Study And Practice Of Image Thinning Algorithm Based On PCNN

The third artificial neural network based on the visual model of mammals—Pulse Coupled Neural Network(PCNN), has been applied to image processing widely. This network is very close to the biological neural network model of human brain, and is an important tool for information processing. Image thinning is that the ridges of a certain width are reduced into lines of single pixel width to extract the "skeleton", and it is one of the important features to describe the nature of image geometry and topology. Pulse coupling and autowaves, PCNN inherent features, make it suitable for image thinning.A thinning algorithm based on PCNN, proposed in this paper, is applied to CD-ROM object lens wires quality inspection system and Automated Fingerprint Identification System. A common core hardware platform is designed based on DSP technology, and different peripheral modules are distributed to set two systems up according to different requirements. The former system can accomplish online automatic detection of wires, and distinguish defects; the latter can accomplish fingerprint collection, preprocessing, thinning and matching, and would be used in a variety of applications such as entrance guard. Besides, embedded fingerprint management system designed in this paper could make the fingerprint identification system have more features and easier operation. The jobs are listed as follows:1. A thinning algorithm is proposed in this paper, based on analysis of PCNN model, parameters and mechanism. The feature of PCNN autowaves is studied and applied to image thinning.2. An online automated CD-ROM object lens wires quality inspection algorithm is proposed. The wire image segmentation, image thinning, setting up a model used for wire quality inspection and judging in accordance with this model are described in chapter 5.3. A thinning algorithm based on PCNN model is used for fingerprint image thinning. On the basis of the research and analysis of conventional fingerprint thinning algorithms, PCNN algorithm is adopted to retain the basic shape and connectivity of fingerprint, to get rid of burrs, and to remove false minutiaes. The experiment shows that, compared with those conventional algorithms, this algorithm based on PCNN model overcomes the shortcomings, and has a better result.4. Designing two hardware systems based on DSP. The common core hardware platform is composed of an algorithm processing module with a core of TMS320VC5509A(5509A for short), a data transfer interface, an HCI module, a power management module and an external memory module. Moreover, a camera module used in CD-ROM object lens wires quality inspection system and a fingerprint sensor module in Automated Fingerprint Identification System are also included.5. An embedded fingerprint management system is also designed. This system is used to manage the fingerprint templates and information of users. In this system, the functions of registration, login, inquiry, modification, deletion, download and clear-out are involved. With this system, users can manage the fingerprint template library easily. In addition, the fingerprint data can be transferred to PC through USB interface for visual observation and debugging.The experiment proves that, the two systems work stably, and both can meet the actual demands well.