1. 研究目的与意义（文献综述）

1.1 purpose andmeaning

in the machine vision algorithmand application has been the vigorous development of contemporary society,image processing and image processing based on real-time behavior recognitionhas become a very popular research topic. in recent years, with the rapiddevelopment of the internet and the popularity of mobile terminals, theexplosive growth of video data to the video data processing has brought newchallenges. human behavior recognition is the basic problem and key link ofvideo processing, which directly affects the research of video contentcomprehension and scene comprehension. behavior recognition as part of thebehavior analysis, has a wide range of application background, such asintelligent monitoring, virtual reality, human-computer interaction, videocoding, motion analysis and so on. this paper studies the real-time behaviorrecognition algorithm based on image processing, which means that thesealgorithms have a wide range of application fields and practicality, and havegreat development prospects in video surveillance, human-computer interface,human-computer interaction, content-based video retrieval the for intelligentdriving of modern cars, human behavior recognition also plays an importantrole, it can be used as a basis for the operation of the adas system. in thispaper, we study the real-time behavior recognition algorithm, and can activelyexplore the intelligent driving of modern automobile, and provide somereference for other aspects of behavior recognition research.

1.2analysis of the status quo of research at home and abroad

2. 研究的基本内容与方案

2.1 basic content:

a. study the methodsand problems of behavior identification

since the 1970s and1980s, behavior recognition has entered its initial research stage, and so farit has evolved from a simplified motion identification to a complex behavioranalysis that is close to a real scene. the 1970s and 1980s were the initialstages of behavioral analysis. during this period, the focus of the study wason the principle of human understanding of the movement, such as 1975.johansson demonstrated that human behavior can be described by using only a fewpoints on the human body. this method of describing the behavior of the bodystructure based on the behavior description algorithm has played a veryimportant guiding role. in the late 1980s, nagel proposed a four-layerdefinition of behavior, further deepening the understanding of behavior in thefield of image analysis. in the 1990s, the related research of behavioranalysis was developed step by step. during this period, some influentialresearch ideas were put forward. for example, the hidden markov model wasintroduced into behavior recognition, and the behavior recognition method basedon state space was started. template matching behavior analysis algorithm isproposed, behavior recognition research methods therefore goes further.however, the study of this period is often limited to the perspective of fixed,simple environment, the rules of the situation. after 2000, the research of allaspects of behavior analysis has been rapidly and deeply developed, theresearch of the angle of behavior recognition has get some research results.however, the current research still focuses on the behavior identification ofsimple rules in finite classes or the detection of abnormal behavior inspecific scenes. few algorithms involve high-level understanding of complexbehaviors. if behavior analysis is wanted to be applied in the real scene,there are still many serious problems.

3. 研究计划与安排

4. 参考文献（12篇以上）

[1]郑胤,陈权崎,章毓晋. 深度学习及其在目标和行为识别中的新进展[j]. 中国图象图形学报,2014,02:175-184.

[2]田国会,尹建芹,韩旭,于静. 一种基于关节点信息的人体行为识别新方法[j]. 机器人,2014,03:285-292.