Figure 1.
(a) Experiment environment; (b) Fish shape model.
Figure 2.
The diagram of the proposed method.
Figure 3.
The illustration of the ellipse parameters.
(a) The raw image of the fish head region; (b) (xo,yo) denotes the extreme point and the violet color shows the grayscale distribution of the extreme point region. Variables length, width and represent the long axis, short axis and angle of the fitted ellipse respectively; (c) The variable contrast in the direction of the z-axis represents the contrast change of the ellipse and its surrounding region.
Figure 4.
Candidate constraints based on width, contrast and angle.
Figure 5.
(a) The model of compensation window; (b) The segmentation model of matching region.
Figure 6.
The process of feature matching.
Figure 7.
Trajectory linking based on time and distance.
Table 1.
Parameter settings in the test process.
Figure 8.
(a) Example of frame image illustrating the detection results of the fish head regions for a group of 40 fish; (b) Some examples of occlusion events efficiency resolved and a rare case where the detection failed.
Table 2.
Detection performance on different groups.
Table 3.
Compared methods.
Figure 9.
Performance of compared methods on two evaluation metrics.
(a) TCF; (b) TFF. As fish density increases, tracking performance of all five methods falls. In comparison, the proposed method offers highest TCF values and lowest TFF values, indicating its performance is the best among the compared methods.
Figure 10.
Tracking results on different groups with 16.7 seconds as duration.
Left column: trajectory acquisition results with the time axis. Right column: trajectory acquisition results without the time axis. (a) A1 (10 fish); (b) A2 (20 fish); (c) A3 (40 fish).