Abstract: It was difficult to find the correct object in the circumstance of noise when detecting the object in the pantograph image . In order to overcome this difficulty, a rectangular template was employed to search the local object edge. The algorithm description model was constructed based on Su-field analysis method in TRIZ theory. The algorithm was analyzed by applying conflict solution method in TRIZ theory, the optimal template dimension was obtained based on the principle of No.1 segmentation, and the big step search strategy was used based on the principle of No.35 parameter changes, and the strategy of search the edge from the object inside to the outside was used based on the principle of No.13 the other way round. Simulation example showed that the algorithm possessed the ability of resisting the noise in pantograph image without increasing computation amount obviously. Finally, the actual pantograph image in scene was used to verify the detection algorithm; the results revealed its rationality and feasibility.

Key words: TRIZ theory, pantograph, substance field model

摘要： 为减少对受电弓图像进行目标检测时噪声的存在对算法发现目标的影响，根据发明问题解决理论的物质-场模型，构建出一种以矩形区域为检测模板的局部目标边缘的检测算法，并应用发明问题解决理论的冲突解决方法对算法进行改进分析，基于No.1分割原理构建了变尺寸区域模板搜索边缘的方法，基于No.35参数变化原理构建了大步长搜索边缘的方法，基于No.13反向原理构建了由目标内向外搜索边缘的方法。模拟算例表明，该算法在不显著提高计算量的条件下具有抵抗噪声的能力。以现场实际图像对算法进行验证，结果表明算法是合理可行的。

关键词: 发明问题解决理论, 受电弓, 物质-场模型, 机器视觉, 图像