Abstract: With continuously extending of pig breeding scale and quantity in China, husbandry informatization will be an important supervise model for pig breeding industry at present and in the future. As an effective support technology for information processing, computer vision technology has provided a behavior recognition method with automatic, non-contact, low-cost, high profit and yet non-injury and non-stress for animals. It is capable for assessing pig health condition, timely preventing and diagnosing diseases. This paper introduced pig behavior recognition vision system; reviewed the application of vision technology in pig target extraction and individual recognition, behavior recognition analysis, and algorithm; and analyzed the existing problems in accuracy, effectiveness and applicability of existing visual system and behavior recognition method. Before ending, the paper put forward several strategy for improving visual system, and key research direction for applying computer vision technology in pig individual and behavior recognition.

Key words: computer vision, pig, target segmention, individual recognition, behavior recognition, tracking

摘要： 中国生猪养殖规模和数量在不断扩大，养殖信息化是今后生猪饲养监管的重要模式。计算机视觉技术作为信息处理的有效辅助技术，提供了一种自动化、非接触式、低成本、高收益且对动物无伤无压力的行为识别方式，可用于考量生猪健康状况并及时预防和发现疾病。介绍了猪行为识别视觉系统，回顾了视觉技术在生猪目标提取和个体识别、行为识别分析中的应用及算法，分析了现存视觉系统及行为识别方法在准确性、有效性与适用性方面存在的问题，提出了视觉系统的改进策略及今后计算机视觉技术在猪体识别和行为识别应用中的重点研究方向。

关键词: 计算机视觉, 猪, 目标分割, 个体识别, 行为识别, 跟踪