论文部分内容阅读
提出一种基于概率潜在语义分析(Probabilistic Latent Semantic Analysis,PLSA)的检测算法,首先通过PLSA将目标表述为潜在成分的概率组合,然后利用统计模式识别方法对获取的潜在成分概率进行判别,从而完成最终的检测。其中,生成的潜在成分反映了目标与特征之间相互出现的频率关系,并以潜在成分在目标中概率差异的形式对上述不对应现象给出了直观描述。实验结果表明,所提出的算法对多种复杂情况下的光学图像舰船检测具有很好的准确性和鲁棒性。
A detection algorithm based on Probabilistic Latent Semantic Analysis (PLSA) is proposed. Firstly, the target is expressed as the probability combination of potential components by PLSA, and then the statistical pattern recognition method is used to determine the probability of potential components obtained, so as to complete The final test. Among them, the potential components generated reflect the frequency relationship of the mutual appearance of the target and the feature, and the intuitive description of the above-mentioned non-correspondence phenomenon takes the form of the probability difference of potential components in the target. The experimental results show that the proposed algorithm has good accuracy and robustness to optical image ship detection under many complicated conditions.