Abstract: Optimal linear combination of RGB color model target recognition is proposed in order to improve the accuracy of target detection. This paper records the target and background on points of sampling of red, green, blue component values, and then designs linear programming objective function and constraints. It makes out of the optimal solution of linear programming problem by using red, green, blue component statistics, and establishes the optimal linear combination of RGB color model. Under four kinds of light shooting conditions such as smooth light, side light, side backlight, backlighting with the experiment of 120 images, it concludes that: this method, by contrast of R-B, 2R-G-B, G-B, has larger gray difference in objectives and background, and shows the lowest false detection rate and missing detection rate. Thus, it can be adopted in target detection which has color difference with the background.

Key words: target detection, linear programming, RGB linear combination, optimal model

摘要： 为提高目标检测的准确性，提出了最优RGB线性组合颜色模型目标识别方法。统计目标及背景代表性点的红色、绿色、蓝色3个分量值，设计线性规划目标函数及约束条件，利用红色、绿色、蓝色3个分量统计值求线性规划问题的最优解，建立最优RGB线性组合颜色模型。顺光、侧光、侧逆光、逆光4种采光拍摄条件下120幅图像的试验结果表明：该方法较R-B、2R-G-B、G-B法的目标与背景灰度差值都大、误检率与漏检率都低，可广泛用于与背景存在颜色差异的目标检测。

关键词: 目标检测, 线性规划, RGB线性组合, 最优模型