| 引用本文: | 高庆一,逯鹏,袁道任.视觉感知的目标识别算法[J].哈尔滨工业大学学报,2013,45(1):101.DOI:10.11918/j.issn.0367-6234.2013.01.019 |
| GAO Qingyi,LU Peng,YUAN Daoren.Target recognition algorithm based on visual perception[J].Journal of Harbin Institute of Technology,2013,45(1):101.DOI:10.11918/j.issn.0367-6234.2013.01.019 |
|
| 摘要: |
| 针对图像的目标识别问题,采用视觉感知的方法,模拟感受野的分层信息处理机制,并引入神经元间的侧抑制机制,对神经元响应进行筛选,通过检测视觉基本特征的方式识别图像中的目标．算法首先在简单细胞的感受野中对图像进行预处理;其次,在复杂细胞的感受野中,将简单细胞的感受野刺激进一步拓扑特征提取,得到感受野刺激响应;最后,通过侧抑制机制对响应神经元筛选,找出对刺激响应较强烈的神经元,将其输出作为目标识别的参数标准．实验结果表明,基于视觉感知的算法可以用少量样本解决大量图像中的目标识别问题,识别率高于边缘检测和图像分割方法,算法的目标识别率达到95.56％． |
| 关键词: 感受野 视觉感知 侧抑制 目标识别 |
| DOI：10.11918/j.issn.0367-6234.2013.01.019 |
| 分类号:TP2 |
| 基金项目:国家自然科学基金资助项目(0,2); 郑州市科技攻关资助项目(112PPTGY219-8). |
|
| Target recognition algorithm based on visual perception |
|
GAO Qingyi1, LU Peng2, YUAN Daoren2
|
|
(1.School of Computer Science and Engineering, Beihang University, 100191 Beijing, China; 2.School of Electrical Engineering, Zhengzhou University, 450001 Zhengzhou, China);1.School of Computer Science and Engineering, Beihang University, 100191 Beijing, China ;2.School of Electrical Engineering, Zhengzhou University, 450001 Zhengzhou, China
|
| Abstract: |
| Aiming at the problem of target recognition in image processing,a novel algorithm based on visual perception with better information processing mechanism was proposed for target recognition, in which the hierarchical information processing mechanism of receptive field were simulated and the lateral inhibition mechanism which was used to filter response of neurons was introduced. Firstly, in receptive field of simple cells, the images are initially processed. Secondly, in receptive field of complex cells, stimulation of receptive field from former layer is further feature extracted, and then a new receptive field is obtained. And finally, the neurons which are strongly responded to the stimulation are found out by lateral inhibition mechanism, and their correspondent content can be output as the standard of aerial target recognition, the recognition rate of algorithm is 95.56％. Experimental results show that this novel algorithm has high efficiency, and can achieve a great number of target recognitions with few samples, and the recognition rate is higher than that of edge detection and image segmentation method. |
| Key words: receptive field visual perception lateral inhibition target recognition |
