摘要： 构建了一种快速鲁棒的图像识别算法,它融合小波分解与遗传算法并行性的优点。对图像和模板进行小波分解,用低频部分代替图像和模板参与识别运算;把对模板进行形变旋转处理的参数编码为染色体,依据染色体表现值对模板进行形变和旋转处理;用模板匹配法判定图像中是否含有形变旋转后的模板,判定函数就是遗传算法的适应度函数。算法终止时的最佳适应度为图像中是否包含模板所示目标的判定依据。实验表明该图像识别方法适应性广、稳定性好、速度快。

关键词: 小波分解, 遗传算法, 低频分量, 模板匹配

Abstract: A new image object recognition method that integrates the merits of wavelet division (WD) and parellel property of genetic algorithm(GA) is proposed. At first, WD is done to the image and the template, substitutes them with corresponding low frequency part to join in recognitionprocess. Then, GA produces the deformation x-direction and y-direction coefficients and the rotation angle, deforms and rotates the templateaccording to the three parameters. Template-matching method is used to decide whether the deformed and rotated template is contained in the image,the decision function is the adaptability function (AF) of GA. At last, the best value of AF is the basis for deciding whether the image contains theobject presented by the template. The practice indicates that the recognition speed is quick and the method is robust.

中图分类号: 

• TP391.41