Plant Density Estimation Using UAV Imagery and Deep Learning
6
Citation
47
Reference
10
Related Paper
Citation Trend
Abstract:
Plant density is a significant variable in crop growth. Plant density estimation by combining unmanned aerial vehicles (UAVs) and deep learning algorithms is a well-established procedure. However, flight companies for wheat density estimation are typically executed at early development stages. Further exploration is required to estimate the wheat plant density after the tillering stage, which is crucial to the following growth stages. This study proposed a plant density estimation model, DeNet, for highly accurate wheat plant density estimation after tillering. The validation results presented that (1) the DeNet with global-scale attention is superior in plant density estimation, outperforming the typical deep learning models of SegNet and U-Net; (2) the sigma value at 16 is optimal to generate heatmaps for the plant density estimation model; (3) the normalized inverse distance weighted technique is robust to assembling heatmaps. The model test on field-sampled datasets revealed that the model was feasible to estimate the plant density in the field, wherein a higher density level or lower zenith angle would degrade the model performance. This study demonstrates the potential of deep learning algorithms to capture plant density from high-resolution UAV imageries for wheat plants including tillers.Keywords:
Density estimation
Plant Density
Индекс листовой площади (LAI) и фитомасса являются основными определителями первичной чистой продукции, которые могут быть определены методами дистанционного зондирования кроны растительности. Индекс LAI отражает взаимосвязь кроны растений с оптической радиацией Солнца иявляется довольно значимым показателем для определения обмена СО2, H2O, и энергии между растениями и атмосферой. LAI также является количественным показателем сезонных изменений кроны а такжефенологии растений, которые признаны в качестве интегрированных показателей реакции растений наклиматические изменения. Этот индекс количественно может быть определен как с помощью методовбортовых измерений, так и методов наземных валидационных измерений. Вместе с тем, информация,добываемая с помощи отраженной от растительности радиации, зависит от таких показателей, как уголсолнечного освещения, фоновое отражение, угол обзора, собственные показатели фитомассы и индексLAI. Статья посвящена исследованию влияния небесной фоновой радиации на значение предлагаемогомодифицированного индекса LAI. Проведенное модельное исследование показало, что неопределенностьпредлагаемого модифицированного индекса листовой площади MLAI минимальна при убывающем видефункции зависимости FAPAR от небесной фоновой радиации, т.е. при больших зенитных углах Солнца.На основании этого сделан вывод о том, что вновь введенный индекс MLAI наиболее устойчив при больших зенитных углах Солнца. Следовательно, при наличии данных о значении небесной фоновой радиацииβ, величину LAI желательно вычислять по индексу MLAI при больших зенитных углах Солнца.
Cite
Citations (0)
å¶é¢ç§¯ææ°LAIï¼Leaf Area Indexï¼æ¯è¡¨å¾æ¤è¢«å ä½ç»æåçé¿ç¶æçéè¦çç©ç©çåæ°ï¼ä¹æ¯é表è¿ç¨æ¨¡åçéè¦è¾å ¥åæ°ï¼å¦ä½è·åé«ç²¾åº¦LAIä¸ç´å¤åå ³æ³¨ãè¿å¹´æ¥ï¼éçé¥ææ°æ®çä¸æ丰å¯ï¼LAIé¥æä¼°ç®ç®æ³å¾å°äºå¿«éåå±ï¼å ¨ç尺度çLAI产å已被广æ³åºç¨äºæ°åä¸çæç¯å¢ååç 究ãç¶èï¼å½å主æµçLAIé¥æ产åçæç®æ³åºæ¬ä¸åºäºå¹³å¦å°è¡¨å设è忽ç¥äºå°å½¢çå½±åï¼å æ¤å¨å°å½¢å¤æçå°åºç²¾åº¦è¾å·®ãè¿æ¯å 为å¨å±±å°ä¸å´å²çå°è¡¨ä¸ä» ä¼å¯¼è´ä¸¥éçè¾å°å¤±çç°è±¡ï¼è¿ä¼å é»è¿çå°å½¢å¯¹å°ç©ç®æ é æé®æ¡ï¼å æ¤æ£®æå¤æ ·çå å±ç»æåå±±å°å¤æå°å½¢çç¸äºå½±åç»LAIé¥æåæ¼å¸¦æ¥äºè¾å¤§çä¸ç¡®å®æ§ãå±±å°ä½ä¸ºä¸ç§ç¹æ®çå°è²ï¼çº¦å å ¨çéå°è¡¨é¢ç1/4ï¼å¨ä¸å½å äºè¿2/3ï¼å¨è¿äºå¤æåºåä¸ä¼°ç®LAIèèå°å½¢å ç´ ååå¿ è¦ãå¨æ¬æä¸ï¼æ们é¦å ç³»ç»å°æ»ç»äºç°æLAIåæ¼ç®æ³åå ¨çé¥æ产åçå辨çã精度çä¿¡æ¯ï¼å¹¶è®¨è®ºäºå°è¿äºç®æ³å产ååºç¨äºå´å²å°å½¢LAIåæ¼ç主è¦ææãç¶åï¼é对山å°æ¤è¢«åºæ¯ä¸åå¨çå°å½¢æåºã尺度æåºï¼æ»ç»åºå±±å°æ¤è¢«å å±LAIåæ¼ççç¥ä¸»è¦å æ¬å°å½¢æ ¡æ£æ¹æ³åå±±å°è¾å°ä¼ è¾æ¨¡åï¼å¹¶è®¨è®ºäºä¸åçç¥çä¼ç¼ºç¹ãæ¥çï¼æç« è®¨è®ºäºéå¤è§æµçLAIæ°æ®å¨å´å²å°å½¢ä¸åå¨çå°å½¢æåºå尺度æåºï¼ä»¥åè¿äºæåºå¯¹åæ¼ç»æéªè¯çå½±åç¨åº¦ãæåï¼ç»¼åæ»ç»åå±æ表æï¼é¥æè§æµãå±±å°è¾å°ä¼ è¾å»ºæ¨¡ãæºå¨å¦ä¹ ææ¯çæ¹é¢çåè°ä½¿ç¨å°æ¥å¯ä»¥ä¸ºå´å²å°è¡¨çLAIç²¾åä¼°ç®åå¯é éªè¯æä¾äºä¸æ¡æå¸æçéå¾ã
Cite
Citations (6)
Different optical instruments are currently available for measuring LAI such as LAI 2000 Plant Canopy Analyser (LAI-2000), Tracing Radiation and Architecture of Canopies (TRAC) and Digital Hemispherical Photography (DHP). Their applicability varies in different ecosystems. This study was devoted to compare LAI measured using four different methods (LAI measured by DHP, LAI measured by TRAC, LAI calculated using effective LAI measured by LAI-2000 and clumping index measured by DHP, and LAI calculated using effective LAI measured by LAI-2000 and clumping index measured by TRAC) in the Maoershan experimental forest farm of Northeast Forestry University located in Shangzhi city of Heilongjiang province. Methods used to measure LAI have considerable effects on observed LAI. The means of LAI measured by four different methods are 3.15, 4.73, 3.97, and 4.24 and corresponding standard deviations are 1.54, 2.39, 1.82, and 1.75, respectively. According to previous studies, the combination of LAI-2000 with TRAC can give the most reliable measurements of LAI. Therefore, DHP tends to underestimate LAI at this area, especially for dense canopies while TRAC tends to overestimate slightly LAI for dense canopies. The fitting of LAI measured using four different methods with normalized difference vegetation index (NDVI) and reduced simple ratio (RSR) calculated from TM data acquired on June 24, 2009 indicated that RSR is a better predictor of LAI than NDVI in this study area. The agreements between measured and estimated LAI using the best fitted models are 58%, 70%, 57% and 68% for these four different methods. Corresponding root mean square errors (RMSE) are 0.80, 0.85, 0.88, and 0.75, respectively. The regional means of LAI retrieved using the empirical models established on the basis of RSR and LAI measured with four different methods are 3.47, 5.26, 4.31, and 4.68, respectively, indicating that if DHP is used as a surrogate of TRAC and LAI-2000, LAI might be underestimated by about 25.7% in this area.
TRAC
Cite
Citations (1)
We have investigated the plant number reactions of three maize hybrids of various genotypes in a small-plot field experiment. The plant numbers were 50, 70 and 90 thousand ha-1, while the row distances were 45 and 76 cm. The experiment was set on the Látókép Experimental Farm of Centre for Agricultural Sciences of the University of Debrecen in four replications on calcareous chernozem soil.
The assimilation area and the leaf area index have important role in development of the crop yield. The studied three different genotype maize hybrids reached its maximum leaf area index at flowering. The maximum leaf area index increased linearly with increasing plant density. The season-hybrids reached less yield and leaf area index. According to our experimental results, we have concluded that with the decrease of the row spacing, the yield increased in the average of the hybrids. The studied hybrids reached the maximum yield at 70 and 90 plants ha-1 plant density. We determined the optimal plant number that is the most favourable for the certain hybrid under the given conditions.The higher plant density was favourable at 45 cm row spacing than 76 cm. The hybrids reached the maximum grain yield at 45 cm row spacing between 76 712–84 938 plants ha-1, while the optimum plant density at 76 cm row spacing changed between 61 875–65 876 plants ha-1.
The leaf area index values between the applied plant density for the flowering period (July 1, 24), we defined a significant differences. In the archived yields were significant differences at the 45 cm row spacing between 50 and 70, 90 thousand ha-1 plant density, while the number for the 76 cm row spacing used did not cause a significant differences in the yield. There were significant differences between the examined hybrids of yields.
Plant Density
Cite
Citations (0)
Interception
Specific leaf area
Spatial heterogeneity
Cite
Citations (4)
Increase of plant density with decreasing cotton row spacing has been suggested as an alternative strategy to optimize cotton profit. Although, the primary goal of this method is to reduce input cost, however, there is limited information about the agronomic and physiological aspects of these systems across the world cotton belt. In this task, three cultivation systems were studied in terms of narrow row high plant density (NRHPD; 48 cm and 32 plants/m 2 ), narrow row low plant density (NRLPD; 48 cm and 16 plants/m 2 ) and conventional row spacing (CR; 96 cm and 16 plants/m 2 ). Effects of these systems on the accumulation and allocation of dry mass as well as on leaf area index (LAI) were examined at critical growth stages during two growing seasons. Independently of row spacing, system with high plant density (NRHPD) produced significantly (P ≤ 0.001) greater dry mass and leaf area index (LAI) compared to lower plant density systems, i.e. CR and NRLPD. These differences became more significant at stage of maximum dry mass and LAI. However, this system of NRHPD partitioned the same or less dry mass to reproductive growth than other systems. Also, significant (P ≤ 0.05) differences were measured between systems with the same plant density and different row spacing, thus total dry mass and LAI were significantly higher in NRLPD than in CR system.
Dry weight
Plant Density
Cropping system
Cite
Citations (35)
Cite
Citations (159)
Global products of vegetation green Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation absorbed by vegetation (FPAR) are being operationally produced from Terra and Aqua Moderate Resolution Imaging Spectroradiometers (MODIS) at 1-km resolution and eight-day frequency. This paper summarizes the experience of several collaborating investigators on validation of MODIS LAI products and demonstrates the close connection between product validation and algorithm refinement activities. The validation of moderate resolution LAI products includes three steps: 1) field sampling representative of LAI spatial distribution and dynamic range within each major land cover type at the validation site; 2) development of a transfer function between field LAI measurements and high resolution satellite data to generate a reference LAI map over an extended area; and 3) comparison of MODIS LAI with aggregated reference LAI map at patch (multipixel) scale in view of geo-location and pixel shift uncertainties. The MODIS LAI validation experiences, summarized here, suggest three key factors that influence the accuracy of LAI retrievals: 1) uncertainties in input land cover data, 2) uncertainties in input surface reflectances, and 3) uncertainties from the model used to build the look-up tables accompanying the algorithm. This strategy of validation efforts guiding algorithm refinements has led to progressively more accurate LAI products from the MODIS sensors aboard NASA's Terra and Aqua platforms
Moderate-resolution imaging spectroradiometer
Spectroradiometer
Land Cover
Photosynthetically active radiation
Reference data
Cite
Citations (370)
Rapid and reliable estimates of leaf area index (LAI) are important for studies of exchanges of energy and gases in the biosphere-atmosphere continuum. This paper evaluates the field performance of SunScan canopy analysis system for rapid estimation of LAI. Direct and indirect measurements of LAI were made in a maize ( Zea mays L.) field at four phenological stages (emergence, vegetative, flowering and physiological maturity) at a tropical site in Ghana during the Glowa Vota Project field campaign ( www.glowa-volta.de ). Similar measurements were repeated in early and late planting seasons with similar crop management practices. The result showed a generally good performance of this sensor at all the phenological stages. Average LAI from the sensor (LAI S ), ranged from 0.40–4.45, and was consistently higher than the actual LAI, which varied from 0.31–4.22, respectively for both seasons. Regression between LAI and LAI S showed a range of significant correlations with R 2 > 0.74 for all the stages and seasons. With combine d datasets for all stages and the two plantings, a simple regression model was fitted to estimate LAI from LAI S with R 2 = 0.97 and standard error of 0.23 ( P The evaluated sensor yielded a good and reliable LAI estimates under maize canopy. Keywords: SunScan probe, field evaluation, leaf area index, maize, Ghana
Growing season
Cite
Citations (8)