Digital Image Moving Object Extraction And Its Application

The rolled-strips,as one form of iron and steel industry main product, has become the raw material of automobile, machine manufacture, chemical industry, aerospace. Each large steel factory is improving the rolled strips quality. Along with the information technology rapid development, computer vision has made considerable progress in the cold-rolled-strip surface defect inspection system. The main goal in this article is to defect the segment at cold-rolled-strip surface with computer vision. Because the field background is complex and uncertain, to remove the disturbance factor with general method is not very effective. To subtract the rolled-strips surface defect with drawing auto-adapted form non-parameter Gauss background model, we obtain the flaw target array for the first time. The utilization ranked-order mathematics morphological filters suppress the spot, and successfully withdraw the flaw from the target image. This article has mainly studied the following content:1. Analyzed and compared several kinds of backgrounds subtraction algorithm, obtained algorithms under complex environment. And proved that Gauss background model subtraction algorithm is a quite suitable solution for the question of getting flaw withdrew.2. Elaborated the non-parameter Gauss background model. Then used the non-parameter Gauss background model to take rolled strips surface blemish is proved to be effective The gradient difference non-parameter Gauss background mode subtraction algorithm and the frame difference non-parameter Gauss background model subtraction algorithm utilizes were the first time to apply the cold-strip steel surface defect withdraws.3. In regarding to fetch in the non-parameter Gauss background model parameter, used many images constitution matrix characteristic value to take the non-parameter Gauss background model the unknown parameter. Has solved this algorithm consumption memory in a big way, parameter renewal difficulty and so on, the experiment indicated this method was effective.4. Has analyzed and designed ranked-order mathematics morphological filters, and...