Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (16): 234-238.

Previous Articles     Next Articles

Design of In-field presorting system for apples based on ARM

XU Libing, ZHU Qibing, HUANG Min   

  1. Key Laboratory of Advanced Process Control for Light Industry, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2015-08-15 Published:2015-08-14

基于ARM的苹果采后田间分级检测系统设计

许立兵,朱启兵,黄  敏   

  1. 江南大学 轻工过程先进控制教育部重点实验室,江苏 无锡 214122

Abstract: In-field presorting for apples can reduce production costs such as storage, packaging, processing etc, and it plays a very important role in improving the economic income of growers and increasing the economic efficiency of enterprise. To overcome the traditional classification detection system disadvantages such as the huge size, difficulty of real-time application in field, apple In-filed presorting and detecting system is designed and implemented based on ARM11 processor which is named S3C6410 and runs pruned Linux software platform. Image acquisition is based on Linux V4L2 framework, CMOS camera acquires apple’s image real-time;ARM is utilized to call OpenCV machine vision processing algorithms which is based on cascade Adaboost target detection algorithm and Haar-like characteristics, detecting apple’s defects and size in time;sending the results to the actuator completing apple's separation. The system control interface is compiled with QT and multi-threading technology, control buttons responding immediately. Preliminary experiments show that the average detection time for an apple is 300 ms and average accuracy is 93%. Compared with the traditional apple detection system, this system has the advantages of high detection speed, low cost, small size, which is suitable for In-field presorting for apples.

Key words: apple detection, QT, ARM, OpenCV, machine vision

摘要: 苹果采摘后的及时分类,可以降低储藏、包装、加工等生产成本,对于增加果农的经济收入,提高企业经济效益具有重要的作用。项目研究并开发了一套基于ARM11+Linux架构,以S3C6410为核心处理器、运行精简的Linux内核的苹果采后田间预分级检测系统,克服了传统分级检测系统体积庞大,难以田间实时应用等缺点。该系统利用CMOS图像传感器,基于Linux下的V4L2编程框架,实现苹果图像的实时采集;采用基于Haar-like特性的级联Adaboost目标检测算法,调用OpenCV机器视觉库,完成检测图像中的苹果缺陷和大小识别;并由执行机构完成不同等级苹果的分离操作。系统控制界面采用QT应用程序开发框架和多线程技术,保证了控制按键的快速响应。实验结果表明,一个苹果的平均检测时间为300 ms,对于各级苹果分类的平均精度为93%,与传统苹果检测系统相比,该系统检测速度快,成本低,体积小,适合苹果的田间预分级检验。

关键词: 苹果检测, QT, ARM, OpenCV, 机器视觉