本論文在探討如何很有效率地設計一個滿足乃曼-皮爾森準則的平行式分散式偵測系統。其中乃曼-皮爾森準則限制假警報機率在一個預先規定的值，並要求使偵測機率達到最大值；而分散式偵測系統使用了多個感應器觀測同樣一個現象，各個感應器在對其觀測資料做過處理之後，再將其決策送往融合中心作最後的判別。 設計分散式偵測系統所需要的時間將隨著感應器的數目成指數性地增加，因此，尋找一個有效率的演算法是非常重要的。在本論文中將討論如何以牛頓法調整感應器的門檻，配合以割線法調整融合規則以及良好的初始值，能夠快速地設計符合乃曼-皮爾森準則的分散式偵測系統。此演算法和傳統上被採用的Gauss-Seidel法相比，的確具有較好的成績。 A study of design of a parallel distributed detection system under Neyman-Pearson criterion, which means one seeks to attain the maximum detection probability with a constrained false alarm probability, is presented in this thesis. The distributed detection system uses multiple sensors to observe the same phenomenon, process observation data, and pass their decisions to the fusion center for the final decision. The computation time for the design of a distributed detection system increases exponentially with the number of sensors, therefore it’s very important to find out an efficient algorithm. A discussion about how to adjust the threshold of sensors by Newton method with regulation of fusion rule and good initial values will be included in this thesis. We design a distributed detection system under Neyman-Pearson criterion in efficiency. This algorithm indeed has a better performance compared with the Gauss-Seidel algorithm used in tradition.
