论文部分内容阅读
水稻品种识别能有效防御假冒伪劣种子,提高水稻种子纯度。利用激光诱导击穿光谱,采用BP神经网络对水稻种子进行了类型识别研究。当波长范围为222.054nm至849.019nm的全谱数据为BP神经网络的输入值时,其识别率为91.2%。将全谱数据进行去噪后,其识别率提高到96.4%。采用分段光谱进行识别时,识别率降低且各段的识别率相差较大,但其识别所用时间大大减小。采用适当的分段光谱组合识别时,能提高其识别率,其识别率可达到92.4%,超过了全谱去噪前的识别率,而识别所用时间远小于全谱识别所用时间。结果表明:利用适当的分段组合激光诱导击穿光谱对水稻品种进行识别时,能在较短的时间内达到满意的识别效果。
Rice variety identification can effectively prevent counterfeit and shoddy seeds and improve rice seed purity. Using laser-induced breakdown spectroscopy, type identification of rice seeds was studied using BP neural network. When the full spectrum data with the wavelength range of 222.054 nm to 849.019 nm is the input value of BP neural network, the recognition rate is 91.2%. When the full spectrum data is denoised, the recognition rate is increased to 96.4%. When using the segmentation spectrum to identify, the recognition rate is reduced and the recognition rate of each segment is quite different, but the recognition time is greatly reduced. When using the appropriate segmentation spectral combination recognition, the recognition rate can be improved, the recognition rate can reach 92.4%, which exceeds the recognition rate before full spectrum de-noising, and the recognition time is far less than the time used to identify the full spectrum. The results show that the satisfactory identification effect can be achieved in a relatively short period of time by identifying the rice varieties with appropriate segment combination laser induced breakdown spectroscopy.