Rapid object detection using a boosted cascade of simple features：（快速目标检测使用了一连串的简单功能）.pdfVIP

ACCEPTED CON FEREN CE ON COMPU TER VI SION AN D PATTERN RECOGN ITION 200 1 Rapid Object Detection using a Boosted Cascade of Simple Features Paul Viola Michael Jones viola@ mjones@ Mitsubishi Electric Research Labs Compaq CRL 201 Broadway, 8th FL One Cambridge Center Cambridge, MA 02139 Cambridge, MA 02142 Abstract tected at 15 frames per second on a conventional 700 MHz Intel Pentium III. In other face detection systems, auxiliary This pap er describes a machine learning approach for vi- information, such as image differences in video sequences, sual object detection which is capable of pr ocessing images or pixel color in color images, have been used to achieve extremely rapidly and achieving high detection rates. This high frame rates. Our system achieves high frame rates work is distinguished by three key contributions. Thefir st working only with the information present in a single grey is the introduction of a new image representation called the scale image. These alternative sources of information can “Integral Image” which allows thefeatur es used by our de- also be integrated with our system to achieve even higher tector to be computed very quickly. The second is a learning frame rates. algorithm, based on AdaBoost, which selects a small num- There are three main contributions of our object detec- ber of