Smartphone Based Fluorescence Imaging for Online Control of Cattle Fodder Preparation

Round 1

Reviewer 1 Report

In this manuscript, the authors proposed a simple and cost-effective technique for online monitoring of grist concentration in cattle food mixture. The similarity in color of the Rapeseed grist and maize makes it difficult to distinguish them from a color image. But the difference in chlorophyll concentration provides a direction for solving this problem. Along this line, the authors proposed a fluorescence approach and obtained a good result.
However, some descriptions are incomplete in several areas, detracting from the clarity of the work. The following points need to be addressed in order to provide more detail required to make the paper appropriate for publication in this journal.
 1. Some grammar problems.
  • For the 6th line of the 2nd paragraph in section ‘Results and Discussion’, It can be clearly observed … correspond to…Here, correspond should be corresponds.
  • For the sentence ‘the and result is presented in Fig. 5 c’ above the Figure 5. I think the author wants to express ‘the result’.
  • For the sentence ‘The total mass of the each sample was about 1 kg’, ‘the’ before each sample should be deleted.
 2. In the ‘References’ section, each reference was numbered twice. There are 38 references in the main manuscript, so ‘39’ should be removed from the ‘References’ section.
 3. For quantitative determination of rapeseed grist mass fraction, the author measured ten times to get statistically meaningful results. Then a linear function fitting operation was used to get the mathematic relationship between the ‘signal (pixel fraction)’ and ‘grist mass fraction’. I suggest the author explain why a linear fitting is used here. Because of the same diameter for all the grist pellets?
 4. Again, for quantitative determination, Let’s imaging two different cases. For a same grist pellet, in the first case, the orientation of the grist is parallel to the camera. In the second case, the orientation is perpendicular to the camera. For these two cases, the grist signal (fraction of red pixels in the processed fluorescence image) is different, but the quantity is the same. I suggest the author explain the limitations of the proposed method.

Author Response

Answer for the referee’s comments

Authors: Vasily N. Lednev, Ivan A. Kucherenko, Vladislav A. Levshin, Pavel A. Sdvizhenskii, Mikhail Ya. Grishin, Alexey S. Dorohov, Sergey M. Pershin

Title: Smartphone based fluorescence imaging for online control of cattle fodder preparation

 

We’ve marked our answers and all the corrections in the manuscript with red color.

In the answers, we used italics to highlight the text which has been added/changed in the manuscript.

 

Reviewer #1:

In this manuscript, the authors proposed a simple and cost-effective technique for online monitoring of grist concentration in cattle food mixture. The similarity in color of the Rapeseed grist and maize makes it difficult to distinguish them from a color image. But the difference in chlorophyll concentration provides a direction for solving this problem. Along this line, the authors proposed a fluorescence approach and obtained a good result.

However, some descriptions are incomplete in several areas, detracting from the clarity of the work. The following points need to be addressed in order to provide more detail required to make the paper appropriate for publication in this journal.

1. Some grammar problems.

• For the 6th line of the 2nd paragraph in section ‘Results and Discussion’, It can be clearly observed … correspond to…Here, correspond should be corresponds.
• For the sentence ‘the and result is presented in Fig. 5 c’ above the Figure 5. I think the author wants to express ‘the result’.
• For the sentence ‘The total mass of the each sample was about 1 kg’, ‘the’ before each sample should be deleted.

Answer:

Thank you for careful reading. We have corrected the text accordingly

 

2. In the ‘References’ section, each reference was numbered twice. There are 38 references in the main manuscript, so ‘39’ should be removed from the ‘References’ section.

Answer:

This is due to reference manager software (Mendeley). In the revised version we have corrected this problem.

 

3. For quantitative determination of rapeseed grist mass fraction, the author measured ten times to get statistically meaningful results. Then a linear function fitting operation was used to get the mathematic relationship between the ‘signal (pixel fraction)’ and ‘grist mass fraction’. I suggest the author explain why a linear fitting is used here. Because of the same diameter for all the grist pellets?

Answer:

With the proper mixing the number of grist pellets at the surface is linearly proportional to the volume fraction of the grist. Supposing constant grist density, its constant diameters (it is true due to production in extrusion machine) and absence of pellets crashing during mixing process (typically, grist pellets did not crashed in the mixing machine used in the farms) the number of pellets at the mixture surface should be linear dependent on mass fraction.

Following the referee’s recommendation we have added small discussion on this topic.

Added:

… due to linear dependence of the number of grist pellets at the surface and grist mass fraction for proper mixing and absence of pellets crushing. For conventional mixers in the farms both conditions (homogeneous mixture and no pellets milling) are typically fulfilled.

 

4. Again, for quantitative determination, Let’s imaging two different cases. For a same grist pellet, in the first case, the orientation of the grist is parallel to the camera. In the second case, the orientation is perpendicular to the camera. For these two cases, the grist signal (fraction of red pixels in the processed fluorescence image) is different, but the quantity is the same. I suggest the author explain the limitations of the proposed method.

Answer:

We agree that orientation of the pellets can influence the results is small area is imaged. However this influence can be diminished by large area mapping (2000 cm2 or greater in our experiments) supposing that grist pellets distributed randomly in the mixture. Typically, commercial mixer machines provide very uniform distribution of pellets in the mixture during few dozens of minutes. In our experiments we used hand mixing of the pellets in the bucket (we added ~1 kg of silage and required amount of grist into 10 l bucket) and achieved uniform distribution after 5-10 minutes of mixing. Then we deposited mixture at the table surface (>50x50 cm2 square) and acquired fluorescent image. Afterward we mixed for 2 minutes and made a new fluorescence image and so on until 10 images were captured. With such sampling we achieved accuracy of 2.5%. In practice the imaged area is greater 1 m2 so getting few fluorescence images is enough to achieve few percent accuracy.

Of course if pellets will be crushed then our instrument will provide false results (underestimated real mass fraction). Poor mixing will resulted in false results provided by the developed instrument. However, the instrument can determine the minimum time required to get uniform distribution of components in the mixture: large oscillations of the measured mass fraction should be flatten out.

Followed referee’s comment we added discussion on this topic in the manuscript:

If pellets will be crushed during mixing then false results will be achieved since pellet surface fraction is not proportional to the mass of the grist. However, poor mixing can be traced by the developed instrument: large oscillations of the measured mass fraction at the mixing start will be continuously flatten out until uniform grist distribution will be achieved.

Author Response File: Author Response.docx

Reviewer 2 Report

The article is framed very carelessly:

- there is no affiliation of A. S. Dorokhov,

- there is no line numbering,

- incorrect descriptions of sources [1] and [2].

 

1. Excessive citation [1-9], [10-16], while there are no sources of information on the lamps and LEDs used.

2. How were the samples mixed? How were the mixing conditions used in agricultural practice taken into account?

3. Incorrect use of the term "pump source". We should not talk about pumping, but about excitement.

4. A tungsten lamp cannot have a color temperature of 5500K.

5. The authors indicate the presence of a tungsten lamp radiation spectrum in Appendix 1, but there is no such spectrum there.

6. Explain in more detail the need to use the OS-13 light filter.

7. The wrong caption to Figure 2: which "Transmission spectra of the LED lamp" are we talking about?

8. It is necessary to separate Figure 3: separate photos and graphs. The photos are very small. Why combine (b) and (c) in Figure 4? Why duplicate (a) in Figure 4?

9. Why is 450nm taken as the excitation wavelength?

10. The authors' statement "The orange color filter absorbs in 400-500 nm spectral range (see Complementary materials A1), so there is almost zero intensity in fluorescence spectrum." is not clear.

11. The authors' statement about the cheapness of the developed device should be accompanied by a comparison of its price with the modern analogues used.

Author Response

Answer for the referee’s comments

Authors: Vasily N. Lednev, Ivan A. Kucherenko, Vladislav A. Levshin, Pavel A. Sdvizhenskii, Mikhail Ya. Grishin, Alexey S. Dorohov, Sergey M. Pershin

Title: Smartphone based fluorescence imaging for online control of cattle fodder preparation

 

Please, check the attached file with the answers due to the figures modified. 

We’ve marked our answers and all the corrections in the manuscript with red color.

In the answers, we used italics to highlight the text which has been added/changed in the manuscript.

Reviewer #2

The article is framed very carelessly:

- there is no affiliation of A. S. Dorokhov,

- there is no line numbering,

- incorrect descriptions of sources [1] and [2].

Answer:

Thank you very much for careful reading. We have corrected the manuscript according to referee’s 2 recommendations.

A.S. Dorohov’s affiliation have been added: FSBSI «Federal scientific agroengineering center VIM», Moscow, Russia

Sorry, for line numbering

We have corrected the references 2 and flushed the reference 1:

National Research Council. Nutrient requirements of small ruminants: sheep, goats, cervids, and new world camelids; The National Academies Press: Washington, DC, USA, 2007; 384 p. ISBN 0309102138. https://doi.org/10.17226/11654

 

1. Excessive citation [1-9], [10-16], while there are no sources of information on the lamps and LEDs used.

Answer: We cannot fully agree with the referee’s comment. It is true that references 1-16 didn’t contain the original papers on LED/lamp utilization for sensing some cattle food control. However, these references contained the important review focused on developing new instrumentation for smart agriculture. In our opinion, it is important to highlight the study motivation in order to get the understanding that smart agriculture is a very fast growing technology nowadays and new sensors are needed to solve new challenges for individual feeding and control in cattle production industry (both milk and meat production). Still, following the referee’s comment we checked the references again and flushed 5 of the 1-16 references.

 

2. How were the samples mixed? How were the mixing conditions used in agricultural practice taken into account?

Answer: We have developed the instrument for online control of homogeneous mixture during cow food preparation in commercial mixer at the farms so in the lab we tried to simulate the conditions as close as possible. Typically, commercial mixer machines provide very uniform distribution of pellets in the mixture during few dozens of minutes. In our experiments we used hand mixing of the pellets in the bucket (we added ~1 kg of silage and required amount of grist into 10 l bucket) and achieved uniform distribution after 5-10 minutes of mixing. Then we deposited mixture at the table surface (>50x50 cm2 square) and acquired fluorescent image. Afterward we mixed for 2 minutes and made a new fluorescence image and so on until 10 images were captured. With such sampling we achieved accuracy of 2.5%. In practice the imaged area is greater 1 m2 so getting few fluorescence images is enough to achieve few percent accuracy.

We have described it in the manuscript:

The total mass of each sample was about 1 kg and in order to get a uniform distribution the sample was mixed for 15 minutes. The sample mixture was deposited at the laboratory table to cover a 50x50 cm square. The captured image almost fit the square for proper representation of the sample. Each sample mixture was deposited and then a fluorescence image was captured. The procedure was repeated ten times (mixing, deposition, fluorescence imaging) to get statistically meaningful results. Then fluorescence images were processed as described above. A grist signal was defined as a fraction of red pixels in the processed a fluorescence image (see Fig.5 d). The grist calibration curve was constructed for the samples with different grist mass fraction, and the result is presented in Fig. 6. The calibration curve was then fitted with a linear function due to linear dependence of the number of grist pellets at the surface and grist mass fraction for proper mixing and absence of pellets crushing. For conventional mixer trailers in the farms both conditions (homogeneous mixture and no pellets milling) are typically fulfilled.

 

3. Incorrect use of the term "pump source". We should not talk about pumping, but about excitement.

Answer: Agree with the comment. It seems that we take it from the laser physics but here term “excitation” is more convenient. We have corrected the text accordingly and flushed away the term “pump” and its derivatives.

Corrected:

The instrument was based on a tungsten incandescent/light emitting diode (LED) lamp as a source for excitation and …

For fluorescence excitation we used a commercially available grow light …

… 450 nm blue LED lamp for fluorescence excitation and …

… LED excitation wavelength …

… with grow light lamp excitation and …

 

4. A tungsten lamp cannot have a color temperature of 5500K.

 

Answer:

We agree with the referee’s comment. We use the tungsten lamp with the temperature of 2200 K. We have corrected the text accordingly:

Color imaging was carried out with the tungsten lamp (color temperature 2200 K) …

 

5. The authors indicate the presence of a tungsten lamp radiation spectrum in Appendix 1, but there is no such spectrum there.

Answer:

Thank you for careful reading. We have added the tungsten lamp spectrum in appendix A. and redraw the figure.

A1 (a) – Tungsten lamp emission spectrum. (b) - Grow light emission spectrum with two types of light emitting diodes (LED) at 450 and 630 nm. (c) – Color glass optical filters transmission spectra to filter out LED excitation wavelength (SZS-8) and to suppress elastic scattering reaching the detector (OS-13).

 

6. Explain in more detail the need to use the OS-13 light filter.

Answer:

The efficiency of elastic scattering of excitation wavelength (450 nm) is significantly higher compared to that for fluorescence emission. The color CMOS (or CCD) camera channels are rather spectroscopically wide (~100 nm) (see Fig. 3). Without OS-13 filter the fluorescence images will be saturated at blue channel while red and green channels (we actually need) will have low level of signal (red and green channel images will be noisy). Moreover, spectral wide of 450 nm excitation band is rather wide so “right” part of spectral wing (near 490-500 nm) will produce strong signal in green channel that is due to elastic scattering rather than fluorescence emission. With an OS-13 filter the backscattered 450 nm photons are suppressed so only fluorescence photons in “red” and “green” channels will be quantified providing highly contrast “red” and “green” channel images.

Added in the manuscript:

OS-13 glass filter is required since camera’s color channels had too wide spectral bands (see Fig. 3 below) and without such filter the “blue” channel will be saturated before “green” and “red” channels will get meaningful signals. Moreover, the “low energy” photons from the spectral wing of 450 nm excitation will produce signal in “green” channel originated by elastic scattering rather than by fluorescence.

 

7. The wrong caption to Figure 2: which "Transmission spectra of the LED lamp" are we talking about?

Answer:

Sorry, we were very harrying preparing the manuscript that resulted in some faulty statements and poor sentences. We have corrected the Fig. 2 capture according to the referee’s comment:

LED lamp spectrum and colored glass filters transmission spectra are presented in Supplementary materials A1.

 

8. It is necessary to separate Figure 3: separate photos and graphs. The photos are very small. Why combine (b) and (c) in Figure 4? Why duplicate (a) in Figure 4?

Answer:

We have modified the Fig. 3 and Fig. 4 and increased the photos 2-fold. In our opinion it is not convenient for the reader to separate photos and spectra in Fig. 3 and 4. All the components are marked by color and number so it is easy to compare the corresponding spectra in Fig. 3 (b) and Fig. 4 (b). If we extract the photos from Fig. 3 then it would be more difficult for the readers to switch attention between different figures. In the current form, it is easy to understand what particular spectrum corresponds to the component of interest (by matching the color or number).

Figure 3. Reflectance spectroscopy: (a) – photographs of rapeseed grist and individual maize silage components (1 – rapeseed grist pellets, 2 – maize grain, 3 – maize green leaves, 4 – maize stalk pulp, 5 – maize stalk leaves); (b) – reflectance spectra of rapeseed grist pellets and maize silage components; (c) – spectral sensitivity of red, green and blue channels of Canon 650D CMOS matrix according to the specification.

Figure 4. Fluorescence spectroscopy: (a) – photographs of rapeseed grist and individual maize silage components (1 – rapeseed grist pellets, 2 – maize grain, 3 – maize green leaves, 4 – maize stalk pulp, 5 – maize stalk leaves); (b) – fluorescence spectra of rapeseed grist pellets and maize silage components with the 450 nm excitation; (c) – spectral sensitivity of red, green and blue channels of Canon 650D CMOS matrix according to the specification.

Fig. 4 is the key figure of the manuscript. It is important to combined (b) and (c) in Fig. 4 since it clearly demonstrated that silage and grist components can be differed by fluorescence imaging with conventional color CMOS camera due to the fact that the green and red color channels fits the major fluorescence bands (530 and 680 nm) which are rather different for the components of interest.

We believe that it would be more convenient for the reader to see raw materials photos in the Fig. 4 (a) rather than check separated photos so losing the alertness.

 

9. Why is 450nm taken as the excitation wavelength?

Answer:

350-550 nm spectral range is a good choice for excitation of fluorescence spectra of plant materials. Here, we were interested in developing compact and economy instrument so we need a diode lamp with rather narrow spectral wide of excitation wavelength. Nowadays, 450 nm diodes are cheap and can provide few watts of power per single element. Such diodes are also rather economy due to mass scale production of grow light lamp for farmers and for home users. So, we choose the 450 nm excitation since it can provide enough power for sampling large areas (30 W power, 2500 cm2 area) with uniform distribution, but also had relatively narrow spectral wide (25 nm) of the excitation wavelength and low costs.

We have added discussion to the manuscript:

Nowadays, 450 nm diodes had rather narrow spectral wide, provided few watts of power and very cheap due to mass scale production of grow light lamp so it is a good choice for constructing low cost and robust instrument.

 

10. The authors' statement "The orange color filter absorbs in 400-500 nm spectral range (see Complementary materials A1), so there is almost zero intensity in fluorescence spectrum." is not clear.

Answer:

With this sentence we want to explain the absence of fluorescence signal in 400-450 nm spectral range. However, we made a typo. The orange color filter (OS-13) is a longpass colored glass filter with the edge at 460 nm (see Fig. A1 (c) in Supplementary materials) so no fluorescence can be detected in 400-450 nm range with our instrument. Still, we need to utilize OS-13 filter to suppress the elastically backscattered excitation preventing detector saturation (see above the discussion on 6^th referee’s question).

We rephrased the sentence to make it more clear for the reader:

Zero intensity of fluorescence spectrum in 400-450 nm range is due to orange color filter utilization (absorbs in 400-450 nm see Supplementary materials A1 (c) for OS-13 filter) that is needed to prevent spectrometer detector saturation by elastically backscattered excitation emission.

 

11. The authors' statement about the cheapness of the developed device should be accompanied by a comparison of its price with the modern analogues used.

Answer:

To the best of our knowledge, no instruments for online control of total mixed ration for cattle food production have been presented so far. Total mixed ration quality is important feature to get high efficient milk production in farms [Halachmi, I., et al. "Feeding behavior improves prediction of dairy cow voluntary feed intake but cannot serve as the sole indicator." Animal 10.9 (2016): 1501-1506.; Moallem, Uzi, and Liliya Lifshitz. "Accuracy and homogeneity of total mixed rations processed through trailer mixer or self-propelled mixer, and effects on the yields of high-yielding dairy cows." Animal Feed Science and Technology 270 (2020): 114708]. Nowadays, food preparation control it is carried out daily utilizing laboratory measurements with wet-chemistry methods or near infrared spectroscopy technique with chemometrics processing. Both mentioned approaches are very time and labor consuming so a new way for feed intake preparation is required. In other word we cannot compare our instrument with analogs due to absence of analogs. The closest concurrent techniques are near infrared spectroscopy (5000-25000 USD per instrument) and wet-chemistry methods (rather cheap chemicals but very expensive chemists labor time, online measurements are not possible).

We have added some discussion on this topic in Conclusions:

The alternative ways of cattle food mixture control (near infrared spectroscopy and wet-chemistry methods) cannot provide online measurements so far but costs 10-25 times greater compared to the developed smartphone based instrument.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

There is still a repeat of the photos of the silo components in Figures 3 and 4.

Author Response

Followed the referee’s comment we flushed the photos from Fig. 4 (a) and redraw the figure and change the figure caption accordingly.

Figure 4. Fluorescence spectroscopy: (a) – fluorescence spectra of rapeseed grist pellets and maize silage components with the 450 nm excitation (1 – rapeseed grist pellets, 2 – maize grain, 3 – maize green leaves, 4 – maize stalk pulp, 5 – maize stalk leaves; see photos in Fig.3 a); (b) – spectral sensitivity of red, green and blue channels of Canon 650D CMOS matrix according to the specification.

Author Response File: Author Response.docx