All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
The reviewers are generally satisfied with the amendments made to the paper. There are some minor comments still regarding the use of the language, which need to be resolved prior to publication.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
I believe that the revised manuscript has resolved all issues and meets the journal's standards. I recommend that the paper be accepted.
no comment
no comment
no comment
no comment
no comment
no comment
This revised manuscript solved my confusion and can be published.
no comment
no comment
no comment
The language needs to be further improved.
Just for examples:
1. "The physical encoder Ephi represents a point by a vector with seven features..." should be "The physical encoder Ephi represents a point as a vector with seven features..."
2. "We conduct much in-depth analysis, showing the necessity and further verifying the effectiveness of our method." can be streamlined to "We conduct in-depth analysis to verify the necessity and effectiveness of our method."
3. "The image generation module generates an image for each point in a trajectory, which contains the contextual information for the point..." This could be clearer as "The image generation module creates an image for each trajectory point, containing the contextual information specific to that point..."
4. "We used the BiLSTM interface provided by Pytorch (a python-based deep learning library)..." should be "We used the BiLSTM interface provided by PyTorch (a Python-based deep learning library)..."
5. "We conducted much in-depth analysis..." should be "We conducted an in-depth analysis..."
6." Figure 5. Results of different field-road classification methods on two harvesting trajectory samples. Satellite imagery ©2024 Google Earth Engine; Map data ©2024 Google." "on" should be changed to "for"
7. "The impact of the data limitation can be alleviated." could be more accurately phrased as "The impact of data limitations can be mitigated."
8. "To effectively extract visual features, we propose a task-specific approach based on a general pre-trained visual model." could be more concise: "We propose a task-specific approach using a pre-trained visual model to effectively extract visual features."
and many more .....
The reviewers have highlighted some weaknesses in the paper of which the Authors are encouraged to look into and revise. Addressing these comments will enhance the manuscript’s clarity and impact.
Authors propose a multi-view field-road trajectory segmentation method for solving movement patterns of agricultural machinery operation. This study presents a certain level of innovation in the model, and the experimental dataset and results substantiate the high practical and scholarly value of the proposed method. However，there are some problems to be solved：
1. In abstract and method parts, as the core contribution to study, the more details regarding to the pretraining-finetuning paradigm required to be presented, including briefly calculation process, etc. In addition, more results data should be presented in the abstract
2. In the introduction, references required for “As observed in previous studies.”.
3. The author proposed in the introduction that Field-road trajectory segmentation can also be called field-road classification, which is not rigorous. From the perspective of model calculation, the research task is more like a classification topic, and it is inappropriate to name it field-road segmentation, and a modification of the name is recommended.
4. In Multi-view encoder part, it appears that only one view of the model is presented?
5. We praise the authors for not deliberately seeking complex models, but proposing a novel feature processing approach to use appropriate models to achieve better performance. However, the reasons for choosing ResNet and BiLSTM need further explanation, that is, why these two classic models are considered in the study.
6. Although this approach is a novel and impressive one, the limitations and future directions of this line of study should be added to the discussion part.
7.The author needs to provide the hardware environment for the experiment
8.We appreciate the author's utilization of a publicly available dataset for the experiment, which enhances the reliability of the obtained results. Nonetheless, it has come to our attention that the majority of trajectories included in the dataset were collected in 2019. We suggest that the author use data from the past two years, i.e. 2023 and 2024, to conduct experiments and public them.
The structure of this paper is clear, the language is fluent, and it provides sufficient background in the field and references. The content of the figures and tables is well organized, and the relevant results for hypotheses are self-contained. Moreover, the authors have shared the raw data.
The experimental design of this paper is comprehensive and presents an original study that falls within the aims and scope of the journal. The generation of background images for trajectory points to aid field-road trajectory segmentation is an innovative idea. The methods are described with detail and information to replicate.
The trajectory data used in this paper has been made open source by the authors, and all the basic data have been provided. The conclusions are appropriately stated and relevant to the original research question.
In this paper, the authors present a multi-view field-road trajectory segmentation method, which extracts two feature vectors to represent a trajectory point: a physical feature vector which encodes the physical motion characteristic of the point, and a visual feature vector which represents the shape characteristic in the context of the point. The experiments have shown performance improvements on both wheat and paddy harvesting trajectory datasets. The main innovation of this paper is the generation of images that reflect the shape characteristics of each trajectory point, opening up new research directions in the field.
The overall work is comprehensive and has achieved favorable experimental results. This paper brings new research methods to the field road trajectory segmentation task. After the author addresses the following issues, this paper can be accepted.
(1) The description of the main contributions in the introduction section should be more detailed.
(2) Present the calculation process of the RGB three-channel values more clearly.
(3) Give the details of the datasets, such as the number of trajectories and points.
(4) Give the software and hardware configurations used in the paper, which is crucial for replicating the experiment.
(5) Figure 1 can be replaced with real trajectory data to reflect the actual scenario of field road segmentation.
1. Review the manuscript for any overly complex sentences and consider simplifying them for better readability. for instance, the sentence "To capture movement patterns specific to agricultural operations, we propose a multi-view field-road trajectory segmentation method, which extracts two feature vectors to represent a trajectory point: a physical feature vector which encodes the physical motion characteristic of the point, and a visual feature vector which represents the shape characteristic in the context of the point." is long and contains multiple clauses, which can make it difficult to follow. All in all, the English writing needs to be improved.
2. The introduction provides a good context for the study, explaining the importance of GNSS trajectories in agricultural vehicle behavior analysis. However, the introduction could benefit from a more detailed explanation of the existing challenges and how the proposed method addresses them and its segmentation performance.
3.The figures included are relevant and support the text well, but some figure captions could be more descriptive. For example, the description of figure 1 could be " An example trajectory of an agricultural vehicle, illustrating the distinction between field and road segments. (a) Field segment: This sub-trajectory, depicted with red points, follows a pattern characteristic of agricultural operations, comprising alternating straight trips and turning curves. The straight trips indicate operational passes within the field, while the curves represent turns at the end of each pass. (b) Road segment: This sub-trajectory, depicted with green points, represents the vehicle's movement on a road, characterized by a more linear and continuous path without the alternating pattern seen in field operations. ", which is more informative and descriptive to readers.
1. Did you compare the model performance between those with pre-training and those without pre-training?
2. I don't see any figures displaying your segmentation results.
1. The manuscript could benefit from a more detailed discussion on the limitations and potential future work
2.The manuscript encourages replication by providing detailed methods and data availability
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.