论文部分内容阅读
针对视频序列的稳健性目标跟踪问题,提出一种基于卷积神经网络(CNN)与一致性预测器(CP)的视觉跟踪算法。该算法通过构建一个双路输入CNN模型,同步提取帧采样区域和目标模板的高层特征,利用逻辑回归方法区分目标与背景区域;将CNN嵌入至CP框架,利用算法随机性检验评估分类结果的可靠性,在指定风险水平下,以域的形式输出分类结果;选择高可信度区域作为候选目标区域,优化时空域全局能量函数获得目标轨迹。实验结果表明,该算法能够适应目标遮挡、外观变化以及背景干扰等复杂情况,与当前多种跟踪算法相比具有更强的稳健性和准确性。
Aiming at robust target tracking problem of video sequence, a visual tracking algorithm based on Convolutional Neural Network (CNN) and Consistency Predictor (CP) is proposed. This algorithm constructs a two-input CNN model, extracts the high-level features of the frame sampling area and the target template synchronously, uses the logical regression method to distinguish the target from the background area, embeds the CNN into the CP framework, and evaluates the reliability of the classification result by using the algorithm random test , And output the classification result in the form of domain under the specified risk level; selecting the high-confidence region as the candidate target region and optimizing the global energy function in the space-time domain to obtain the target trajectory. The experimental results show that the proposed algorithm can adapt to complex situations such as target occlusion, appearance changes and background interference, and it is more robust and accurate than many current tracking algorithms.