基于自适应遗传算法和BP网络的物重监测模型

采用自适应遗传算法与误差反向传播算法(BP)相结合,建立一个通过图像监测物体重量的模型。先对图像进行特征提取,然后用遗传算法进行全局训练,再用BP算法进行精确训练,使网络收敛速度加快并避免局部极小。作为实例,利用圆柱体、锥形体、梯形体等物体图像相关资料建立了数据库,将图像的特征因素作为样本对网络进行训练,并用训练好的网络预测未知物体重量。由实例表明,该方法在预测物体重量中是可行的,误差较小,为物体重量监测提供了一种新思路和新方法,可用于大型生产线上的物体重量在线监测和质量控制。

Object-weight forecasting model based on adaptive genetic algorithm and BP neural network

The adaptive genetic algorithm (GA) and back-propagation (BP) neural network are applied to forecast object weight by image database. A weight forecasting model is established. First, some characters are obtained from the object pictures. Second, adaptive genetic algorithm is used for global training. Finally, BP algorithm is used for accurate training. With these steps, the network convergence speed is increased and local minimum is avoided. As an example, the database for the purpose of training the BP network is established based on the picture data of cylindrical, conical and trapezoic objects. The trained BP network is applied in weight forecasting. Experimental results demonstrate the forecasting model is effective and the error rate is low. It provides a new method and new thinking for forecasting object weight and is suitable for real-time monitoring and mass control of the large-scale production line.