BP neural network target detection method based on histogram statistics

The invention provides a BP neural network target detection method based on histogram statistics, and the method comprises the steps: converting two-dimensional target slice information into one-dimensional histogram information, building a BP neural network model, and enabling the image histogram information to serve as the BP neural network input. The method has the advantages of being low in operation complexity, high in operation efficiency, simple in network model and capable of achieving rapid target detection and classification. Under the constraint of limited on-orbit resources of satellites, target detection and recognition work is completed on orbit quickly with low cost, high efficiency and high accuracy. 本发明提出一种基于直方图统计的BP神经网络目标检测方法，该算法将二维目标切片信息转换为一维直方图信息，通过建立BP神经网络模型，将图像直方图信息作为BP神经网络输入，算法具有运算复杂度低、运行效率高、网络模型简单、可实现快速目标检测分类的优点。在卫星有限的在轨资源约束下，低成本、高效率、高准确率、快速地在轨完成目标检测识别工作。