论文部分内容阅读
为加强空中回转体目标识别的智能化程度,提高目标识别率,对目标图像的不变矩特征与BP神经网络相结合的空中回转体目标识别方法进行了研究。以给出权值的方式引入Relief算法对目标图像的7个Hu氏不变矩、3个仿射不变矩的识别性能进行科学的评估。选出权值较大的特征量作为BP神经网络的输入特征来训练网络,识别样本。模拟实验表明:引入Relief算法对空中回转体目标不变矩选择的目标识别方案是有效的,应用Relief算法选择出的特征项作为神经网络的输入特征不但减少了特征量提取的采集次数,降低了算法的计算量,而且,可使网络更易于收敛,且提高了目标物的识别率。
In order to enhance the intelligence of target recognition in air swivels and improve the target recognition rate, a method of target recognition of air swivels combined with invariant moment features of target images and BP neural network was studied. The Relief algorithm is introduced to evaluate the performance of seven Hu’s moment invariants and three affine invariant moments in the target image. Select the features with larger weights as the input features of BP neural network to train the network and identify the samples. The simulation results show that the proposed Relief algorithm is effective for target recognition of target invariant moments in aerial gyroscopes. Using the feature selected by Relief algorithm as the input feature of neural network not only reduces the number of acquisitions, but also reduces The amount of calculation of the algorithm, moreover, the network can be more easily converged and the recognition rate of the target is improved.