基于Softmax的采摘机器人目标识别技术研究

Research on Target Recognition Technology of Picking Robot Based on Softmax

首先,介绍了目标水果的形态、采摘机器人识别技术与试验平台;然后,设计了双目视觉系统,并基于Softmax分类模型设计了苹果的目标识别算法,实现了一套基于Softmax的采摘机器人目标识别模型。试验结果表明:采摘机器人对无遮挡的苹果目标识别率为94%,对部分遮挡的苹果目标正确识别率为88.7%,稳定性较好,精度较高。

It first introduces the shape of the target fruit,picking robot recognition technology and experimental platform,then designs the binocular vision system,and designs the apple target recognition algorithm based on Softmax classification model,and realizes a set of picking robot target recognition model based on Softmax.The experimental results show that the recognition rate of the picking robot is 94% for the uncovered Apple target and 88.7% for the partially occluded Apple target,with good stability and high accuracy.