Identify a Specified Fish Species by the Co-occurrence Matrix and AdaBoost - Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing - page 97

Fig. 1 Fish examples for study

Before staring the Learning process, we check the created positive image by
human eyes and determine whether or not it is a reasonable data set for studying.
e. Processing of Machine Learning
After getting a seasonal data set, we feed them in to a haartraining com-
mand, finally we get the learned cascade feature file.

f. Object Detection

Now we can detect the target object by using the feature file created by step e.

The detection result is shown in Fig. 2. All the fishes are detected correctly, and
the deferent species are not detected. Due to the experiment is executed under an
ideal environment, there are still thought many miss detection exist when for practi-
cal use. Here, we think this step can be correctly accomplished, then go to the next
steps.

3 Targeted Invasive Alien Fish Species Decision

In order to identify the extermination target species, black bass, from other fishes,
we proposed a quantifying the texture analysis approach from body surface pattern
of fish.