白丝美女被狂躁免费视频网站,500av导航大全精品,yw.193.cnc爆乳尤物未满,97se亚洲综合色区,аⅴ天堂中文在线网官网

Data processing apparatus, data processing method, and non-transitory storage medium

專利號
US10867162B2
公開日期
2020-12-15
申請人
NEC Corporation（JP Tokyo）
發(fā)明人
Jianquan Liu; Shoji Nishimura; Takuya Araki; Yasufumi Hirakawa
IPC分類
G06K9/00; G06F16/783; G06F16/00
技術(shù)領(lǐng)域
person,data,extraction,in,analyzed,appearance,image,unit,moving,be
地域： Tokyo

摘要

A data processing apparatus (1) of the present invention includes a unit that retrieves a predetermined subject from moving image data. The data processing apparatus includes a person extraction unit (10) that analyzes moving image data to be analyzed and extracts a person whose appearance frequency in the moving image data to be analyzed satisfies a predetermined condition among persons detected in the moving image data to be analyzed, and an output unit (20) that outputs information regarding the extracted person.

說明書

For example, all possible pairs of all persons (all detection IDs) detected in previously processed frames and the detection ID of the person detected from the frame to be processed may be created, and similarity determination may be performed for each pair. In this case, however, the number of pairs would become enormous. As a result, the processing speed may be reduced.

Therefore, for example, the person extraction unit 10 may index a person detected from each frame as shown in FIG. 7, and perform the above-described determination using the index. By using the index, it is possible to increase the processing speed. The details of the index and the generation method are disclosed in Patent Documents 2 and 3. Hereinafter, the structure of the index in FIG. 7 and its usage will be briefly described.

The indexes shown in FIG. 7 hierarchize persons detected from a plurality of frames, specifically, FIG. 7 hierarchizes detection IDs.

In the third layer, nodes corresponding to each of all the detection IDs obtained from all the frames processed up to that point are arranged. Then, the plurality of nodes arranged in the third layer are grouped such that those having a similarity (similarity between the feature values shown in FIG. 5) equal to or higher than a predetermined level are grouped together.

權(quán)利要求

1
微信群二維碼
意見反饋