Abstract:Phytoplankton, a crucial component of aquatic ecosystems, requires efficient monitoring to understand marine ecological processes and environmental conditions. Traditional phytoplankton monitoring methods, relying on non-in situ observations, are time-consuming and resource-intensive, limiting timely analysis. To address these limitations, we introduce PhyTracker, an intelligent in situ tracking framework designed for automatic tracking of phytoplankton. PhyTracker overcomes significant challenges unique to phytoplankton monitoring, such as constrained mobility within water flow, inconspicuous appearance, and the presence of impurities. Our method incorporates three innovative modules: a Texture-enhanced Feature Extraction (TFE) module, an Attention-enhanced Temporal Association (ATA) module, and a Flow-agnostic Movement Refinement (FMR) module. These modules enhance feature capture, differentiate between phytoplankton and impurities, and refine movement characteristics, respectively. Extensive experiments on the PMOT dataset validate the superiority of PhyTracker in phytoplankton tracking, and additional tests on the MOT dataset demonstrate its general applicability, outperforming conventional tracking methods. This work highlights key differences between phytoplankton and traditional objects, offering an effective solution for phytoplankton monitoring.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is the limitations of existing phytoplankton monitoring methods. Traditional methods rely on non - in - situ observations, that is, collecting water samples and bringing them back to the laboratory for manual observation. This method is time - consuming and resource - intensive and cannot analyze the dynamics of phytoplankton in a timely manner. To overcome these limitations, the paper proposes an intelligent in - situ tracking framework named PhyTracker, aiming to automatically track phytoplankton.
### Specific Problems and Solutions
1. **Unconspicuous Appearance**:
- Phytoplankton usually has a tiny size, light color, irregular shape and simple texture, which makes their identification more challenging than that of ground objects.
- **Solution**: The Texture - enhanced Feature Extraction (TFE) module is introduced. It uses dilated convolution and SRM filters to enhance feature extraction, thereby improving the detectability of phytoplankton.
2. **Complex Monitoring Environment**:
- Water samples usually contain impurities scattered throughout the field of view. These impurities are very similar to phytoplankton, which poses a challenge to accurate monitoring.
- **Solution**: The Attention - enhanced Temporal Association (ATA) module is proposed. Through the attention mechanism, it effectively correlates the features between consecutive frames, eliminates the interference of impurities, and distinguishes between phytoplankton and impurities.
3. **Different Monitoring Processes**:
- The tracking of ground objects (such as pedestrians and vehicles) is usually carried out in the field scene using ordinary cameras, while the monitoring of phytoplankton involves extracting water samples from the ocean and gradually analyzing them through a microscope. In this case, the movement of phytoplankton is mainly driven by water currents, and the trajectories are highly consistent.
- **Solution**: The Flow - agnostic Movement Refinement (FMR) module is designed. It restores the movement characteristics of each phytoplankton, reduces the feature confusion of similar movement trajectories between different tracking entities, and retains the original movement offset information, thereby enhancing the sensitivity to individual movement characteristics.
### Summary
PhyTracker solves the key challenges in phytoplankton monitoring through three innovative modules (TFE, ATA, and FMR) and realizes online real - time tracking. The experimental results show that PhyTracker performs better than traditional tracking methods on the PMOT dataset and also demonstrates its wide applicability on the general - purpose object tracking dataset (MOT).
### Formula Display
1. **Cross - Attention Operation**:
\[
CA(Q, K, V)=\phi_q(Q)(\phi_k(K)^T V)
\]
where,
\[
\phi_q(Q)=\text{softmax}_{\text{row}}(Q), \quad \phi_k(K)=\text{softmax}_{\text{col}}(K)
\]
2. **Offset Calculation**:
\[
OX_t = C_{i,j}^X T_{i,j}^X, \quad OY_t = C_{i,j}^Y T_{i,j}^Y
\]
where,
\[
T_{i,j}^X(l)=(l - j)\times s, \quad 1\leq l\leq w'
\]
\[
T_{i,j}^Y(k)=(k - i)\times s, \quad 1\leq k\leq h'
\]
Through these improvements, PhyTracker provides an effective solution for the efficient monitoring of phytoplankton.