Intent Recognition Research Based On Machine Vision

The ability to recognize the plans and goals of other agents enables humans to reason about what other people are doing, why they are doing it, and what they will do next. This fundamental cognitive capability is also a critical ability of the interpersonal interactions. With the help of machine visual to understand real life human’s intention and inference, this technology has broad application prospect and economic value in the area of intelligent monitoring, natural human-computer interaction, human motion analysis, virtual reality.This paper’s study revolves around human intent recognition which based on visual device, mainly include three aspects: human action recognition, related object positioning and intent recognition algorithm. Human action recognition is the basis of the visual human behavior and intention studying. This paper uses the visual sensor device to capture color and depth information from image in real situations, then construct human body skeleton parameters which independent with the distance and direction. Through the Logistic regression to classify human behaviors, and the classification of the motion information is accurate. Intent recognition of human behavior need to understand not only the human actions, but also relevant object information. But different objects have different characteristics, it is not appropriate to analysis and identify objects single from one perspective. This paper propose a kind of object recognition algorithm based on D-S evidence theory for robot would be provided, merge together SIFT and color histogram features to recognize the object. And the paper proposes a proof technique to distinguish whether target object is the known object, that improve the robustness of the recognition. Because of the randomness of human intentions, there are some problems such as abandoning plans, partially ordered plans and observations of failed actions which serious influenced the recognition results. Based on the idea of rough set, this paper set up a hierarchical model to express human intents, then using stratified hierarchical reduction method to extract the intentions of the core information, finally generate the intent decision set, it can reduce the interference result from the above problems.