束平 吴洪昊 孙娟 唐晓东

摘要 为推进农业信息化，实现猪的智能化养殖以及对多猪只的智能跟踪，设计猪只检测阶段和跟踪阶段。在检测阶段，该设计提出了基于高斯混合建模和均值分割算法相结合的信息融合算法，有效地解决猪只静止或运动缓慢以及背景噪声对检测结果的影响;在跟蹤阶段，传统粒子滤波算法并不能对猪只重叠进行运动跟踪，对重要性粒子滤波结果进行序列化，并将其结果利用KNN算法进行轨迹跟踪。最后进行了处理试验，结果显示算法真实有效。该成果可用于猪只养殖信息化。

关键词 农业信息化;猪养殖;目标跟踪

中图分类号 S-058   文献标识码 A  文章编号 0517-6611（2021）16-0230-03

doi：10.3969/j.issn.0517-6611.2021.16.060   开放科学（资源服务）标识码（OSID）：

Pig Tracking Based on Improved Particle Filter Algorithm

SHU Ping， WU Hong-hao， SUN Juan et al

（Yancheng Bioengineering Branch of Jiangsu Union Technical Institute， Yancheng， Jiangsu 224051）

Abstract To promote agricultural informatization and to realize intelligent pig breeding and intelligent tracking of pigs， two stages were designed， which were pig detection stage and tracking stage. In the detection phase， the information fusion algorithm based on Gauss mixture modeling and mean segmentation algorithm was proposed to effectively solve the influence of pig static or slow motion and background noise on the detection results. In the tracking stage， the traditional particle filter algorithm could not track the overlap of pigs and was of great importance. The results of particle filter were serialized， and the KNN algorithm was used to track the trajectory. Finally， the processing experiments were carried out. The results showed that the algorithm was real and effective. The results could be used in pig farming informatization.

Key words Agricultural informatization;Pig breeding;Target tracking

农业信息化是整个农业发展的趋势，每头猪的信息从出生到死亡的每一件事都记录在与其对应的唯一号码上，相当于身份证一样，因此对猪运动信息的跟踪也就显得尤为重要。由于猪是一种社会性动物，猪通过打架来分出社会地位，由于猪只运动的复杂性，普通的车辆跟踪算法无法对多猪只进行轨迹跟踪，鉴于此，笔者介绍了基于信息融合的猪只检测算法和结合粒子滤波的猪只踊跃算法，从猪只检测和猪只跟踪2方面进行改进，一方面利用图像信息融合算法实现了猪只的精确检测，另一方面利用重样性粒子滤波的序列化对猪只的运动轨迹进行跟踪，解决了猪只重叠等问题。

1 基于信息融合的猪只检测算法

目……