Scholarlink

<p>Most yield estimation in orchards is performed manually, with the vital limitations of requiring experts, making subjective decisions and incurring elevated charges. This problem has been gradually overcome through modern farming techniques by developing an intelligent model for fruit yield estimation using deep learning, one of the emerging areas due to its high precision results. This study focuses on estimating yield through MangoYieldNet, which detects and counts mangoes from tree images collected from the orchards using DeepLabv3+ (a DL-based semantic segmentation architecture). After the sampling process, a total of 556 images (from 278 trees) were captured and annotated. Using image augmentation techniques (reflection, rotation and translation) a dataset of 1152 images was developed; it was split in an 80:20 ratio for training and validation, respectively. DeepLabv3+ with ResNet18 consists of 100 deep layers including convolutional layers, ReLU activation function, batch normalization, pooling, fully connected and softmax layers by which feature extraction can be done smoothly without missing any important feature that corresponds to fruits even in most of the challenging conditions of orchards. The accuracy values of 99.96% and 99.47% were obtained in the training and validation phases, respectively. The proposed method outperformed other DL-based semantic segmentation architectures in terms of accurate object delineation and shorter execution time. The architecture was tested with 30 unobserved images and revealed a mean accuracy and a mean intersection over union of 96.5% and 95.97%, respectively. Finally, the segmented fruits were counted using the circle Hough transform method. The mango count was compared with the manual count obtained from the farmers and provided the regression coefficient value of 0.98. Then, the final yield in kilograms was estimated for the sampled trees. The error percentage between actual and predicted weight is 4.99%. A user-friendly graphical user interface was developed for estimating mango yield.</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

MangoYieldNet: Fruit yield estimation for mango orchards using DeepLabv3 + with ResNet18 architecture

  • 期刊论文

  • 发表日期：

    2025年4月2日

查看全文
  • Prabhakar Maheswari,
  • Purushothaman Raja,
  • Sudha Natarajan

摘要

Most yield estimation in orchards is performed manually, with the vital limitations of requiring experts, making subjective decisions and incurring elevated charges. This problem has been gradually overcome through modern farming techniques by developing an intelligent model for fruit yield estimation using deep learning, one of the emerging areas due to its high precision results. This study focuses on estimating yield through MangoYieldNet, which detects and counts mangoes from tree images collected from the orchards using DeepLabv3+ (a DL-based semantic segmentation architecture). After the sampling process, a total of 556 images (from 278 trees) were captured and annotated. Using image augmentation techniques (reflection, rotation and translation) a dataset of 1152 images was developed; it was split in an 80:20 ratio for training and validation, respectively. DeepLabv3+ with ResNet18 consists of 100 deep layers including convolutional layers, ReLU activation function, batch normalization, pooling, fully connected and softmax layers by which feature extraction can be done smoothly without missing any important feature that corresponds to fruits even in most of the challenging conditions of orchards. The accuracy values of 99.96% and 99.47% were obtained in the training and validation phases, respectively. The proposed method outperformed other DL-based semantic segmentation architectures in terms of accurate object delineation and shorter execution time. The architecture was tested with 30 unobserved images and revealed a mean accuracy and a mean intersection over union of 96.5% and 95.97%, respectively. Finally, the segmented fruits were counted using the circle Hough transform method. The mango count was compared with the manual count obtained from the farmers and provided the regression coefficient value of 0.98. Then, the final yield in kilograms was estimated for the sampled trees. The error percentage between actual and predicted weight is 4.99%. A user-friendly graphical user interface was developed for estimating mango yield.