AI and Stereo Cameras in Offshore Salmon Aquaculture

Modern offshore aquaculture relies heavily on underwater stereo cameras and artificial intelligence (AI) to manage salmon populations efficiently and sustainably. By deploying these advanced optical systems inside sea cages, fish farmers can accurately estimate salmon biomass, automatically count parasitic sea lice, and prevent costly feed waste. This article provides a straightforward breakdown of how 3D imaging and machine learning algorithms work together beneath the waves to automate these three critical tasks without stressing the fish.

Estimating Salmon Biomass with Stereo Cameras

Traditional biomass estimation requires netting fish and weighing them manually, which causes high stress and mortality. Underwater stereo cameras eliminate the need for physical handling. A stereo camera system uses two lenses placed a set distance apart to capture simultaneous images of the salmon, mimicking human binocular vision.

AI-driven computer vision models process these dual images to create a 3D depth map. The software identifies specific anatomical landmarks on the salmon, such as the snout, tail fork, and dorsal fin. By calculating the precise distance between these points in three-dimensional space, the system measures the exact length and height of the fish as it swims past. The AI then applies established biological growth formulas to convert these physical dimensions into an accurate weight estimate. Tracking thousands of fish daily allows the system to calculate the total biomass and weight distribution of the entire cage in real time.

Automated Sea Lice Counting

Sea lice are a major threat to farmed salmon, and strict regulations require farmers to keep parasite levels below specific thresholds. Manual counting is labor-intensive and only samples a tiny fraction of the population.

To automate this, high-resolution cameras continuously scan the salmon as they swim through the cage. Deep learning algorithms, specifically trained on hundreds of thousands of annotated underwater images, analyze the video feed frame by frame. The AI isolates individual fish and scans their skin for the distinct visual signatures of sea lice, such as shape, color, and attachment location. The system is precise enough to differentiate between harmless debris, skin blemishes, and actual parasites. It also categorizes the lice by their life stages—distinguishing between mobile, adult female, and attached stages. This provides farmers with continuous, statistically significant data on infection rates, allowing for targeted and timely treatments.

Stopping Feed Waste

Feed accounts for the largest operational cost in salmon farming, and uneaten pellets sinking to the ocean floor cause both financial loss and environmental damage. AI and camera systems prevent this by automating the feeding process based on real-time fish behavior and pellet tracking.

During feeding, cameras monitor the water column. AI algorithms track two main metrics: the speed and movement of the salmon, and the presence of sinking feed pellets. Active, fast-swimming fish near the surface indicate high appetite. As the salmon become full, their swimming speed decreases and they disperse deeper into the cage. Simultaneously, object-detection algorithms scan the water beneath the feeding zone for uneaten pellets. The moment the AI detects a specific threshold of pellets sinking past the fish, it sends a signal to the automated feeding system to slow down or stop dispensing food immediately. This closed-loop system ensures the salmon receive optimal nutrition while reducing feed waste to near zero.