Research On Hand Gesture Segmentation Algorithm With Complex Background

Human-computer interaction(HCI)technology provides a good and convenient way for information communication between people and computers,thus has become a research hotspot in the current computer field.Visual gesture recognition(HGR),as a widely-recognized technology,can be applied to human-computer interaction scenes in terms of concept and practice.Specifically,the vision-based gesture recognition technology refers to acquiring a video image containing an operation gesture by a capture device such as a camera,and then performing corresponding processing on the video image,such as gesture segmentation,gesture feature extraction,gesture feature classification,and gesture tracking,etc.,thus enabling information exchange between the user and the smart device.Gesture segmentation is the initial key step of the gesture recognition algorithm,and ensuring its good effect is a prerequisite for accurate gesture recognition.The methods of gesture segmentation are varied,but from the point of view of practice and application,almost without exception,these methods still face enormous challenges in accuracy,stability,and speed.In a complex application scenario,cluttered backgrounds and non-uniform lighting will have a huge impact on the effect of gesture segmentation.In addition,gestures are non-rigid objects,and the diversity of their outlines,angles,sizes,and positions also brings difficulties for accurate segmentation.Therefore,how to integrate effective target detection methods and prior information of gestures In a complex application scenario,and design an efficient gesture region segmentation algorithm to overcome the above problems is the goal that this article should focus on.Based on the analysis of previous researches,this paper proposes a static gesture segmentation algorithm based on saliency and skin color detection and a static gesture segmentation algorithm based on dark channel priors.? A static gesture segmentation algorithm based on saliency and skin color detection.For the backgrounds in complex environments(indoor/outdoor,with/without near-skin background interference,different hands,etc.),SLIC-based image pre-segmentation,a pixel-level gesture region detection method,a regional level gesture region detection method,a Bayesian framework-based fusion of multiple saliency maps have been used successively.Finally,through the control experiments with other algorithms,the high stability and practical value of the algorithm are verified.? A static gesture segmentation algorithm based on dark channel priors.Aiming at the characteristics of image background and lighting conditions under the"projector-camera system",a static gesture segmentation algorithm framework based on dark channel priors is proposed,including definition of dark channel prior,pre-segmentation based on graph representation,and the fusion of saliency map and segmentation map.Finally,through the self-built database and the control experiment based on this database,the obvious advantages of the algorithm in the"projector-camera system" environment are reflected.In this algorithm,the "dark channel prior" defined by us has a very good effect on the positioning of the gesture area,which provides a useful idea for future related research.