电子监控探测人物身高的分析

依据重庆市公安局刑警大队提供的监控数据,分析监控器摄象时影响身高的因素,利用多元回归的原理建立了三个预测身高的回归模型,并对其作出了优劣评价,最终确定出犯罪嫌疑人实际身高与其在照片上的像高、电子监控器距地面的高度、人与监控器的垂直和偏离距离四个因素的最优回归模型.该最优模型的可决系数R2=0.9931,预测身高与实际身高有97.89%的误差点在±0.02m范围内,73.09%在±0.01m范围内.并对最优模型中的预测身高与实际身高进行配对样本的t检验,两者间无显著性差异.

An Analysis of Human Stature Detection by Electronic Vidicon

According to the monitoring data provided by the division of criminal police force in Public Security Bureau of Chongqing municipality,analyses were made on factors that affected the prediction of statures which captured by Vidicon.Three regression models that could predict human statures were built with the theory of multiple regression.After an assessment of the three models,an optimal one was decided,based on factors including suspects′ actual body heights,suspects′ body length on photos,the installing heights of Vidicons,and the perpendicular and deviated distance between the suspects and Vidicons The R Square of the optimal model was 0.9931.In the error analysis made between the predicted and the actual stature,97.89% error points were obtained in the range of ±0.02m while 73.09% in the range of ±0.01m Finally,in paired t-test, both the predicted and the actual stature in the optimal model showed no significant difference.