from the aircraft, and if the aircraft can be flown at higher altitudes, the risks can be eliminated.
We have tailored our EO/IR systems for general littoral observations. Two systems are used together—the multi-spectral EYE-5g system along with the hyper-spectral EYE-HSswir system. The EYE-5g uses narrowband filters in the blue, green, red, and near-IR spectrum. The blue and green bands have a water-penetration capability similar to the whale-detection algorithms. The red and near-IR bands are used in a vegetation index to separate out sea grass and kelp clutter that is common in southern California's ocean waters. The hyper-spectral system can monitor and assess oil seeps and slicks, which also can be southern California issues. In addition, both systems can be used to directly detect anything on the ocean surface using specific signatures such as the international orange of floatation devices and survival suits, and the oil sheen and debris fields from aircraft or boating accidents.
Military missions over the last decade have been hampered by the ease at which insurgent elements in asymmetric warfare can change and redirect their threats, and the difficulty of deploying new sensor packages to counter those changes. Although littoral threats have not played a role in recent conflicts, we still need to apply the lessons learned in desert and mountain environments to littoral surveillance and reconnaissance.
Development of multi-channel systems, including multi-spectral and hyper-spectral ones, has shown that these systems can address a wide range of potential threats. They are designed to be used with automatic target detection, and thus reduce both the bandwidth needed to transmit imagery and the manpower needed to assimilate it. We have devised a candidate process that packages these multi-channel systems into a size, weight, and power envelope suitable for small UASs and capable of addressing a variety of commercial and military missions in the littoral environment.
We have developed a series of modular, multi-channel imaging systems incorporated into small, six-inch TASE300 Cloud Cap gimbals.2 Three of the systems, the EYE-5f (a three-channel visible near-IR multi-spectral with a long wavelength IR microbolometer), the EYE-5g (a four-channel multi-spectral), and the EYE-HSswir (a near-IR/short-wavelength IR hyper-spectral with a 75mm lens), all have a zoom video system (see Figure 2).
The gimbals have common electrical and mechanical interfaces. They can connect to the same data acquisition and processing unit and thus are interchangeable. The cameras in each gimbal can also be easily changed, as can the associated optics (lenses, filters, polarizers), so tailoring a system for specific applications is a straightforward process.
Figure 3 shows the GRIP-Delta general-purpose ruggedized integrated processor from Vision4ce that is used to control all sensors and perform a series of baseline spectral algorithms. The current processing stream includes sensor radiometric calibration, multi-spectral band-to-band registration, anomaly detection, spectral-matched-filter detection, and post-detection processing that outputs real-time target detection decisions.
The systems can work with standard sensor packages on board small, unmanned aircraft or stand alone. They do not need additional data link bandwidth and will not add work for the data analyst. The systems have shown they can help in border patrol, search and rescue, and littoral reconnaissance operations, and can detect (in real time) and extract data from subsurface targets in maritime operations.
Over the next year or so we will continue to pursue military applications for these systems. We also will be working with SciFly LLC, an aircraft services company in San Diego, to further use the technology in both of the above applications and in new uses such as fish spotting for the San Diego Tuna Fleet and basic research in physical oceanography, such as monitoring chlorophyll and sediment loads in coastal waters.
