Echoview software for tracking and counting fish detected with a dual frequency identification sonar (DIDSON) imaging sonar: A flexible approach to signal enhancement and object detection prior to target tracking

Abstract

The data from the DIDSON (Dual frequency IDentification SONar) are of startling resolution, and show swimming fish in unprecedented detail. Although the fish objects are visually striking, object detection must be achieved in post processing and is a challenge when compared with conventional split beam systems. Data processing can be applied to automatically remove static objects and noise and to enhance the signal from moving objects in DIDSON data. Moving objects are then detected as targets within individual DIDSON frames and converted to X−Y−Z-time position data suitable for input to a tracking algorithm. The initial data manipulation for noise removal and object detection is achieved through the application of successive modular operators. The results of each operation can be viewed as a ‘‘virtual variable’’ enabling each step to be scrutinized and optimized by the operator. This modular approach to data manipulation, target detection, and fish tracking will enable the rapid development of new techniques and tools. It is a ‘‘future-proof’’ data processing solution.