摘要
以一起典型案件为例,探寻低质模糊视频人像的综合性检验方法。将多个检材与样本视频链接为单一视频文件以便回放、搜寻和分析人像特征,根据人像特征的反映情况,确定以人像头部、五官形态、衣着、体型、体态五个方面作为检验内容与方向;从连续截取的视频原始画面中裁切出各类人像特征图片,按姿态、视角、方向相似的原则从中筛选用于比较分析的图片;分别对人像头部与五官形态特征、衣着款式特征、人像在行进过程中的头部姿态、摆臂姿态、步幅姿态、人像起落脚的步态特征以及人像的习惯性动作进行比较检验,并对人像身高进行测量。在上述检验过程中,采用制作慢速播放的分屏比对视频的方法对反映在时间连续性上的动态特征进行呈现、比对和记录。最后,综合评断了各类人像特征的一致性。各类人像特征分别在各自维度上具有高度一致性,他们相互联系、相互补充、相互印证,优化并提升了各自的证明价值和证明力,构成了本案人像同一认定的客观依据。搜集充分的、可比性强的各种影像资料与相关物证,采取多种方法,从多个维度进行比对检验,是模糊视频人像检验与同一认定的可行且有效途径。
Objective To explore a comprehensive method for examining and identifying human image in poor-quality videos based on a typical case. Methods In advance of the examination, a variety of available videos and related evidence were collected and submitted for assessment. The videos were organized and linked together as a single file to facilitate playback, visual inspections and analysis. According to the features depicted in videos, the head and facial morphology, clothing, somatotype and posture were determined as the examination contents and directions. Then the pictures reflecting multiclass human image features were extracted and cropped from the previously captured continuous video-images, and those used for comparison were chosen under similar pose, angle and orientation. Afterwards, a series of “side-by-side” comparisons and evaluations were conducted on the following aspects： the head and facial morphology, clothing features, body postures at key nodes in a gait cycle, foot landing direction, foot motion changes and habitual behavior. A photogrammetric analysis was also conducted to determine the height of the person depicted in the submitted videos. In the above-mentioned process, a slow motion and split screen video were made and used for presenting, comparing and recording the observed dynamic human image characteristics reflected with time lapsing. Finally, a comprehensive assessment was made according to the correspondence/discordance of the human image characteristics. Results All kinds of human image characteristics were consistent in their respective dimensions and thus they were mutually interrelated, complemented and supported. Their probative value and proving ability were optimized and improved, leading to certain identification. Conclusions It is a feasible and effective way for examination and identification of human image in poor-quality video by collecting sufficient image data and related evidence with comparability and conducting photographic comparisons from multiple dimensions with
出处
《刑事技术》
2015年第4期340-344,共5页
Forensic Science and Technology
基金
四川省教育厅项目(No.13TD0016
No.13ZA0116)
关键词
模糊视频
人像鉴定
人像同一认定
poor-quality video
human image examination
human image identification
