201. A Positioning Method for Apple Fruits Based on Information Fusion
- Author
-
Chen Yuan, Lihua Zheng, Changyi Xiao, Mai Chunyan, and Li Minzan
- Subjects
Mean squared error ,Computer science ,business.industry ,ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION ,Rgb image ,Information fusion ,GNSS applications ,Position (vector) ,Assisted GPS ,Global Positioning System ,Computer vision ,Information acquisition ,Artificial intelligence ,business - Abstract
In order to provide effective data for yield estimation of orchard, a positioning method for apple fruits based on information fusion was proposed. An information acquisition system which consists of a Microsoft Kinect, a GPS (M600 GNSS receiver, Sinan corp.) receiver and an attitude sensor (MPU6050, Inven Sense corp.)was designed, and the world coordinates calculation method for apple fruits in a RGB image was developed according to this system. In this information acquisition system, the Microsoft Kinect (Kinect for Windows V2 i¼Œ Microsoft) is used to acquire both the RGB image and depth information image of apple trees in the apple orchard, the GPS receiver is used to get the position information of the Kinect, and the attitude sensor is used to obtain the attitude information of the Kinect. First, RGB image and depth image is fused to calculate the coordinates of centers of apples using relative positioning model. Secondly, the position and attitude information of the camera are added to establish an absolute positioning model to calculate the world coordinates of apples based on three-dimension coordinate converting method. Thus, each apple gets only one space position in the world. The experiment results showed that the mean error of the relative positioning reached to 3.75cm; and for absolute positioning, the mean distance error of the longitude was 117mm i¼Œ that of the latitude was 437mm i¼Œ and that of the altitude was 145mm.
- Published
- 2015