Development of a Support System for Judging the Appropriate Timing for Grape Harvesting

Amemiya, Tatsuyoshi; Akiyama, Kodai; Leow, Chee Siang; Buayai, Prawit; Makino, Koji; Mao, Xiaoyang; Nishizaki, Hiromitsu

doi:10.1109/cw52790.2021.00040

Cited by 3 publications

(9 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, 9) passing the inference results back to the AI server and 10) the color estimation result is sent back to the Hololens 2 through the private 5G network for displaying the result real-time and with high accuracy was required. In our previous study [24], we used a dataset called the Embrapa Wine Grape Instance Segmentation Dataset (WGISD) [32] to train YOLOv5. This dataset contains 300 photographs of grape farms with the corresponding label data (grape bounding boxes).…”

Section: Grape Bunch Detectionmentioning

confidence: 99%

“…Considering the model size and processing speed, YOLOv5 may be suitable. Moreover, we used the same photographed bunches of grapes harvested at a Shine Muscat farm in a small darkroom from a previous study [24]. To train the YOLOv5 model for grain and pest grain detection, we added additional grain images cropped from the grape photographs mentioned in Sect.…”

Section: Normal and Pest Grain Detectionmentioning

confidence: 99%

“…This autoencoder model was trained with 23,000 normal grain images which were cut out from the grape bunch images Fig. 3 The autoencoder model for detecting diseased grains taken in a small dark room, the same condition in the previous study [24].…”

Section: Diseased Grain Detectionmentioning

confidence: 99%

“…Therefore, our proposed method used color judgments made by experienced farmers at the harvest site as labels, which were then employed to train the DNN model. In our previous study [24], we proposed a color estimation model that considered how sunlight shines on grapes. However, because the number of grapes used in the evaluation was small and multiple cuts of images (> 30) were required for each bunch to achieve accurate estimations, it was clear that this would take too much time and could not be used in practice.…”

Section: Introductionmentioning

confidence: 99%

“…In practice, the color estimation result must be output within 2 s after the grape image is captured, meaning only one image cut could be used. Accordingly, we propose a new color estimation model for the previous study [24]. Although RGB color space [25] is often used in standard image recognition tasks, we need to estimate the color of green grapes when the brightness of the pictures differs, depending on the weather and other factors.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Appropriate grape color estimation based on metric learning for judging harvest timing

et al. 2022

Self Cite

View full text Add to dashboard Cite

The color of a bunch of grapes is a very important factor when determining the appropriate time for harvesting. However, judging whether the color of the bunch is appropriate for harvesting requires experience and the result can vary by individuals. In this paper, we describe a system to support grape harvesting based on color estimation using deep learning. To estimate the color of a bunch of grapes, bunch detection, grain detection, removal of pest grains, and color estimation are required, for which deep learning-based approaches are adopted. In this study, YOLOv5, an object detection model that considers both accuracy and processing speed, is adopted for bunch detection and grain detection. For the detection of diseased grains, an autoencoder-based anomaly detection model is also employed. Since color is strongly affected by brightness, a color estimation model that is less affected by this factor is required. Accordingly, we propose multitask learning that uses metric learning. The color estimation model in this study is based on AlexNet. Metric learning was applied to train this model. Brightness is an important factor affecting the perception of color. In a practical experiment using actual grapes, we empirically selected the best three image channels from RGB and CIELAB (L*a*b*) color spaces and we found that the color estimation accuracy of the proposed multi-task model, the combination with “L” channel from L*a*b color space and “GB” from RGB color space for the grape image (represented as “LGB” color space), was 72.1%, compared to 21.1% for the model which used the normal RGB image. In addition, it was found that the proposed system was able to determine the suitability of grapes for harvesting with an accuracy of 81.6%, demonstrating the effectiveness of the proposed system.

show abstract

Section: Grape Bunch Detectionmentioning

confidence: 99%

Section: Normal and Pest Grain Detectionmentioning

confidence: 99%

Section: Diseased Grain Detectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Appropriate grape color estimation based on metric learning for judging harvest timing

et al. 2022

Self Cite

View full text Add to dashboard Cite

show abstract

Usability of Instrument for a Farm Product in a Real Farm

Makino

Nishizaki

Usui

et al. 2022

2022 IEEE 4th Global Conference on Life Sciences and Technologies (LifeTech)

View full text Add to dashboard Cite

Freshness Grading of Agricultural Products Using Artificial Intelligence

Elbir

Caferoglu

Cihan

2022

Artificial Intelligence Applications in Agriculture and Food Quality Improvement

View full text Add to dashboard Cite

This chapter presents a deep learning model that can be used to determine the levels of freshness of three agricultural products: strawberries, lemons, and tomatoes. For this purpose, YOLO, a state-of-the-art object detection algorithm is utilized. The data for training, validation, and testing are collected from online sources, and by applying image augmentation techniques, a sufficient number of images are obtained. Test results show that the model is performing quite well, and the speed of the model is fast. These results are promising and can be utilized to reduce a significant amount of agricultural waste and increase customer satisfaction once it is utilized by online groceries.

show abstract

Development of a Support System for Judging the Appropriate Timing for Grape Harvesting

Cited by 3 publications

References 17 publications

Appropriate grape color estimation based on metric learning for judging harvest timing

Appropriate grape color estimation based on metric learning for judging harvest timing

Usability of Instrument for a Farm Product in a Real Farm

Freshness Grading of Agricultural Products Using Artificial Intelligence

Contact Info

Product

Resources

About