A Tracking Scheme for Norway Lobster and Burrow Abundance Estimation in Underwater Video Sequences
; Yee, L.
; Fonseca, P.F.
; Campos, A.C.
A Tracking Scheme for Norway Lobster and Burrow Abundance Estimation in Underwater Video Sequences, Proc International Workshop on Advanced Image Technology - IWAIT, Tainan, Taiwan, Vol. ., pp. . - ., January, 2015.
Digital Object Identifier: 0
Underwater imaging constitutes an alternative approach for stock assessment in some commercial fisheries. Herein, the Norway lobster, Nephrops norvegicus, a benthic burrowing crustacean, is used as our case study. An unified tracking scheme is presented to track our object of interest, either the lobsters themselves or their burrows, across video footage obtained from an underwater Remotely Operated Vehicle (ROV) during a survey on deep-water crustacean grounds off the Portuguese South coast. The objective of the tracking module is to prevent the over-counting in consecutive frames thus ensuring a correct estimation of the actual abundance of Norway lobster and their burrows appearing in the whole video footage. The proposed tracking scheme involves two different methods for lobster(s) and burrow(s) respectively. Lobster candidates are tracked using a particle filter-based
Method since the lobsters are moving freely under the moving camera. Contrastingly, the burrow locations, although static, are under dynamic observation. Thus burrow candidates are tracked based on the motion model of the observer, where the affine transformation model is used as the motion model. The latter is estimated using a set of matching feature points over pairs of consecutive images by the iterative Lucas-Kanade’s optical flow method. The experimental results demonstrate the efficiency of the proposed method based on MOTA metric.