분할 이미지 셀을 이용하는 아다부스트 기반의 보행자 검출 방법 및 장치
    7.
    发明授权
    분할 이미지 셀을 이용하는 아다부스트 기반의 보행자 검출 방법 및 장치 有权
    使用分割图像细胞的基于AdaboOST的对象检测的方法和装置

    公开(公告)号:KR101419837B1

    公开(公告)日:2014-07-21

    申请号:KR1020130051268

    申请日:2013-05-07

    CPC classification number: G06K9/00369 G06K9/6256 G06K9/6261

    Abstract: According to the present invention, a method for detecting a pedestrian based on an Adaboost algorithm comprises the steps of: dividing an input image into unit cells; counting the number of pixels where each of first to third color components is greater than a threshold value, with respect to the respective unit cells, and generating the characteristic values of the first to third color components which are true or false, according to whether a ratio of the counted number of pixels to the total pixel number of unit cells exceeds a predetermined ratio; counting the number of black and white pixels when the respective pixels are divided into the black and white pixels, based on the threshold value, with respect to the respective unit cells, and generating binary characteristic values that are true or false, according to whether the ratio of the counted number of pixels to the total pixel number of unit cells exceeds the predetermined ratio; generating contour characteristic values that are true or false, according to whether contours are extracted, with respect to the respective unit cells; setting a strong classifier as classification weights selected to minimize a combined error of weak classifiers which generate the characteristic values of the first to third color components, the binary characteristic values, and the contour characteristic values from respective sample images, according to an Adaboost algorithm; and classifying whether a pedestrian object is included in the input image by the set strong classifier.

    Abstract translation: 根据本发明,一种基于Adaboost算法检测行人的方法包括以下步骤:将输入图像划分为单位单元; 对相对于各个单位单元计数第一至第三颜色分量中的每一个的像素数量大于阈值的数量,并且根据第一至第三颜色分量是否为真或假产生第一至第三颜色分量的特征值 计数的像素数与单位像素的像素数的比例超过预定比例; 对于各个单位单元,基于阈值,将各像素分割成黑白像素的黑白像素的数量进行计数,根据是否生成真或假的二值特征值 计数的像素数与单位像素的像素数的比例超过预定比例; 根据轮廓是否相对于各个单位单元产生真或假的轮廓特征值; 将强分类器设置为选择的分类权重,以根据Adaboost算法最小化生成来自各样本图像的第一至第三颜色分量,二进制特征值和轮廓特征值的特征值的弱分类器的组合误差; 并通过设置的强分类器对步行对象是否包含在输入图像中进行分类。

    암모니아 처리를 통한 은 나노와이어 필름의 제조방법 및 그에 의해 제조된 은 나노와이어 필름
    8.
    发明公开
    암모니아 처리를 통한 은 나노와이어 필름의 제조방법 및 그에 의해 제조된 은 나노와이어 필름 有权
    用于制造纳米线膜的方法用氨处理,从而由所述纳米线膜产生

    公开(公告)号:KR1020170052781A

    公开(公告)日:2017-05-15

    申请号:KR1020150154247

    申请日:2015-11-04

    Abstract: 본발명은암모니아처리를통한은 나노와이어필름의제조방법및 그에의해제조된은 나노와이어필름에관한것으로, 기판위에코팅된은 나노와이어에암모니아를단시간처리하여투과도저해없이전도도를향상시킴으로써, 은나노와이어필름의전도도향상을위한별도공정인고온열처리를대체할수 있는장점이있다.

    Abstract translation: 本发明通过改善没有传输抑制是通过氨用氨处理快速处理在导电性已经涉及一种方法,用于制造纳米线膜和由此产生的纳米线膜,在基材上的涂层是纳米线,银线 存在这样的优点,可以取代的高温热处理的单独的步骤用于提高膜的导电性。

Patent Agency Ranking