多特征Camshift和Kalman滤波结合的猪只智能跟踪

Pig intelligent tracking based on multi-feature Camshift algorithm combined with Kalman filter

为了实现猪只的智能跟踪,提出了一种多特征Camshift和Kalman滤波结合的猪只跟踪算法。首先采用最大类间方差法从背景中分割出猪只,从而统计出猪只的颜色特征和纹理特征;然后获取当前帧图像的联合特征概率分布图,利用Camshift算法求得猪只的位置;最后使用Kalman滤波预测猪只下一帧位置,实现多猪只跟踪。研究结果表明,猪只跟踪算法具有较强的鲁棒性,且更好满足实时性的要求。研究结果可为猪只健康养殖提供技术上的支持。

In order to implement the intelligent tracking of pig, pig intelligent tracking algorithm based on multi-feature Camshift combined with Kalman filter is proposed. Firstly, Ostu method was used to divide pig from background, and add up the color feature and texture feature of pig; Then, multi-feature probability distribution of the current frame was produced, and Camshift algorithm was applied to compute pig position; Finally, the next frame＇s positions are predicted by Kalman filter for multi-pig tracking. The experimental results showed that this pigs tracking algorithm is better in term of robustness and real-time needs. This research can provide technical support for healthy breeding of pigs.