Research On Object Shape Recognition Based On Fourier Descriptors

With the rapid development of multimedia technology, computer communication technology and network, people widely used electronic devices of performance in life,also come into contact with more need to deal with digital image information. Digital image recognition technology has been increasingly widely used in all aspects in our life and work safety,etc. So,study image target recognition is of great significance, the research has very broad application prospects. Shape recognition of the research is the key of image target recognition, good shape feature matching of poor directly determines the effect of target recognition. Fourier descriptor is one of the main methods for shape recognition, and this topic focuses on the Fourier descriptor in the object shape recognition, and improve the optimization. The main research content is as follows:1.mainly studied the application of Fourier descriptor in the object shape recognition. Based on the curve continuous Fourier transform of polygonal approximation method to calculate the Fourier descriptor, and through the shape of the main direction to eliminate the influence of the initial point of phase boundary method, defines the new rotation, translation and scaling invariance of normalized Fourier descriptor. And using discrete Fourier transform and modulus normalized,compared to traditional Fourier descriptor. The experiments show that this new normalized Fourier descriptor than traditional Fourier descriptor can more efficiently and accurately identify the shape of the object.2.Traditional Fourier description method, the boundary of the target contour extract Fourier transform Fourier descriptor with the same characteristics to identify the target. The algorithm is simple, small computational complexity, suitable for real-time occasion, but susceptible to noise. The lack of stability. In this paper,combined with the feature of moment and the shape of the Fourier descriptor identification method. This method takes the centroid as the center is divided into several fan-shaped area, the object are calculated torque characteristic values for different segments of the fan said the object shape torque characteristics of the sequence, and is obtained by discrete Fourier transform with translation, rotation and scale invariant normalized moment features of Fourier descriptor.