The fish is unwell, SingTown Technology's OpenMV knows first.

In aquaculture and ornamental fish management, common diseases such as fish skin diseases, fin and tail rot, and water mould, if not detected promptly, can severely impact the health and survival rates of fish populations. Traditional detection primarily relies on manual visual observation, which is not only inefficient but also susceptible to interference from turbid water, varying light conditions, and fish movement, making it difficult to accurately and timely assess pathological conditions.

To this end, aquaculture enterprises have adopted SingTown Technology's OpenMV intelligent cameras to establish a machine vision-based fish health monitoring system for real-time identification of abnormalities on fish surfaces. The system is installed on the exterior glass walls of fish tanks or aquaculture ponds, capable of monitoring fish activity within a one-metre range and performing automatic identification and lesion detection on small goldfish approximately five centimetres in length.

The system aims to detect pathological features such as white spots, white fuzz, water mould, or fin rot on fish surfaces through image recognition, thereby providing early health warnings and assisting breeders in timely intervention.

Aquaculture enterprises have developed a fish health identification system based on the OpenMV smart camera, integrating AI image recognition, colour segmentation, texture analysis, and feature comparison algorithms to accurately detect fish body lesions underwater.

The system captures real-time images of the aquarium via a high-definition camera. On the device side, AI utilises a colour recognition algorithm to extract the fish body area and filter out background reflections. The algorithm then identifies abnormal reflective areas, such as white spots or water mould, by analysing the brightness distribution and texture continuity on the fish's surface. Simultaneously, the algorithm can assess the presence of fin rot, fin decay, or mould attachment by combining morphological features, such as irregular fin edges, local whitening, or blurring.

To accommodate the movement of fish in water and fluctuations in light, the system incorporates dynamic threshold adjustment and multi-frame fusion algorithms, providing stable judgment for short-term continuous footage and effectively reducing false alarms. During testing, the recognition distance was maintained within one metre, and the system can simultaneously identify multiple individuals, achieving an accuracy rate exceeding 92% in detecting water mould disease in small fish such as goldfish. Detection results can be output in real-time to the host computer or app, with alerts for suspected lesions sent via the network.

SingTown's OpenMV smart camera is easy to operate, has low power consumption, and offers strong real-time detection capabilities. It is suitable for deployment on edge devices in aquaculture, pet aquariums, and university research laboratories, helping users achieve timely, visual, and intelligent monitoring of fish health.