Prof John Watson FIEEE, VP OCEANS
Engineers at the University of Aberdeen, Scotland, have developed one of the most advanced subsea holographic cameras in the world, capable of rapid 3D imaging of marine organisms and microparticles. Affectionately dubbed “weeHoloCam”, it is one of the most compact devices of its type, as well as offering rapid image capture and processing of the reconstructed holographic images.
weeHoloCam was developed and funded by the Defence & Security Accelerator (DASA) of (UK) DSTL and is adaptable for a variety of subsea applications – for example, studying marine habitats and organisms in precise detail, or for monitoring ocean microplastic pollution.
The project is led by Dr Thangavel Thevar (MIEEE) from the University’s School of Engineering, along with Emeritus Professor John Watson (FIEEE), Dr Nick Burns, Dr Amer Syed, and consultant Mike Ockwell. (Hi-Z 3D).
A key feature of weeHoloCam (Fig 2) is its size (hence the name!) – at just 9 cm in diameter, 60 cm long and weighing 3.5 kg, it is ideally suited for rapid deployment on remotely operated underwater vehicles (ROVs), autonomous underwater vehicles (AUVs) or towed and (as was done in these trials) lowered to specific depths. It has been designed for deployment down to 500 m, but only used to 50 m in these trials.
The holocamera consists of two water-tight housings: one containing the illuminating laser and the other containing the CCD image sensor. Holograms are recorded in the so-called “in-line” mode whereby the collimated laser beam passes through a window into the water, through another window and onto the electronic sensor in the other half of the housing. Particles in the water scatter light towards the sensor where it combines with background light from the laser to form the hologram. The recording image volume is 12 cm3 (8.5 mm by 7.1 mm and 200 mm path length).
The holograms would normally be reconstructed in software using an algorithm such as the Angular Spectrum Method to form an image at any specific plane in the recording volume. Each hologram can be considered as a stack of thousands of 2D images, However, rather than performing this in software, we have implemented the algorithm using FPGA hardware to give a dramatic increase in processing and reconstruction speed to rapidly reconstruct these 2D images, and to extract focused particles to create 3D versions in almost real-time.
Using a pulsed laser diode, holograms are recorded at a rate of up to 20 Hz. The processing suite autofocuses particles as the holograms are captured in real-time, and stores copies of focused particles as bitmap images. Processing speed is about 1 s to 4 s per 5 MB hologram (2464 x 2056 pixels) depending on the particle density. Fig 3 shows a screen dump of the processing system output.
For its first deployment, weeHoloCam was inserted into a submersible frame (Fig 4) and lowered by rope to depths of 50 m in the North Sea, about 2 miles east of Aberdeen Harbour, from the catamaran Sea Cab. We recorded around 100,000 holograms over four dives totalling about 90 mins of actual recording time.