Optical counting platform of shrimp larvae using masked k-means and a side window filter

Abstract

Accurate and efficient counting of shrimp larvae is crucial for monitoring reproduction patterns, assessing growth rates, and evaluating the performance of aquaculture. Traditional methods via density estimation are ineffective in the case of high density. In addition, the image contains bright spots utilizing the point light source or the line light source. Therefore, in this paper an automated shrimp counting platform based on optics and image processing is designed to complete the task of counting shrimp larvae. First, an area light source ensures a uniformly illuminated environment, which helps to obtain shrimp images with high resolution. Then, a counting algorithm based on improved k-means and a side window filter (SWF) is designed to achieve an accurate number of shrimp in the lamp house. Specifically, the SWF technique is introduced to preserve the body contour of shrimp larvae, and eliminate noise, such as water impurities and eyes of shrimp larvae. Finally, shrimp larvae are divided into two groups, independent and interdependent, and counted separately. Experimental results show that the designed optical counting system is excellent in terms of visual effect and objective evaluation.