為探索步態(tài)識別技術(shù)在實際中的應用，構(gòu)建了基于步態(tài)圖像特征進行身份識別的智能
門禁系統(tǒng)硬件平臺，且開發(fā)了相應的門禁系統(tǒng)軟件。其技術(shù)流程為：首先從攝像機捕獲視頻輸入計算機進行監(jiān)控，發(fā)現(xiàn)后對人體進行檢測與跟蹤，分割人體輪廓并將其規(guī)格化疊加處理獲取步態(tài)特征圖，然后提取包含人體整體模型信息的邊界矩參數(shù)，以此作為識別參量輸入支持向量機（SVM）進行分類識別。為評價步態(tài)識別性能，分別對正確識別率（PCR）、錯誤接受率（FAR）和錯誤拒絕率（FRR）進行了測試，結(jié)果表明該實驗平臺能較好地對人體目標進行自動監(jiān)控跟蹤、提取步態(tài)特征且有效地進行分類識別。
關(guān)鍵詞：步態(tài)識別 門禁系統(tǒng) 特征提取 不變矩
Abstract: In order to explore the practical application of gait recognition, this experimental
platform of access control system based on gait feature was built for human recognition. The
software of experimental platform was compiled. This study deals with the following process. The video frequency was collected from the camera and put into computer. Then the object was detected and also tracked. Body silhouette sequences of gait were extracted and normalized in this study. The sequences were added together and gait character image could be obtained. At the same time moment invariants was used in extracting gait character parameters involving integrated model. The technique of support vector machines (SVM) was presented for gait recognition. In order to test the performance of gait recognition, three items were performed for Probability of Correct Recognition (PCR), False Rejection Rate (FRR) and False Acceptance Rate (FAR).
According to the results, it proves that the experimental platform was beneficial for watching the
object and at the same time tracking the human body. It also revealed great value for feature
extract and gait classification.
Keywords: Gait recognition; Access control system; Feature extract; Invariant Moments