Research Advances in Underground Bamboo Shoot Detection Methods

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article provides an overview of modern methods for accurate search and detection of winter bamboo shoots in shallow soil. Considering that the manual collection process is labor-intensive and low-productive, the authors conducted a series of field tests to evaluate the potential of alternative methods such as ground-penetrating radar (GPR), microwave detection, resistivity method, parallel electrical method, electronic nose technology, etc. Technical and economic aspects of their applicability are analyzed. The article is well structured and the conclusions are confirmed by experimental data. The obtained results are undoubtedly significant for further development of the studied problem.

Comments 

For better visual comparison it is desirable to combine Figures 2 and 3. Also, in the text of the article, please describe in more detail the features and differences of models a, b and c.

All the methods analyzed in the article for the detection of underground bamboo shoots have a number of limitations due to their cost and insufficient detection accuracy. In this regard, would it be possible to use visual features of bamboo rhizome potential locations (via leaf density and color analysis, etc.) in intelligent image analysis systems (e.g. drone or mobile device camera data) using neural network models with subsequent mapping of promising search areas?

In addition to the presented methods, it would be useful to analyze the prospects of growing bamboo shoots under artificial conditions, e.g. using aeroponics technologies.

Author Response

Dear Editors and Reviewers,

Thank you for the comments. We extend our appreciation to the editor and reviewers for dedicating their time and effort to reviewing the earlier version of the manuscript. Their valuable suggestions have facilitated improvements in our work. Based on the instructions provided from the reviewers, we uploaded the file of the revised manuscript. Accordingly, we have uploaded a file including the highlighted tracked changes in PDF. Our point-by-point response to the comments raised by the reviewers is appended to this letter. The comments are reproduced in black, and our responses are provided directly afterward in a different color (red). Furthermore, we would like to express our gratitude for granting us the opportunity to resubmit the revised manuscript.

Comment 1: For better visual comparison it is desirable to combine Figures 2 and 3. Also, in the text of the article, please describe in more detail the features and differences of models a, b and c.

Response 1: Thanks for your suggestion. We have combined Fig. 2 & 3 into anew Fig. 2 and gave more detailed description about the features and differences of the models in subsection 2.2: “Figures 2 depicts a model diagram and scanning echo diagram. The model is set up for winter bamboo shoots with the same burial depth but varying sizes. The models from left to right (a, b, c) represent winter bamboo shoots with lengths of 15, 10, and 5 cm, respectively, all buried at a depth of 5 cm. Multiple echoes are reflected from the top and bottom interfaces of the winter bamboo shoots. Longer winter bamboo shoots exhibit greater interval distances between the hyperbolic signatures at their upper and lower boundaries. Consequently, the longitudinal dimension of winter bamboo shoots can be estimated by measuring the time interval between the two hyperbolic arrivals.”

Comment 2: All the methods analyzed in the article for the detection of underground bamboo shoots have a number of limitations due to their cost and insufficient detection accuracy. In this regard, would it be possible to use visual features of bamboo rhizome potential locations (via leaf density and color analysis, etc.) in intelligent image analysis systems (e.g. drone or mobile device camera data) using neural network models with subsequent mapping of promising search areas?

Response 2: Thanks for your suggestion. The integration of visual features with intelligent imaging systems and neural network models to locate bamboo rhizome potential zones and map bamboo shoot search hotspots is a promising approach. We have added a new subsection 3.3.4 Integration of visual features with intelligent imaging analysis to discuss more about this topic: “Integrating visual features with neural network models, such as CNNs and intelligent imaging systems offers a promising solution to locate bamboo rhizome zones and map bamboo shoot search hotspots. The approach leverages deep learning architectures, such as YOLOv5 and Transformer-based models, to automate defect detection in bamboo strips, adapting principles from agricultural AI for rhizome path prediction [71]. U-Net architectures with ResNet-50 backbones are deployed for radar image segmentation. For example, a modified U-Net trained on multispectral radar data can detect subtle soil anomalies (e.g., moisture gradients) linked to rhizome activity, achieving higher precision in simulated bamboo forest trials [72]. By analyzing multispectral imagery and color anomalies, the system identifies potential shoot clusters via spatial feature sampling and query recycling mechanisms. Key advantages include cost efficiency (reducing reliance on costly ground-penetrating radar), non-invasive monitoring, and scalability for large-area surveys. Challenges such as canopy occlusion and model generalization across bamboo species (e.g., Phyllostachys vs. clumping bamboo) can be mitigated via transfer learning with cross-species datasets and attention mechanisms to focus on critical regions. Future work should prioritize edge computing for real-time processing and multimodal fusion to enhance precision. By embedding domain-specific AI architectures (CNNs, decision trees), addressing data limitations (synthetic augmentation, transfer learning), and formalizing traditional knowledge (rule engines, Reinforcement Learning), this framework strengthens the technical rigor and practical relevance of the present study. Future work should explore explainable AI to visualize model decisions and foster farmer trust.”

Comment 3: In addition to the presented methods, it would be useful to analyze the prospects of growing bamboo shoots under artificial conditions, e.g. using aeroponics technologies.

Response 3: Thanks for your suggestion. We have added a new subsection 3.3.5 about aeroponic technology for bamboo shoot cultivation: “Cultivating bamboo shoots in controlled environments, such as aeroponic systems, revolutionizes traditional farming practices and creates unique opportunities to optimize and integrate bamboo shoot detection technologies. The high controllability of aeroponic systems—including precise regulation of temperature, humidity, light cycles, and nutrient solution formulations—enables the generation of standardized growth parameter datasets [73]. These datasets provide high-quality training samples for deep learning-based phenotyping detection models. For instance, multispectral imaging sensors integrated into aeroponic facilities can capture real-time 3D morphological dynamics of bamboo shoot root systems. When combined with CNN algorithms, the data facilitate the establishment of growth rate prediction models to identify optimal harvesting windows. Moreover, the reduced morphological variability of bamboo shoots in controlled environments minimizes interference from complex natural backgrounds on image recognition models, greatly improving detection accuracy. To address the unique nutrient management requirements of aeroponics, decision tree-based rule engines can correlate root image features with nutrient absorption efficiency, enabling early warnings of growth anomalies. Future advancements should focus on developing cross-modal data fusion frameworks to synergistically analyze aeroponic environmental sensor data with visual detection metrics. This integration can support the creation of closed-loop feedback systems, transitioning detection technologies from passive recognition to proactive regulation. Such innovations can provide comprehensive, full-cycle intelligent decision-making support for intelligent bamboo shoot detection and industrial-scale production.”

We would like to express our sincere gratitude for the reviewers' insightful comments. The manuscript has undergone professional language editing by native English-speaking experts to ensure clarity and proper academic expression. Should any further refinements be required, we remain fully committed to addressing them promptly.

Sincerely,

Huimin Yang

Reviewer 2 Report

Comments and Suggestions for Authors

First, I would like to commend the author for addressing a very relevant and timely topic. The review provides a valuable perspective on the future of bamboo agriculture, especially in the context of decreasing local knowledge among traditional farmers. Your initiative to explore detection and harvesting methods for underground bamboo shoots is both innovative and necessary. However, there are a few points in the manuscript that need clarification:

• Sub-heading 2. Current status of winter bamboo shoot detection technology. This sub-chapter lists several bamboo shoot detection methods/technologies. However, if checked, the technology listed in sub-heading 2.2 to 2.4 is the potential or theory that can be used to develop bamboo shoot detection technology. Sub-heading 2.2-2.4 do not match the title "current status". The existing technology in the form of traditional methods is what is called current. Meanwhile, sub-heading 2.2 to 2.4 can be new tittle.
• In the sub-heading summary of method, the author has made a balanced review between the advantages and disadvantages of each method (Table 1). The review will be more weighty if it is associated with the diversity of landforms and climates, for example tropical areas, mountains, lowlands, etc. This review will be an introduction in the sub-heading development prospect (line 319).
• Sub heading development prospect should be enriched with diversity of land forms and climate

Author Response

Dear Editors and Reviewers,

Thank you for the comments. We extend our appreciation to the editor and reviewers for dedicating their time and effort to reviewing the earlier version of the manuscript. Their valuable suggestions have facilitated improvements in our work. Based on the instructions provided from the reviewers, we uploaded the file of the revised manuscript. Accordingly, we have uploaded a file including the highlighted tracked changes in PDF. Our point-by-point response to the comments raised by the reviewers is appended to this letter. The comments are reproduced in black, and our responses are provided directly afterward in a different color (red). Furthermore, we would like to express our gratitude for granting us the opportunity to resubmit the revised manuscript.

Comment 1: Sub-heading 2. Current status of winter bamboo shoot detection technology. This sub-chapter lists several bamboo shoot detection methods/technologies. However, if checked, the technology listed in sub-heading 2.2 to 2.4 is the potential or theory that can be used to develop bamboo shoot detection technology. Sub-heading 2.2-2.4 do not match the title "current status". The existing technology in the form of traditional methods is what is called current. Meanwhile, sub-heading 2.2 to 2.4 can be new tittle.

Response 1: Thanks for your suggestion. We have modified the title of section 2 to “Existing technologies for winter bamboo shoot detection” for exact phrasing.

Comment 2: In the sub-heading summary of method, the author has made a balanced review between the advantages and disadvantages of each method (Table 1). The review will be more weighty if it is associated with the diversity of landforms and climates, for example tropical areas, mountains, lowlands, etc. This review will be an introduction in the sub-heading development prospect (line 319).

Sub heading development prospect should be enriched with diversity of land forms and climate.

Response 2: Thanks for your suggestion. We have added a new paragraph in subsection 3.3 to discuss the technologies addressing the diversity of land forms and climate: “The efficacy of bamboo shoot detection systems is profoundly influenced by terrain and climatic variations, driving innovations in adaptive technologies and multi-sensor integration. In tropical lowlands, high humidity and soil acidification reduce microwave radar penetration depth, necessitating hybrid systems that combine shortwave infrared spectroscopy (SWIR) for moisture-compensated biomass mapping. Mountainous regions demand cold-resistant probes and solar-wind hybrid power solutions to counteract low temperatures and UV radiation. Flood-prone lowlands face salt crust interference, prompting salt-penetrating SWIR bands and AI-driven waterlogging alerts. In extreme climates, IoT vibration sensors can be employed to monitor rhizome water stress thresholds, while hurricane-affected Philippine groves require light detection and ranging-based 3D canopy reconstruction post-storm. These regional adaptations highlight a shift from single-method solutions to multimodal architectures that integrate climate-resilient hardware and adaptive algorithms.”

We would like to express our sincere gratitude for the reviewers' insightful comments. The manuscript has undergone professional language editing by native English-speaking experts to ensure clarity and proper academic expression. Should any further refinements be required, we remain fully committed to addressing them promptly.

Sincerely,

Huimin Yang

Reviewer 3 Report

Comments and Suggestions for Authors

1. In the introduction, the review presents four primary categories of bamboo shoot detection methods manual, microwave-based, resistivity-based, and biomimetic. While the discussion provides a clear overview of these approaches, it largely remains descriptive rather than analytical. A more thorough comparative evaluation would enhance the contribution of the paper. Incorporating quantitative data such as detection accuracy, depth penetration, adaptability to varying soil conditions, and cost-effectiveness would help readers better assess the advantages and limitations of each method.
2. Although the review highlights the growing role of artificial intelligence (AI) in this domain, the treatment of AI applications is somewhat superficial. It would strengthen the manuscript to include specific examples of AI models that could be applied, such as convolutional neural networks (CNNs) for radar-based imaging or decision trees for fusing multi-sensor inputs. Further discussion of challenges related to data availability, the need for real-time inference, and strategies for integrating the practical knowledge of experienced bamboo farmers perhaps through rule-based or hybrid learning systems would also be beneficial.
3. The manuscript briefly notes that much of the current research is still confined to laboratory settings or simulations. However, it would be valuable to more directly address the discrepancy between experimental findings and real-world applicability. Challenges such as diverse soil types across geographic regions, signal interference from environmental noise, and complex terrain should be acknowledged to give a more complete picture of the practical limitations and readiness of these technologies.
4. Labor-related issues, such as the aging workforce in bamboo shoot harvesting, are mentioned, but the economic feasibility of implementing intelligent detection systems is not thoroughly explored. A discussion of factors like return on investment (ROI), potential subsidies, and accessibility for small-scale farmers could provide useful insights for stakeholders considering adoption of such technologies.
5. Finally, while the manuscript is well-structured and clearly written overall, a few editorial improvements would further enhance readability. For example, figure captions could be expanded to better explain their content and relevance. The summary table might be strengthened by adding a column that reflects the current development stage of each method (concept, prototype, commercially available). Additionally, breaking Section 3 into clearly labeled subsections would help improve navigation and comprehension.

Author Response

Dear Editors and Reviewers,

Thank you for the comments. We extend our appreciation to the editor and reviewers for dedicating their time and effort to reviewing the earlier version of the manuscript. Their valuable suggestions have facilitated improvements in our work. Based on the instructions provided from the reviewers, we uploaded the file of the revised manuscript. Accordingly, we have uploaded a file including the highlighted tracked changes in PDF. Our point-by-point response to the comments raised by the reviewers is appended to this letter. The comments are reproduced in black, and our responses are provided directly afterward in a different color (red). Furthermore, we would like to express our gratitude for granting us the opportunity to resubmit the revised manuscript.

Comment 1: In the introduction, the review presents four primary categories of bamboo shoot detection methods manual, microwave-based, resistivity-based, and biomimetic. While the discussion provides a clear overview of these approaches, it largely remains descriptive rather than analytical. A more thorough comparative evaluation would enhance the contribution of the paper. Incorporating quantitative data such as detection accuracy, depth penetration, adaptability to varying soil conditions, and cost-effectiveness would help readers better assess the advantages and limitations of each method.

Response 1: Thanks for your suggestion. We have added a new table 2 with quantitative data to illustrate the advantages and limitations of each method.

Comment 2: Although the review highlights the growing role of artificial intelligence (AI) in this domain, the treatment of AI applications is somewhat superficial. It would strengthen the manuscript to include specific examples of AI models that could be applied, such as convolutional neural networks (CNNs) for radar-based imaging or decision trees for fusing multi-sensor inputs. Further discussion of challenges related to data availability, the need for real-time inference, and strategies for integrating the practical knowledge of experienced bamboo farmers perhaps through rule-based or hybrid learning systems would also be beneficial.

Response 2: Thanks for your suggestion. We have added a new subsection 3.3.4 to discuss more about this topic: “Integrating visual features with neural network models, such as CNNs and intelligent imaging systems offers a promising solution to locate bamboo rhizome zones and map bamboo shoot search hotspots. The approach leverages deep learning architectures, such as YOLOv5 and Transformer-based models, to automate defect detection in bamboo strips, adapting principles from agricultural AI for rhizome path prediction [71]. U-Net architectures with ResNet-50 backbones are deployed for radar image segmentation. For example, a modified U-Net trained on multispectral radar data can detect subtle soil anomalies (e.g., moisture gradients) linked to rhizome activity, achieving higher precision in simulated bamboo forest trials [72]. By analyzing multispectral imagery and color anomalies, the system identifies potential shoot clusters via spatial feature sampling and query recycling mechanisms. Key advantages include cost efficiency (reducing reliance on costly ground-penetrating radar), non-invasive monitoring, and scalability for large-area surveys. Challenges such as canopy occlusion and model generalization across bamboo species (e.g., Phyllostachys vs. clumping bamboo) can be mitigated via transfer learning with cross-species datasets and attention mechanisms to focus on critical regions. Future work should prioritize edge computing for real-time processing and multimodal fusion to enhance precision. By embedding domain-specific AI architectures (CNNs, decision trees), addressing data limitations (synthetic augmentation, transfer learning), and formalizing traditional knowledge (rule engines, Reinforcement Learning), this framework strengthens the technical rigor and practical relevance of the present study. Future work should explore explainable AI to visualize model decisions and foster farmer trust.”

Comment 3: The manuscript briefly notes that much of the current research is still confined to laboratory settings or simulations. However, it would be valuable to more directly address the discrepancy between experimental findings and real-world applicability. Challenges such as diverse soil types across geographic regions, signal interference from environmental noise, and complex terrain should be acknowledged to give a more complete picture of the practical limitations and readiness of these technologies.

Response 3: Thanks for your suggestion. We have added a new paragraph to explain the discrepancy between experimental methods and real application in subsection 3.1: “Existing bamboo shoot detection methods offer foundational insights but lack critical comparative metrics and real-world validation. Manual detection achieves not low accuracy but is labor-intensive and prone to human error, while microwave radar reaches much higher accuracy for shallow rhizomes but struggles in saline soils due to signal attenuation. Resistivity sensors face interference from root systems in organic soils, and biomimetic technologies require extensive training data and struggle with cross-species generalization. Field validations reveal significant gaps: the accuracy of laboratory-optimized microwave systems is dropped much in heterogeneous terrains, and biomimetic technologies overfit synthetic data, underperforming in real forests. Challenges like canopy occlusion and soil variability further limit practicality. Future work must prioritize edge-optimized AI and hybrid sensor fusion to address these translational bottlenecks, alongside explainable AI frameworks to enhance farmer trust and ecological compliance.”

Comment 4: Labor-related issues, such as the aging workforce in bamboo shoot harvesting, are mentioned, but the economic feasibility of implementing intelligent detection systems is not thoroughly explored. A discussion of factors like return on investment (ROI), potential subsidies, and accessibility for small-scale farmers could provide useful insights for stakeholders considering adoption of such technologies.

Response 4: Thanks for your suggestion. We have added a new subsection 3.2 to explain the feasibility of intelligent detector: “The economic viability of intelligent bamboo shoot detector hinges on three pillars: return on investment (ROI), policy incentives, and accessibility for smallholders. From an ROI perspective, these systems reduce labor costs by automating shoot detection and health monitoring, resulting in 40%–60% labor savings and ¥800–1,200/ha annual savings, coupled with yield improvements of 20%–30% via optimized harvesting. Initial deployment costs (≈¥15,000–30,000/ha) are typically offset within 3–5 years in high-yield regions. Policy supports further enhances feasibility: governments offer subsidies (e.g., 30% Internet of Things (IoT) equipment rebates) and low-interest loans, while insurance programs mitigate revenue risks. For smallholders, cooperative procurement models (e.g., 40% cost-sharing for sensor networks) and shared-use platforms (e.g., mobile detection units) lower entry barriers. However, challenges persist, including limited rural internet coverage, technical training gaps, and maintenance costs. Solutions such as leasing models (¥200–500/month/ha) and vernacular AI interfaces address these barriers to a certain extent. Strategic investments in rural digital infrastructure and cross-sector partnerships will be critical to democratize access and ensure sustainable adoption.”

Comment 5: Finally, while the manuscript is well-structured and clearly written overall, a few editorial improvements would further enhance readability. For example, figure captions could be expanded to better explain their content and relevance. The summary table might be strengthened by adding a column that reflects the current development stage of each method (concept, prototype, commercially available). Additionally, breaking Section 3 into clearly labeled subsections would help improve navigation and comprehension.

Response 5: Thanks for your suggestion. 1) We have checked and modified the figure captions for better understanding; 2) We have added an extra column about the status of intelligent methods in Table 1; 3) Section 3 is separated into three subsections, 3.1, 3.2 and 3.3 which consists of five subsections 3.3.1~3.3.5.

We would like to express our sincere gratitude for the reviewers' insightful comments. The manuscript has undergone professional language editing by native English-speaking experts to ensure clarity and proper academic expression. Should any further refinements be required, we remain fully committed to addressing them promptly.

Sincerely,

Huimin Yang

Reviewer 4 Report

Comments and Suggestions for Authors

The study “Research advances in underground bamboo shoot detection methods” is an interesting review article. The authors discuss the use of artificial intelligence and 3D methods for harvesting, data evaluation, optimizing planting practices, etc. However, there is room for improvement, which can be addressed.

1. First of all, I would like to recommend revising the abstract carefully to include the main objectives of the overview, specifically the scientific findings, rather than writing general statements. The current condition is too general and not scientific.
2. I must say, the introduction is too short and does not adequately discuss the topics, nor does it sufficiently explain why a review of this particular topic is needed. Also, the focus should be global rather than only on China.
3. Some content seems to be generated by a large language model. Need to revise
4. While the authors discussed some methods in detail, it's also useful to know some AI algorithms used in this area.
5. All the figures must be self-explanatory. For example, figure 9 is missing details for captions (a) and (b), as well as information on the x and y axes.

Author Response

Dear Editors and Reviewers,

Thank you for the comments. We extend our appreciation to the editor and reviewers for dedicating their time and effort to reviewing the earlier version of the manuscript. Their valuable suggestions have facilitated improvements in our work. Based on the instructions provided from the reviewers, we uploaded the file of the revised manuscript. Accordingly, we have uploaded a file including the highlighted tracked changes in PDF. Our point-by-point response to the comments raised by the reviewers is appended to this letter. The comments are reproduced in black, and our responses are provided directly afterward in a different color (red). Furthermore, we would like to express our gratitude for granting us the opportunity to resubmit the revised manuscript.

Comment 1: First of all, I would like to recommend revising the abstract carefully to include the main objectives of the overview, specifically the scientific findings, rather than writing general statements. The current condition is too general and not scientific.

Response 1: Thanks for your suggestion. We have rewritten the abstract with explicit objectives of this review and detailed proposals for potential technologies of intelligent detection.

Comment 2: I must say, the introduction is too short and does not adequately discuss the topics, nor does it sufficiently explain why a review of this particular topic is needed. Also, the focus should be global rather than only on China.

Response 2: Thanks for your suggestion. We have rewritten the introduction and fully discussed the topic. Meanwhile, we have expanded the scope of the discussion on a global scale.

Comment 3: Some content seems to be generated by a large language model. Need to revise

Response 3: We have invited a professional language editing team to revise the language of the manuscript.

Comment 4: While the authors discussed some methods in detail, it's also useful to know some AI algorithms used in this area.

Response 4: Thanks for your suggestion. We have added subsection 3.3.4 & 3.3.5 covering AI technologies, especially CNN models, for signal processing and intelligent image analysis.

Comment 5: All the figures must be self-explanatory. For example, figure 9 is missing details for captions (a) and (b), as well as information on the x and y axes.

Response 5: Thanks for your suggestion. We have added remarks after all the figure titles to make the figures self-explanatory.

We would like to express our sincere gratitude for the reviewers' insightful comments. The manuscript has undergone professional language editing by native English-speaking experts to ensure clarity and proper academic expression. Should any further refinements be required, we remain fully committed to addressing them promptly.

Sincerely,

Huimin Yang

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Accepted

Reviewer 4 Report

Comments and Suggestions for Authors

In this article authors reported existing methods for detecting and harvesting underground winter bamboo shoots, including manual, microwave, resistivity, and biomimetic techniques. It found, while manual harvesting is still common, it is inefficient and damaging; microwave and resistivity methods show promise but have practical limitations, and biomimetic approaches are innovative but not yet fully effective. The authors proposed an integrated intelligent system using AI, multi-sensor fusion, and smart cultivation to improve detection accuracy, reduce labor, and increase yield.
In this version, authors addressed all of my concerns and comments, with substantial improvement.