Estimating Norway lobster abundance from deep-water videos: an automatic approach

Mosaics for Nephrops detection in underwater survey videos

“…A K-Nearest Neighbour classifier is then used to select the respective Nephrops from these candidate regions. Our final decision accuracy at 87.5% recall and precision shows a corresponding 31.5% and 79.4% improvement compared with previous work [1]. …”

“…A segmentation approach is used to accomplish this task because these images are usually very blurry, and detecting parts of the objects with techniques such as edges [1], might not be effective in some cases. Additionally, the scientifically important features such as the diameter of the creature is most effectively extracted from the entire region.…”

“…This non-parametric classifier is used because of its simplicity. The key advantage this scheme offers in comparison to the previously used Decision Tree scheme developed by Lau et al [1], is that it incorporates the use of training data into its classification procedure. The use of this data not only allows the system to identify a large variety of Nephrops, but also facilitates easy adoption to new data sets.…”

“…The proposed lobster recognition system is accomplished in four stages involving: i) Mosaic Generation, ii) Object detection and Grouping, iii) Feature Choice and Extraction, and iv) Classification. Each of these stages are discussed in the following sections, after which a comparison with the previous work by Lau et al [1] is given.…”

“…This new area of research has recently been tackled in the literature by authors Lau et al [1]. In their approach, lobster detection is addressed using video, where candidate regions are detected in the gray scale images via their edges, and then classified with a decision tree framework.…”