基于红外热成像树上板栗机器识别研究

Study on Machine Recognition of Chinese Chestnut on Trees Based on Infrared Thermography

针对普通视觉系统无法分类识别的特点，将红外热成像技术引入到树上板栗的机器识别检测当中．建立了果同树上板栗红外热成像系统。记录了白天树上板栗果实与冠层温度变化情况，发现下午2：00～4：00之间．二者温度差异最为显著，最大温差可达3℃。研究了正常板栗果、坏死果以及空心果的红外辐射差异，发现坏死果表面温度最高．一常果次之，空心果温度最低。最后对获取的热图像进行温度校正、阈值分割、特征提取、目标标识。结果表明．树上板栗的有效识别率在85％1．2上。

As common machine vision systems are not capable of effectively distinguishing fruits, infrared thermography techniques were introduced to detect and identify Chinese chestnut on trees. Testing on a special infrared thermography system in orchard, the fruit and leaf canopy temperature profile during the daytime were recorded. It was found fruits have higher temperatures than the canopy from 2： 00 PM to 4：00 PM, and the maximum temperature difference reaches to 3~C.The infrared radiation properties of sound fruits, dead fruits and hollow fruits were also studied, and found that the temperature on dead fruit is highest, sound fruit comes second, and hol- low fruit has lowest temperature. Finally, image processing procedures such as temperature correction, thresholding segmentation, feature extraction and object identification were implemented to recorded thermal images. Test results showed that the effective recognition rate is up to 85%.