Scholarlink

Purpose <p>Autonomous combine harvesters require accurate crop-row detection for real-time tracking. However, the high cost and reliance on predetermined paths limit the use of real-time kinematic (RTK) global navigation satellite system (GNSS) units, particularly in small-scale farming. Laser range finders (LRFs) with pan-tilt units offer a cost-effective alternative by generating precise three-dimensional profiles using the time-of-flight principle. Therefore, this study aimed to develop an optimized driving baseline modeling algorithm utilizing LRF technology to enhance autonomous navigation in combine harvesters.</p> Methods <p>Various automatic edge detection methods were compared based on accuracy and processing time. A crawler-type, motor-driven mobile platform equipped with a real-time controller and artificial landmarks was developed to simulate the driving baseline algorithm. Additionally, a field-test platform using a 42&#xa0;kW combine harvester was established to validate the algorithm under agricultural conditions. Noise in the raw data was filtered using a modified random sample consensus (RANSAC) method. Crop edge detection was performed using a pan-tilt unit, followed by a second RANSAC filtering step to predict the driving baseline model and calculate lateral and heading deviations in real time.</p> Results <p>Experimental results demonstrated that the proposed algorithm was robust against disturbances and achieved higher accuracy than Otsu’s method, valley-emphasis method, k-means clustering, and the original triangle method under real-time conditions. The average lateral and heading deviations from straight-running tests at speeds of 2, 3, and 4&#xa0;km/h were 0.139&#xa0;m and 7.36°, respectively.</p> Conclusions <p>Given the standard divider interval of 30&#xa0;cm in combines, the achieved accuracy confirms that this LRF-based crop edge detection algorithm is well-suited for integration into autonomous combine harvester systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development of a Real-Time Crop Edge Detection Algorithm Based on Laser Scanner for Autonomous Combine Harvesters

  • 期刊论文

  • 发表日期：

    2025年8月1日

查看全文
  • Chan-Woo Jeon,
  • Yong-Hyun Kim,
  • Hak-Jin Kim,
  • Xiongzhe Han,
  • Junghun Kim,
  • HeeChang Moon

摘要

Purpose

Autonomous combine harvesters require accurate crop-row detection for real-time tracking. However, the high cost and reliance on predetermined paths limit the use of real-time kinematic (RTK) global navigation satellite system (GNSS) units, particularly in small-scale farming. Laser range finders (LRFs) with pan-tilt units offer a cost-effective alternative by generating precise three-dimensional profiles using the time-of-flight principle. Therefore, this study aimed to develop an optimized driving baseline modeling algorithm utilizing LRF technology to enhance autonomous navigation in combine harvesters.

Methods

Various automatic edge detection methods were compared based on accuracy and processing time. A crawler-type, motor-driven mobile platform equipped with a real-time controller and artificial landmarks was developed to simulate the driving baseline algorithm. Additionally, a field-test platform using a 42 kW combine harvester was established to validate the algorithm under agricultural conditions. Noise in the raw data was filtered using a modified random sample consensus (RANSAC) method. Crop edge detection was performed using a pan-tilt unit, followed by a second RANSAC filtering step to predict the driving baseline model and calculate lateral and heading deviations in real time.

Results

Experimental results demonstrated that the proposed algorithm was robust against disturbances and achieved higher accuracy than Otsu’s method, valley-emphasis method, k-means clustering, and the original triangle method under real-time conditions. The average lateral and heading deviations from straight-running tests at speeds of 2, 3, and 4 km/h were 0.139 m and 7.36°, respectively.

Conclusions

Given the standard divider interval of 30 cm in combines, the achieved accuracy confirms that this LRF-based crop edge detection algorithm is well-suited for integration into autonomous combine harvester systems.