Enteromorpha Prolifera Detection in Aerial Images Based on Image Retrieval

In this paper, a simple image retrieval based enteromorpha prolifera detection method is introduced. First, the enteromorpha images are converted into NTSC space from RGB color space and are then separated into enteromorpha and background class by means of k-means algorithm. Then, all those images which only consist of enteromorpha class are stored as sample images. Next, in the stage of detection, an n×n representative patch is chosen from the detecting and a sample image respectively. And Sum of Squared Difference (SSD) of these two patches is calculated. If the value of SSD is smaller than a threshold which obtained by doing trial and error test, the detecting image is regarded as containing enteromorpha. Otherwise, the next sample image will be checked in the same way. If all sample images do not meet the requirement, as a result, there is not enteromorpha in the detecting image.