On-line Nave Surface Flaw Inspection System

Nave is an important part of a train and the manufactories have paid more and more attention to the inspection of nave flaw. The inspection work can be mainly divided into inner inspection and surface flaw inspection. The former is carried out by the way of ultrasonic, etc and has got preferable results. But, surface inspection is a still a challenging problem. Traditional visual inspector tends to miss flaws and is with low efficiency so that it couldn't be adopted in an on-line inspection. Consequently, developing automatic surface flaw inspection system has become a common demand of the nave manufactories.Surface flaw inspection techniques are investigated in this thesis. An on-line nave surface flaw inspection system is designed and developed. Hardware and software of the system and algorithms for flaw inspection are described. The details of the work, which has been done, are as follows:1. An on-line nave surface flaw inspection system is designed and developed. Hardware mainly includes capturing system, computer system and so on while software mainly includes image capturing module, object (or flaw) inspection module, data exchange module and output module. The system has the following features:(1) Multiple CCD matrix cameras are equipped to capture surface images of a fluorescence and magnetic particle processed nave, which leads to high resolution, quality and precision of inspection;(2) Ultraviolet lights are adopted as illumination methods to improve the contrast of object and background so that flaws can be easily detected;(3) Parallel process, such as image capturing and object inspection, are done by the lower computers. Furthermore, each lower computer is with responsibility for three sets of capturing equipment (one camera and one image capture card);(4) The two lower computers, which are equipped with cameras and capturing cards, are linked with the upper computer through an RGB video capturing card. In this way, it leads to centralized control and management;(5) The edge and region information are used together for nave image segmentation, which leads to high accuracy of inspection;(6) All the algorithms are implemented by software to provide flexibility of application and upgrading.2. Algorithms for flaw inspection are developed. Experiments with samples from production lines show that object can be extracted from the background, which is filled with different types of noises.3. Multithreading technique is applied in the process of object inspection to improve the speed of data processing, which satisfies real-time processing.4. The upper and the two lower computers are linked as a LAN (local area network), which leads to data exchange and resource sharing.In conclusion, the proposed algorithms can reduce the effect of background asymmetric illumination, blade lines, water spots, oil stains, etc, while getting satisfying results of flaw inspection. The system, which meets the need of on-line flaw inspection, has the following features: computationally efficient, low flaw missing rate and false alarm rate, real time and so on.