大尺度混合畸变的目标识别技术

Recognition Technology of Large-Scale Hybrid-Distortion Target

为解决目标识别技术中混合畸变目标的识别难题，提出了一种面积极坐标变换算法．该算法利用相似图形尺寸与面积的满射函数关系和极坐标变换的旋转不变特性，对目标与模扳同时进行面积变换和极坐标变换．将面积极坐标变换算法与基于形态学膨胀的边缘加粗处理算法相结合应用于光电混合联合变换相关器，实现了大尺度混合畸变目标的识别．以空中飞机为例，光学相关实验表明，该方法可显著提高混合畸变目标的相关峰强度，旋转畸变容差为0°～100°，比例畸变容差为100％～200%，具有较大尺度的混合畸变容差，扩大了识别范围．其混合畸变识别能力优于极坐标对数变换算法与综合鉴别函数法其衍生方法，其“时间复杂度”远小于综合鉴别函数及其衍生方法．

In this paper,area-polar algorithm is firstly proposed to solve the problem of hybrid-distortion target recognition technology.This method makes area transform and polar coordinate transform for both target and template,based on the surjective function relationship between the scale and the area of similar figures and the invariance of rotation of polar coordinate.Applied on hybrid optoelectronic joint transform correlator,a combined method can achieve large-scale hybrid-distortion target recognition,which includes area-polar algorithm and edge extracting process based on morphological dilation.Presented with examples of planes,The optical correlation experiments show that this method can enhance the brightness of the correlation peaks for hybrid-distortion target,which possesses large hybrid-distortion tolerance(with 0°~100°rotation distortion tolerance and 100%~200%scale distortion tolerance),and expands the scope of target recognition.The hybrid-distortion target recognition capability by this method is better than by Log-polar transform algorithm,and also better than by synthetic discriminant function and by derivative methods,besides its time complexity is far less than synthetic discriminant function′s and derivative methods′.