張鶴 孫瑜



摘? 要: 為了解決以往身份認證中的人臉識別問題,提出了一種基于可見光與紅外攝像頭獲取人像差異的方式來實現活體人臉檢測。首先利用近紅外相機成像特性防止假臉攻擊問題,然后針對近紅外相機獲取的圖像不能利用現有的大量可見光照片進行人臉匹配問題,利用擴大近紅外圖像中檢測到的活體人臉特征框,放到可見光圖像中同坐標位置等比例“裁剪”的方式來確定活體人臉在可見光圖像中的位置,最后對“裁剪”后的圖像進行人臉檢測,獲取面積最大的人臉(即活體人臉)。在自建樣本庫中場景實驗分析,所提出的方法能夠抵御假臉攻擊的同時,又能保證活體人臉檢測的準確度。
關鍵詞: 雙攝像頭;活體人臉檢測;近紅外
中圖分類號: TP391.41 ???文獻標識碼: A??? DOI:10.3969/j.issn.1003-6970.2020.07.010
本文著錄格式:張鶴,孫瑜. 基于雙攝像頭下的活體人臉檢測方法[J]. 軟件,2020,41(07):51-56
Live Face Detection Method Based on Dual Camera
ZHANG He, SUN Yu*
(School of information, Yunnan Normal University, Kunming 650000, China)
【Abstract】: The paper proposed a obtaining human image difference between visible light camera and infrared camera in order to solve the problems of face recognition in the past identity authentication. First, Use of near-infrared camera imaging characteristics to prevent false face attack. Then, for the image acquired by the near-infrared camera, the existing large number of visible light photos cannot be used for face matching, and the feature frame of the living face detected in the near-infrared image is enlarged. Then, we use the Utilizing enlarged live face feature frames detected in near-infrared images. the enlarged face coordinate information was put into the color image to be “clipped” in the same proportion as the coordinate position, and the largest face (i.e. the living face) was obtained according to the "clipped" image. Finally, In the self-built sample database, the method proposed by the scene experiment analysis under different error factors can resist the false face attack and ensure the accuracy of the living face detection.
【Key words】: Dual cameras; Face liveness detection; Near infrared
0? 引言
人臉檢測[1]是指所有人臉在輸入圖像中的位姿、大小的過程。人臉檢測作為一項重要的人臉信息處理技術,成為模式識別與計算機視覺領域一項熱門的研究課題。然而人臉檢測技術容易受到外界因素的干擾,比如光照的充足情況、人體的靜態與動態情況、是否用偽人像等[2-3]。通過研究分析表明[4],其中是否用偽人像是最常見的干擾因素,所謂偽人像就是利用一些圖像或者視頻等手段來偽造成真實的活體人像,因而其實施起來比較方便,所以這也是人臉檢測所要面對的重要攻擊手段。……