Abstract In recent years, the tracking model based on the Siamese Network has been widely used in the object tracking field to model the object tracking task as a similarity matching problem, which balances the tracking speed and accuracy. However, there are insufficient robustness, discriminative ability and generalization ability for object deformation and complex background interference. In this paper, an improved Fully-convolutional Siamese Network is proposed. The Triplet Loss function is used as the model objective function instead of logistic loss, and the multi-channel attention mechanism is introduced to make the model pay more attention to the tracking related information and enhance the model discriminating ability. In the offline training phase, an effective data augmentation strategy is used to control the uneven distribution of sample categories and improve the generalization ability of the model. In the tracking phase, the Distractor-aware module is used to transfer the general feature representation domain to a specific object domain, thereby improving model discriminating ability. In experiments, the results on VOT2016 tracking benchmark shows that our model has a significant improvement over the SiamFC tracker in multiple evaluation indicators.

中文翻译：

---

基于改进的全卷积孪生网络的实时目标跟踪

摘要 近年来，基于Siamese Network的跟踪模型被广泛应用于目标跟踪领域，将目标跟踪任务建模为相似性匹配问题，平衡了跟踪速度和精度。然而，对于物体变形和复杂背景干扰的鲁棒性、判别能力和泛化能力不足。在本文中，提出了一种改进的全卷积连体网络。采用Triplet Loss函数代替logistic loss作为模型目标函数，并引入多通道attention机制，使模型更加关注跟踪相关信息，增强模型判别能力。在离线训练阶段，采用有效的数据增强策略来控制样本类别的不均匀分布，提高模型的泛化能力。在跟踪阶段，使用 Distractor-aware 模块将一般特征表示域转移到特定对象域，从而提高模型判别能力。在实验中，在 VOT2016 跟踪基准上的结果表明，我们的模型在多个评估指标上比 SiamFC 跟踪器有显着的改进。