The Research Of Image Recognition System Based On The Tree Images Taken By A CCD Camera Mounted On A Theodolite

Forest survey is an important basis for the development of policies and scientific management for the forestry sector, its workload and labor intensity are very intensive. With emerging discipline based on machine vision evolving, high-precision, real-time, non-contact measurement technology has been applied. However, this technology is not widely employed because of its high cost, higher environmental requirements, and other many factors. To solve these problems, a kind of system for measuring DBH and tree height based on principle of binocular vision, using a CCD camera mounted on a theodolite was presented. The forest image recognition system can be divided into three sub-systems, including system design, image processing and trees measurement.The main research of the system design is machine vision, on the basis of this theory, build relationships in the coordinate system of the spatial point, and propose a software system, in which CCD and theodolite as a data input and output parameters. The camera was calibrated, internal and external parameters was got; a mathematical model of the system was established; the mathematical model to analyze the accuracy of the system was established, and the results showed that the system structure parameters should be held in the vicinity of1.3and the focal length of the camera as the use of large during the measurement.The main research of image processing function have achieved reading images and processing operations in the Visual C++6.0platform, including algorithms of image enhancement, image segmentation processing, mathematical morphology, etc. A feature extraction method was proposed, based on median filtering, thresholding, opening operation and edge detection.The main research of trees measurement function have achieved extraction and matching of feature points after solving mathematical models and image processing. An interactive feature matching algorithm was proposed to improve the efficiency and reduce the rate of false matches. The DBH and height of trees were measured by this method. Finally, large amounts of data of trees were taken in the experiment. The parameters of trees are calculated by the system, and compared with the actual measured value, the mean relative error was less than5%, which can completely meet the accuracy requirements of technical regulations for inventory for forest management planning and design(GB/T26424). Therefore, the effectiveness of the above algorithms were proved, and the system used in parameter measurements of forest is feasible.