Cross Mul-Strip Net: advanced identification for enhanced understanding of algae biological communities

View at publisher Summarise with VIRAT
Authors: YueXia Xu, Chenyang Yin, Qiushuang Li
Cited by: 0
References: 24

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Abstract With the increasing impact of human activities and climate change on aquatic environments, the eutrophication of water bodies, particularly lakes and rivers, has been worsening at an alarming rate, leads to complex and unpredictable distributions and changes in algal biocenoses. Algal biocenoses, as a crucial component of the ecosystem, play a pivotal role in maintaining aquatic ecological balance, providing ecosystem services (such as water purification and nutrient cycling), and supporting biodiversity. Due to the great diversity of aquatic organisms, their generally elongated morphological forms, significant size disparities, and the minute size of many algal species, traditional identification methods struggle to comprehensively and accurately cover all species. To address the aforementioned issues and achieve precise target detection of algal biological communities, this paper proposes a Cross-Multi-Strip Network (CMSN). Firstly, a Cross-Multi Block (CMB) module is introduced to enhance the network’s multi-scale perception capability, effectively tackling the challenge posed by the complex morphological variations and scale changes of algae. Secondly, a new Cross-Strip Block (CSB) module is proposed, which effectively extracts features of elongated objects through multi-scale strip convolutions. Finally, the Embed-wise Pooling Attention structure (PATT) based on the Transformer architecture and four parallel detection heads are incorporated, significantly enhancing the model’s ability to perceive and localize targets of different sizes. Through validation on public datasets and a self-built database, this model achieves the highest accuracy, and a detailed comparison with mainstream models is provided. In addition, we verified the effectiveness of each model component through ablation experiments.

Fields of Study & Subjects

Advanced Neural Network Applications Water Quality Monitoring Technologies Remote-Sensing Image Classification

Publication details

Journal / source
Engineering Research Express
Volume
8
Issue
7
Pages
075326–075326
Publication date
23 March 2026
Language
EN