应用于禽蛋壳体裂纹声检测的嵌入式神经网络方法

Application of an Embedded Neural Network Method for Acoustic Crack Detection of Poultry Eggshells

针对禽蛋裂纹检测系统复杂识别算法受时间和空间的约束很难在嵌入式平台上实现的缺陷,提出一种基于ARM处理器(STM32F407)的禽蛋裂纹检测系统的研究方法。该方法的主要思想为在嵌入式平台ARM处理器的基础上运用BP神经网络结构实现禽蛋裂纹检测,并对原有算法进行优化,其中嵌入式神经网络结构和优化算法具有减少硬件资源占用和降低实施复杂程度的优点。实验结果表明,基于ARM平台的BP神经网络的方法测试完好蛋和裂纹蛋的准确率较原有PC端上的测试准确率一致,但该方法可以更好地适应工业化要求。

To address the defect that the complex identification algorithm of egg crack detection system is difficult to be realized on the embedded platform due to time and space constraints,a research method of egg crack detection system based on ARM processor(STM32F407)is pro⁃posed in this paper.The main idea of this method is to use BP neural network structure to realize the poultry egg crack detection on the ba⁃sis of ARM processor on embedded platform,and optimize the original algorithm,in which the embedded neural network structure and opti⁃mization algorithm have the advantages of reducing hardware resource occupation and implementation complexity.The experimental results show that the accuracy of the BP neural network based on ARM platform to test intact eggs and cracked eggs is consistent with the original accuracy on the PC,but the method can better adapt to industrial requirements.