Research Progress on Fruit Maturity Detection Technologies for Intelligent Harvesting

doi:10.16420/j.issn.0513-353x.2025-0864

Acta Horticulturae Sinica ›› 2026, Vol. 53 ›› Issue (10): 3148-3166.doi: 10.16420/j.issn.0513-353x.2025-0864

Previous Articles Next Articles

Research Progress on Fruit Maturity Detection Technologies for Intelligent Harvesting

QIAN Tingting¹^,², JIA Xiaoze³^,⁴, LI Yiyang²^,³, SUN Yue¹^,², LI Guo³^,⁴, LU Shenglian³^,⁴^,^**()

¹ Institute of Agricultural Science and Information， Shanghai Academy of Agricultural Sciences， Shanghai 201403， China
² Shanghai Digital Agriculture Engineering Technology Research Center， Shanghai 201403， China
³ School of Computer Science and Engineering， Guangxi Normal University，Guilin， Guangxi 541004， China
⁴ Guangxi Key Laboratory of Multisource Information Mining and Security， Guilin， Guangxi 541004， China

Received:2026-07-16 Revised:2026-08-17 Online:2026-10-25 Published:2026-10-26
Contact: LU Shenglian

Abstract

Abstract:

Fruit maturity detection is one of the key technologies for automated fruit harvesting. Accurate assessment of fruit maturity status can provide reliable decision-making information for harvesting robots，thereby improving harvesting efficiency，and fruit commercial value. Traditional approaches，which relied on manual inspection or rule-based image analysis，were found to be inefficient and unreliable in complex orchard environments due to occlusion，variable illumination，and diverse fruit characteristics. In this review，unimodal strategies based on RGB images，hyperspectral sensing，and depth information were discussed in terms of their ability to capture appearance，internal quality，and structural features. Multimodal fusion approaches，integrating RGB，hyperspectral，depth，tactile，and olfactory data，were discussed to enhance robustness under complex field conditions. Advances in attention mechanisms，lightweight model design，and edge-cloud collaborative deployment were highlighted as effective solutions to meet the requirements of real-time agricultural operations. Publicly available datasets covering multiple fruits were also surveyed，with emphasis on their annotation strategies and contributions to model generalization. Remaining challenges，including limited cross-scenario adaptability，high annotation cost，and the need for efficient deployment on resource-constrained devices，were summarized，and future directions toward the deep integration of fruit maturity detection with intelligent harvesting robots and digital agriculture platforms are proposed.

Key words: fruit maturity, object detection, deep learning, multimodal fusion, intelligent picking, computer vision

CLC Number:

QIAN Tingting, JIA Xiaoze, LI Yiyang, SUN Yue, LI Guo, LU Shenglian. Research Progress on Fruit Maturity Detection Technologies for Intelligent Harvesting[J]. Acta Horticulturae Sinica, 2026, 53(10): 3148-3166.

Figures/Tables 10

References 50

[1]	Attri I, Awasthi L, Sharma T P, Rathee P. 2023. A review of deep learning techniques used in agriculture. Ecological Informatics,77:102217.
[2]	Blekos A, Chatzis K, Kotaidou M, Chatzis T, Solachidis V, Konstantinidis D, Dimitropoulos K. 2023. A grape dataset for instance segmentation and maturity estimation. Agronomy, 13 (8):1995. doi: 10.3390/agronomy13081995 URL
[3]	Chen S, Xiong J, Jiao J, Xie Z, Huo Z, Hu W. 2022. Citrus fruits maturity detection in natural environments based on convolutional neural networks and visual saliency map. Precision Agriculture, 23 (5):1515-1531. doi: 10.1007/s11119-022-09895-2
[4]	Chen W, Rao Y, Wang F, Zhang Y, Wang T, Jin X, Hou W, Jiang Z, Zhang W. 2024. MLP-based multimodal tomato detection in complex scenarios:insights from task-specific analysis of feature fusion architectures. Computers and Electronics in Agriculture,221:108951.
[5]	Coleman G R Y, Kutugata M, Walsh M J, Bagavathiannan M V. 2024. Multi-growth stage plant recognition:a case study of Palmer amaranth (Amaranthus palmeri) in cotton(Gossypium hirsutum). Computers and Electronics in Agriculture,217:108622.
[6]	Cossio-Montefinale Luis, Ruiz-Del-Solar Javier, Verschae Rodrigo. 2024. Cherry CO Dataset:a dataset for cherry detection,segmentation and maturity recognition. IEEE Robotics and Automation Letters, 9 (6):5552-5558. doi: 10.1109/LRA.2024.3393214 URL
[7]	Dai Haichen, Wei Xinyu, Xu Yixin, Chen Yuanping, Xu Yuanyuan, Ji Ying. 2024. Multimodal fusion detection method of tomato fruit based on phase and hyperspectral. Acta Photonica Sinica, 53 (7):0710001. (in Chinese) doi: 10.3788/gzxb URL
	戴海宸, 韦鑫宇, 徐一新, 陈元平, 徐媛媛, 季颖. 2024. 基于相位和高光谱的番茄果实多模态融合检测方法. 光子学报, 53 (7):0710001.
[8]	Dai Shuo, Bai Tao, Zhao Yunjie. 2025. Keypoint detection and 3D localization method for ridge-cultivated strawberry harvesting robots. Agriculture, 15 (4):372. doi: 10.3390/agriculture15040372 URL
[9]	Elhariri Esraa, El-Bendary Nashwa, Mahmoud Saleh Samir. 2023. Strawberry-DS:dataset of annotated strawberry fruits images with various developmental stages. Data in Brief,48:109165.
[10]	Fass E, Shlomi E, Ziv C, Glikman O,& Helman D. 2025. Machine learning models based on hyperspectral imaging for pre-harvest tomato fruit quality monitoring. Computers and Electronics in Agriculture,229:109788.
[11]	Friha O, Ferrag M A, Shu L, Maglaras L, Wang X. 2021. Internet of things for the future of smart agriculture:a comprehensive survey of emerging technologies. IEEE/CAA Journal of Automatica Sinica,8:718-752.
[12]	Garillos-Manliguez C A, Chiang J Y. 2021a. Multimodal deep learning via late fusion for non-destructive papaya fruit maturity classification//18th International Conference on Electrical Engineering, Computing Science and Automatic Control(CCE). IEEE:1-6.
[13]	Garillos-Manliguez C A, Chiang J Y. 2021b. Multimodal deep learning and visible-light and hyperspectral imaging for fruit maturity estimation. Sensors, 21 (4):1288. doi: 10.3390/s21041288 URL
[14]	Gené-Mola J, Ferrer-Ferrer M, Hemming J, van Dalfsen P, de Hoog D, Sanz-Cortiella R, Rosell-Polo J R, Morros J-R, Vilaplana V, Ruiz-Hidalgo J, Gregorio E. 2024. AmodalAppleSize_RGB-D dataset:RGB-D images of apple trees annotated with modal and a modal segmentation masks for fruit detection,visibility and size estimation. Data in Brief,52:110000.
[15]	Huang M L, Chen Y A. 2023. Citrus dataset for image classification. Data in Brief,51:109628.
[16]	Husein I R, Shiddiq M, Sari D L, Putri A. 2022. Wavelength dependence of optical electronic nose for ripeness detection of oil palm fresh fruits. Science,Technology,and Communication Journal, 2 (3):79-86.
[17]	Jiang Yefan, Jiang Zhaohui, Zhu Rui, Yong Yan. 2025. Strawberry ripeness in a complex environment based on PCIA-YOLO lightweight detection method. Joumal of Nanjing Agricultural University, 48 (6):1451-1463. (in Chinese)
	姜业帆, 江朝晖, 朱瑞, 闫勇. 2025. 基于PCIA-YOLO的复杂环境下草莓成熟度轻量化检测方法. 南京农业大学学报, 48 (6):1451-1463.
[18]	Kang S, Fan J, Ye Y, Li C, Du D, Wang J. 2025. Maturity recognition and localisation of broccoli under occlusion based on RGB-D instance segmentation network. Biosystems Engineering,250:270-284.
[19]	Liu Y, Wei C, Yoon S C, Ni X, Wang W, Liu Y, Wang D, Wang X, Guo X. 2024. Development of multimodal fusion technology for tomato maturity assessment. Sensors, 24 (8):2467. doi: 10.3390/s24082467 URL
[20]	Liu Yongmin, Zhang Wei, Ma Haizhi, Yuan liu, Yi Zhang. 2024. Lightweight YOLO v5s blueberry detection algorithm based on attention mechanism. Journal of Henan Agricultural Sciences, 53 (3):151-157. (in Chinese) doi: 10.15933/j.cnki.1004-3268.2024.03.016
	刘拥民, 张炜, 麻海志, 刘原, 张毅. 2024. 基于注意力机制的轻量化YOLO v5s蓝莓检测算法. 河南农业科学, 53 (3):151-157.
[21]	Lu S L, Lv Y B, Qian T T, Ren W Y, Li X M, Li Y Y, Li G. 2025. Lettuce architectural phenotypes extraction from multimodal images by low-light sensitivity and strong spatial perception. Computers and Electronics in Agriculture,232:109928.
[22]	Naqvi L H, Balasubramaniam B, Li J, Liu L, Li B. 2025. Four-dimensional hyperspectral imaging for fruit and vegetable grading. Agriculture, 15 (15):1702. doi: 10.3390/agriculture15151702 URL
[23]	Pai C, Lin H Y. 2025. Maturity and yield estimation of tomatoes using RGB and multispectral images// 2024 IEEE International Conference on Industrial Technology(ICIT). Bristol:IEEE:1-6.
[24]	Qiao J, Su G, Liu C, Zou Y, Chang Z, Yu H, Wang L, Guo R. 2022. Study on the application of electronic nose technology in the detection for the artificial ripening of crab apples. Horticulturae, 8 (5):386. doi: 10.3390/horticulturae8050386 URL
[25]	Ren Jingqiu, Wan Enhan, Shan Mi, Zhang Guanghua, Lu Weidang. 2025. Improved YOLO v8n network for real-time detection of tomato maturity. Transactions of the Chinese Society for Agricultural Machinery, 56 (3):374-382,450. (in Chinese)
	任晶秋, 万恩晗, 单蜜, 张光华, 卢为党. 2025. 基于改进YOLO v8n网络的番茄成熟度实时检测算法. 农业机械学报, 56 (3):374-382,450.
[26]	Saranya N, Srinivasan K, Kumar S K Pravin. 2022. Banana ripeness stage identification:a deep learning approach. Journal of Ambient Intelligence and Humanized Computing, 13 (8):4033-4039. doi: 10.1007/s12652-021-03267-w
[27]	Summaira J, Li X, Shoib A M, Abdul J. 2022. A review on methods and applications in multimodal deep learning. ACM Transactions on Multimedia Computing, Communications,and Applications, 19 (2s):1-41.
[28]	Shamrat F M J M, Shakil R, Idris M Y I, Bonna A, Zhou X J. 2024. FruitSeg30_Segmentation dataset & mask annotations:a novel dataset for diverse fruit segmentation and classification. Data in Brief,56:110821.
[29]	Song Huaibo, Ma Baoling, Shang Yuying, Wen Yuchen, Zhang Shujin. 2023. Detection of young apple fruits based on YOLO v7-ECA model. Transactions of the Chinese Society for Agricultural Machinery, 54 (6):233-242. (in Chinese)
	宋怀波, 马宝玲, 尚钰莹, 温毓晨, 张姝瑾. 2023. 基于YOLO v7-ECA模型的苹果幼果检测. 农业机械学报, 54 (6):233-242.
[30]	Tyagi P, Semwal R, Sharma A, Tiwary U S, Varadwaj P. 2023. E-nose:a low-cost fruit ripeness monitoring system. Journal of Agricultural Engineering,213:108202.
[31]	Visentin F, Castellini F, Muradore R. 2023. A soft,sensorized gripper for delicate harvesting of small fruits. Computers and Electronics in Agriculture,213:108202.
[32]	Wang Hongen, Feng Guohong, Xu Huadong, Zhang Runze. 2024. Identification of blueberry ripeness based on visible-near spectroscopy and spectral analysis and deep forest. Spectroscopy and Spectral Analysis, 44 (11):3280-3286. (in Chinese)
	王宏恩, 冯国红, 徐华东, 张润泽. 2024. 基于可见—近红外光谱和深度森林的蓝莓成熟度判别. 光谱学与光谱分析, 44 (11):3280-3286.
[33]	Wang J X, Liu M, Du Y R, Zhao N H, Jia H L, Guo Z, Su Y H, Lu D Z, Liu Y C. 2024a. PG-YOLO:An efficient detection algorithm for pomegranate before fruit thinning. Engineering Applications of Artificial Intelligence,134:108700.
[34]	Wang L, Zhu Y, Gu S, Chen Z. 2023. Real-time fruit detection method based on RGB-D image fusion// 2023 10th International Conference on Dependable Systems and Their Applications(DSA). IEEE:156-166.
[35]	Wang N, Qian T T, Yang J, Li L Y, Zhang Y Y, Zheng X G, Xu Y Y, Zhao H Q, Zhao J Y. 2022. An enhanced YOLOv5 model for greenhouse cucumber fruit recognition based on color space features. Agriculture,12,1556.
[36]	Wang S, Jiang H, Yang J, Ma X, Chen J, Li Z and, Tang X. 2024b. Lightweight tomato ripeness detection algorithm based on the improved RT-DETR. Frontiers in Plant Science,15:1415297.
[37]	Wei J, Ni L, Luo L, Chen M, You M, Sun Y, Hu T. 2024. GFS-YOLO11:a maturity detection model for multi-variety tomato. Agronomy, 14 (11):2644. doi: 10.3390/agronomy14112644 URL
[38]	Wu M, Qiu Y, Wang W, Su X, Cao Y, Bai Y. 2025. Improved RT-DETR and its application to fruit ripeness detection. Frontiers in Plant Science,16:1423682.
[39]	Wu Z, Liu M, Sun C, Wang X. 2023. A dataset of tomato fruits images for object detection in the complex lighting environment of plant factories. Data in Brief,48:109291.
[40]	Xi Xiaobo, Ding Jieyuan, Weng Xiaoweng, Wang Yu, Han Lianjie, Zou Yunhan, Tang Zihao, Zhang Ruihong. 2025. Tomato maturity detection method based on lightweight YOLO v5s-MCA. Transactions of the Chinese Society for Agricultural Machinery, 56 (3):383-391,436. (in Chinese)
	奚小波, 丁杰源, 翁小祥, 王昱, 韩连杰, 邹贇涵, 唐子昊, 张瑞宏. 2025. 基于轻量化YOLO v5s-MCA的番茄成熟度检测方法. 农业机械学报, 56 (3):383-391,436.
[41]	Xiong Juntao, Han Yonglin, Wang Xiao, Li Zexing, Chen Haoran, Huang Qiyin. 2023. Method of maturity detection for papaya fruits in natural environment based on YOLO v5-Lite. Transactions of the Chinese Society for Agricultural Machinery, 54 (6):243-252. (in Chinese)
	熊俊涛, 韩咏林, 王潇, 李泽星, 陈浩然, 黄启寅. 2023. 基于YOLO v5-Lite的自然环境木瓜成熟度检测方法. 农业机械学报, 54 (6):243-252.
[42]	Xu D, Ren R, Zhao H, Zhang S. 2024. Intelligent detection of muskmelon ripeness in greenhouse environment based on YOLO-RFEW. Agronomy, 14 (6):1091. doi: 10.3390/agronomy14061091 URL
[43]	Yan B, Li X M. 2024. RGB-D camera and fractal-geometry-based maximum diameter estimation method of apples for robot intelligent selective graded harvesting. Fractal and Fractional, 8 (11):649. doi: 10.3390/fractalfract8110649 URL
[44]	Ye B F, Xue R Z, Xu H Q. 2025. ASD-YOLO:a lightweight network for coffee fruit ripening detection in complex scenarios. Frontiers in Plant Science,16:1484784.
[45]	Yin Shuyu, Ruan Ningjun. 2025. A lightweight strawberry ripeness detection method based on YOLO11n. Jiangsu Journal of Agricultural Science, 41 (10):1997-2008. (in Chinese)
	尹书宇, 阮宁君. 2025. 基于YOLO11n的轻量级草莓成熟度检测方法. 江苏农业学报, 41 (10):1997-2008.
[46]	Zahidi U A, Łukasik K, Cielniak G. 2024. Dual-band feature selection for maturity classification of specialty crops by hyperspectral imaging. arXiv preprint arXiv:2405.09955.
[47]	Zaid O, Abdul Majeed A P P, Rosbi M, Ghazalli S A, Selamat H. 2024. Outdoor oil palm fruit ripeness dataset. Data in Brief,55:110667.
[48]	Zhao J, Almodfer R, Wu X, Wang X. 2023a. A dataset of pomegranate growth stages for machine learning-based monitoring and analysis. Data in Brief,50:109468.
[49]	Zhao Z, Hicks Y, Sun X, Luo C. 2023b. Peach ripeness classification based on a new one-stage instance segmentation model. Computers and Electronics in Agriculture,214:108369.
[50]	Zheng Wenxuan, Yang Ying. 2024. Mature stage pear detection method based on frequency domain data augmentation and lightweight YOLO v7 model. Transactions of the Chinese Society for Agricultural Machinery, 55 (5):244-253. (in Chinese)
	郑文轩, 杨瑛. 2024. 基于频域数据增强与轻量化YOLO v7模型的成熟期香梨目标检测方法. 农业机械学报, 55 (5):244-253.

果实种类 Fruit species	成熟阶段特征描述 Descriptions of different maturity stages	形状变化 Shape variation	纹理变化 Texture changes
番茄 Tomato	坐果期（绿色）、膨大期（绿色）、转色期（粉红/浅红色）、成熟期（红色）Fruit setting stage（green)，fruit enlargement stage（green)，color-breaking stage （pink/light red)，ripening stage（red）	前期快速膨胀，后期略膨胀 Rapid enlargement in the early stage，followed by slight enlargement	表皮光滑、发亮 Smooth and glossy skin
草莓 Strawberry	白熟期（白色）、转熟期（白+红）、全熟期（红色）White stage（white），turning stage（white + red），fully ripe stage（red）	逐渐变大、椭圆趋近球形 Gradually enlarges，with the elliptical shape approaching a spherical shape	果面籽下沉、毛刺减少 Seeds sink into the fruit surface，with fewer protruding hairs
香蕉 Banana	绿熟期（绿色）、黄熟期（黄绿色）、完熟期（黄色）、过熟期（褐色）Green-ripe stage（green），yellow-ripe stage（yellow-green），fully ripe stage（yellow），overripe stage（brown）	长度和粗度增加，弯曲加剧 Increased length and thickness，with more pronounced curvature	出现褐斑 Brown spots appear
柑橘 Citrus	绿果期（绿色）、半成熟期（黄绿色）、成熟期（黄色）（Chen et al.，2022）Green-fruit stage（green），semi-ripe stage（yellow-green），ripe stage（yellow）	前期快速膨胀，后期略膨胀 Rapid enlargement in the early stage，followed by slight enlargement	果皮光滑、变薄 Smooth and thinner peel

果实种类 Fruit species	成熟阶段特征描述 Descriptions of different maturity stages	形状变化 Shape variation	纹理变化 Texture changes
番茄 Tomato	坐果期（绿色）、膨大期（绿色）、转色期（粉红/浅红色）、成熟期（红色）Fruit setting stage（green)，fruit enlargement stage（green)，color-breaking stage （pink/light red)，ripening stage（red）	前期快速膨胀，后期略膨胀 Rapid enlargement in the early stage，followed by slight enlargement	表皮光滑、发亮 Smooth and glossy skin
草莓 Strawberry	白熟期（白色）、转熟期（白+红）、全熟期（红色）White stage（white），turning stage（white + red），fully ripe stage（red）	逐渐变大、椭圆趋近球形 Gradually enlarges，with the elliptical shape approaching a spherical shape	果面籽下沉、毛刺减少 Seeds sink into the fruit surface，with fewer protruding hairs
香蕉 Banana	绿熟期（绿色）、黄熟期（黄绿色）、完熟期（黄色）、过熟期（褐色）Green-ripe stage（green），yellow-ripe stage（yellow-green），fully ripe stage（yellow），overripe stage（brown）	长度和粗度增加，弯曲加剧 Increased length and thickness，with more pronounced curvature	出现褐斑 Brown spots appear
柑橘 Citrus	绿果期（绿色）、半成熟期（黄绿色）、成熟期（黄色）（Chen et al.，2022）Green-fruit stage（green），semi-ripe stage（yellow-green），ripe stage（yellow）	前期快速膨胀，后期略膨胀 Rapid enlargement in the early stage，followed by slight enlargement	果皮光滑、变薄 Smooth and thinner peel

技术 Technology	光谱分辨率（示例） Spectral resolution（example）	空间分辨率（示例）Spatial resolution（example）	采集速度 Acquisition speed	典型成本 Typical cost	适用场景 Application scenarios	优缺点要点 Key advantage and disadvantage
高光谱 Hyperspectral imaging（HSI）	高（≥ 100个连续波段，带宽0.5 ~ 10.0 nm） High（≥ 100 contiguous bands，bandwidth 0.5-10.0 nm）	中 ~ 高（取决于镜头与传感器） Medium-high（depending on the lens and sensor）	慢（线扫描或立方体采集，受带数影响） Slow（line scanning or hyperspectral cube acquisition；affected by the number of bands）	高 High	实验室分析、品质检测、科研 Laboratory analysis，quality assessment，scientific research	优：光谱信息丰富、用于定量分析缺：成本高、体积大、对运动敏感 Advantages：rich spectral information，suitable for quantitative analysis Disadvantages：high cost，large size，sensitive to motion
多光谱 Multispectral imaging（MSI）	中（通常3 ~ 10个窄带） Medium（typically 3-10 narrow bands）	高（可采用面阵成像） High（area-array imaging can be used）	快（面阵快门） Fast（area-array shutter）	中 Medium	田间监测、快速筛查 Field monitoring，rapid screening	优：成本适中、实时性好缺：波段有限，定量能力不如HSI Advantages：moderate cost，good real-time performance Disadvantages：limited number of bands，weaker quantitative capability than HSI
近红外 Near-infrared（NIR）	低 ~ 中（单波段或少数波段，700 ~ 1 100 nm常见） Low-medium（single or a few bands，commonly 700-1 100 nm）	高（与单目相机一致） High（similar to a monocular camera）	快（面阵） Fast（area-array）	低 ~ 中 Low-medium	水分/糖度粗略预测、工业分拣 Rough prediction of moisture/sugar content，industrial sorting	优：穿透浅层组织、便携缺：信息量有限，受表皮影响大 Advantages：penetrates shallow tissues，portable Disadvantages：limited information，strongly affected by the skin

技术 Technology	光谱分辨率（示例） Spectral resolution（example）	空间分辨率（示例）Spatial resolution（example）	采集速度 Acquisition speed	典型成本 Typical cost	适用场景 Application scenarios	优缺点要点 Key advantage and disadvantage
高光谱 Hyperspectral imaging（HSI）	高（≥ 100个连续波段，带宽0.5 ~ 10.0 nm） High（≥ 100 contiguous bands，bandwidth 0.5-10.0 nm）	中 ~ 高（取决于镜头与传感器） Medium-high（depending on the lens and sensor）	慢（线扫描或立方体采集，受带数影响） Slow（line scanning or hyperspectral cube acquisition；affected by the number of bands）	高 High	实验室分析、品质检测、科研 Laboratory analysis，quality assessment，scientific research	优：光谱信息丰富、用于定量分析缺：成本高、体积大、对运动敏感 Advantages：rich spectral information，suitable for quantitative analysis Disadvantages：high cost，large size，sensitive to motion
多光谱 Multispectral imaging（MSI）	中（通常3 ~ 10个窄带） Medium（typically 3-10 narrow bands）	高（可采用面阵成像） High（area-array imaging can be used）	快（面阵快门） Fast（area-array shutter）	中 Medium	田间监测、快速筛查 Field monitoring，rapid screening	优：成本适中、实时性好缺：波段有限，定量能力不如HSI Advantages：moderate cost，good real-time performance Disadvantages：limited number of bands，weaker quantitative capability than HSI
近红外 Near-infrared（NIR）	低 ~ 中（单波段或少数波段，700 ~ 1 100 nm常见） Low-medium（single or a few bands，commonly 700-1 100 nm）	高（与单目相机一致） High（similar to a monocular camera）	快（面阵） Fast（area-array）	低 ~ 中 Low-medium	水分/糖度粗略预测、工业分拣 Rough prediction of moisture/sugar content，industrial sorting	优：穿透浅层组织、便携缺：信息量有限，受表皮影响大 Advantages：penetrates shallow tissues，portable Disadvantages：limited information，strongly affected by the skin

技术 Technology	精度（典型） Typical accuracy	有效测距范围 Effective measurement range	光照敏感性 Lighting sensitivity	功耗/复杂度 Power consumption/ Complexity	典型成本 Typical cost	适用场景/备注 Application scenarios/ Remarks
结构光（如Kinect v1） Structured light（e.g.，Kinect v1）	中 ~ 高（毫米级，近距离）Medium-high（millimeter-level，short range）	0.5 ~ 4.0 m	高（受环境光影响）High（affected by ambient light）	中 Medium	低 ~ 中 Low- medium	室内/受控环境效果好；对阳光直射敏感 Good performance indoors/in controlled environments；sensitive to direct lighting
TOF（如Intel RealSense） Time-of-flight（TOF；e.g.，Intel RealSense）	中（毫米 ~ 厘米级）Medium（millimeter-centimeter level）	0.2 ~ 10.0 m（取决于型号） 0.2-10.0 m （depending on the model）	低 ~ 中（抗光照干扰较佳）Low-medium（relatively good resistance to illumination interference）	中 ~ 高 Medium-high	中 ~ 高 Medium-high	适合户外短距测距，抗光照设计差异较大 Suitable for short-range outdoor distance measurement；resistance to illumination varies considerably by design
双目立体视觉 Binocular stereo vision	可变（依算法与基线，误差随深度增长）Variable（depending on the algorithm and baseline；error increases with depth）	可变（取决于相机基线）Variable （depending on the camera baseline）	低（依赖纹理；逆光/低纹理场景表现差）Can be long（depending on the camera baseline）	低 ~ 中 Low（texture- dependent；poor performance under backlighting/low- texture conditions）	低 Low- medium	成本低，适合远距、多视角场景；需要较强的匹配算法 Low cost；suitable for long-distance and multi-view scenarios；requires strong matching algorithms

Research Progress on Fruit Maturity Detection Technologies for Intelligent Harvesting

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 50

Related Articles 2

Recommended Articles

Metrics

Comments

对象 Object	模型 Model	主要改进与优化 Main improvement and optimization	性能指标 Performance metrics	文献 Reference
草莓 Strawberry	YOLOv11	利用mobileNetV4替换主干网络，提出新的特征融合方法Bi-Freq替换原模型中颈部特征融合方法。 Uses MobileNetV4 to replace the backbone network and proposes the Bi-Freq feature fusion method to replace the original neck feature fusion method.	mAP：97.8%	尹书宇和阮宁君,2025；
番茄 Tomato	GFS-YOLO11	提出C3k2_Ghost模块、FRM特征细化模块，引入SPPFELAN模块进行多尺度特征融合 Proposes the C3k2_Ghost module and FRM feature refinement module，and introduces the SPPFELAN module for multi-scale feature fusion	mAP50：93.4%	Wei et al.,2024
	PDSI-RTDETR	PConv_Block模块、可变形注意力机制、slimneck-SSFF特征融合结构、Inner-EIoU损失函数 PConv_Block module，deformable attention mechanism，slimneck-SSFF feature fusion structure，and Inner-EIoU loss function	mAP50：86.8%	Wang et al.,2024b
番茄、香蕉 Tomato，banana	RT-DETR	HG Block增强、引入Rep Block、PConv替代传统卷积、EMA引入 HG Block enhancement；introduces Rep Block；replaces conventional convolution with PConv；introduces EMA	mAP0.5提高2.9%，模型大小减少5.5%，计算复杂度降低9.6% mAP0.5 increased by 2.9%；model size reduced by 5.5%；computational complexity reduced by 9.6%	Wu et al.,2025
甜瓜 Melon	YOLO-RFEW	RFAConv替换主干Conv，C2f-FE模块结合EMA注意力，优化损失函数为WIoU Replaces the backbone Conv with RFAConv；combines the C2f-FE module with EMA attention；optimizes the loss function to WIoU	mAP：90.82%	Saranya et al.,2022；Xu et al.,2024
石榴 Pomegranate	PG-YOLO	ShuffleNetv2替换骨干网络，Depthwise separable卷积替换标准卷积，引入MHSA，数据增强 Replaces the backbone network with ShuffleNetV2；replaces standard convolution with depthwise separable convolution；introduces MHSA and data augmentation	mAP：93.4%	Wang et al.,2024a
咖啡 Coffee	ASD-YOLO	设计L-Norm Attention机制，引入SPD-Conv，用DySample替换上采样 Designs an L-Norm Attention mechanism；introduces SPD-Conv；replaces upsampling with DySample	mAP：80.1%	Ye et al.,2025

技术类型 Technology type	检测对象 Detection object	典型方法与模型 Typical method and model	成本 Cost	适用场景 Application scenario	典型优势 Key advantage	主要局限 Main limitation
RGB图像检测 RGB image detection	外观成熟度（颜色、形态） External ripeness（color，morphology）	Faster R-CNN、YOLOv8、ResNet分类网络 Faster R-CNN，YOLOv8，ResNet classification networks	低 Low	果园实时检测、移动机器人 Real-time orchard detection，mobile robots	结构简单、成本低、部署灵活 Simple structure，low cost，flexible deployment	受光照和遮挡影响大，无法反映内在成熟度 Strongly affected by illumination and occlusion；unable to reflect internal ripeness
多光谱检测 Multispectral detection	果皮色变、早期成熟信号 Peel color changes，early ripening signals	基于SVM或轻量CNN的波段特征融合模型 Band-feature fusion models based on SVM or lightweight CNNs	中 Medium	实验室检测、初级分选 Laboratory detection，primary sorting	波段可调、抗光照干扰 Adjustable bands，resistance to illumination interference	波段有限，难以捕捉连续光谱特征 Limited bands，making it difficult to capture continuous spectral features
高光谱检测 Hyperspectral detection	内在理化特征（糖度、硬度、含水量） Internal physicochemical properties（sugar content，firmness，moisture content）	1D-CNN、3D-CNN、Transformer光谱模型 1D-CNN，3D-CNN，Transformer spectral models	高 High	实验室精细分析 Fine-grained laboratory analysis	信息丰富、可量化化学指标 Rich information，quantifiable chemical indicators	成本高、体积大、采集速度慢 High cost，large size，slow acquisition speed
近红外检测 Near-infrared detection	水分、糖度、内部结构 Moisture，sugar content，internal structure	PLSR、LSTM光谱回归模型 PLSR，LSTM spectral regression models	中 Medium	分选线检测、局部检测 Sorting-line detection，local inspection	穿透性强、稳定性好 Strong penetration，good stability	空间分辨率低、对表皮状态敏感 Low spatial resolution，sensitive to skin condition

数据集名称 Dataset name	果实种类 Fruit type	图像数 Number of images	标注类型 Annotation type	模态 Modality	采集地 Acquisition location	采集设备/传感器型号 Acquisition device/sensor model	采集时间跨度 Acquisition time span	环境干扰类型 Type of environmental interference	数据源 Data source
Tomato- Plantfactory	番茄 Tomato	520	VOC/YOLO	RGB	中国 China	Canon EOS 5D Mark IV（静态光照Static illumination）	单季（夏季） Single season（summer）	背景复杂、光照变化 Complex background，varying illumination	https://data. mendeley.com/datasets/8h3s6jkyff
Strawberry- DS	草莓 Strawberry	247	分类/框标注 Classification/ bounding-box annotation	RGB	埃及 Egypt	Nikon D750 + LED	多季（温室与露地） Multiple seasons（greenhouse and open field）	遮挡严重、反光 Severe occlusion，specular reflection	https://data. mendeley.com/datasets/z6dtfdpzz8
CitrusFruit	柑橘 Citrus	1 379 1,379	分类 Classification	RGB	中国台湾 Taiwan China	Canon EOS 600D	单季（室外） Single season（outdoor）	阳光直射、阴影干扰 Direct sunlight，shadow interference	https://data. mendeley.com/datasets/bxfgvsn9kw
OilPalm-FFB	油棕 Oil palm	466	分类 Classification	RGB	马来西亚 Malaysia	Sony α7R II	多季（热带高湿）Multiple seasons（tropical，high humidity）	雨雾、光照不均 Rain/fog，uneven illumination	https://zenodo.org/records/11114885
Amodal- Apple_ RGB-D	苹果 Apple	6 656	分割 + 尺寸估计 Segmentation + size estimation	RGB-D	西班牙/荷兰 Spain / the Netherlands	Intel RealSense D435	多年（2018— 2021） Multiple years（2018—2021）	遮挡、光照不均 Occlusion，uneven illumination	https://dataverse.csuc.cat/dataset.xhtml?persistentId=doi:10.34810/data916
Cherry CO	樱桃 Cherry	3 006	分割 + 成熟度 Segmentation + maturity	RGB	智利 Chile	Canon EOS 5D Mark III	多季（露地） Multiple seasons（open field）	背景杂枝、光照强 Cluttered branches/background，strong illumination	https://rodrigo.verschae.org/cherryCO
Pomegranate	石榴 Pomegranate	5 857	框标注 Bounding box annotation	RGB	中国 China	Huawei P30 Pro	单季（田间自然光） Single season （natural field illumination）	光照变化、杂草干扰 Varying illumination，weed interference	https://data.mendeley.com/datasets/kgwsthf2w6
CERTH Grape	葡萄 Grape	2 500	掩码/分割 Mask / segmentation	RGB	希腊 Greece	Basler ace camera + light-controlled enclosure	多季（实验室与田间） Multiple seasons （laboratory and field）	遮挡、反光 Occlusion，specular reflection	https://zenodo.org/records/10777647
NinePeach	桃 Peach	3 849	实例分割 Instance segmentation	RGB	中国 China	Canon EOS M50	单季（果园） Single season （orchard）	枝叶遮挡 Occlusion by branches and leaves	https://drive.google.com/drive/folders/1vCSoqVGWhy4pvyEVlW-oLH8uN2S_nUKo

[1]	SUN Yuchao, LI Shouhao, XIA Xiubo, YANG Wei, LI Minzan, ZHANG Huanchun. Detecting Tomato Fruit Ripeness and Appearance Quality Based on Improved YOLOv5s [J]. Acta Horticulturae Sinica, 2024, 51(2): 396-410.
[2]	CHAI A-li;LI Bao-ju;SHI Yan-xia;CEN Zhe-xin;HUANG Hai-yang;and LIU Jun. Recognition of Tomato Foliage Disease Based on Computer Vision Technology [J]. ACTA HORTICULTURAE SINICA, 2010, 37(9): 1423-1430.