Arrow Game Projection Interactive System Based On Improved YOLOv3 Algorithm

With the continuous development of human-computer interaction technology,projection interaction technology has begun to enter people's field of vision and has been widely used in many fields.In view of the problems of the current projection interaction system,such as the large number of system equipment,high hardware cost,and low recognition algorithm robustness,this paper combines a neural network technology and image processing technology in machine vision,and proposes a projection interaction based on the improved YOLOv3 algorithm Method,and developed a projection interactive system under the bow and arrow game scene.The system has the characteristics of simple hardware implementation and accurate interaction.The main work of this article includes the following three parts:(1)Research on arrow positioning algorithm.The key to the realization of the projection interactive system is how to accurately obtain the interactive information in the complex background.In this paper,in order to reduce the system's dependence on equipment,the camera is used as a sensor to obtain interactive information.In order to solve the problem that traditional image processing techniques are vulnerable to scene constraints,this paper studies the convolutional neural network technology,improves its loss function based on the YOLOv3 object detection algorithm,and reduces the characteristics of YOLOv3 through an adaptive pruning algorithm Extract the part,and finally use the improved YOLOv3 algorithm to complete the real-time detection of the arrow position.(2)Research on the automatic positioning and correction algorithm of the projection area.Accurate positioning and correction of the projected scene area in the image is the prerequisite for achieving precise interaction between man and machine.In this paper,a straight line detection algorithm based on large-scale Laplace filter and straight line filter operator is used to detect the boundary of the projection area.This algorithm effectively solves the problem of boundary blur.In this paper,by studying the coordinate transformation relationship,the homography transformation is used to solve the mapping relationship from the image coordinate system to the game scene coordinate system.(3)Realization of the projection interactive system for bow and arrow games.In terms of hardware,the system consists of cameras,projectors,and computers.Compared with other projection interaction systems,this system has fewer devices and is easier to promote.In terms of software,under the Windows system,this article implements a projection interaction system software that encapsulates the core algorithm of this article,and develops a bow and arrow game scene to show the interaction effect.Finally,the actual test of the projection interactive system was carried out.The experimental results show that the system can quickly and accurately detect the arrow target and achieve a good interactive effect.