CN102830072A - Identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy - Google Patents
Identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy Download PDFInfo
- Publication number
- CN102830072A CN102830072A CN2012102864555A CN201210286455A CN102830072A CN 102830072 A CN102830072 A CN 102830072A CN 2012102864555 A CN2012102864555 A CN 2012102864555A CN 201210286455 A CN201210286455 A CN 201210286455A CN 102830072 A CN102830072 A CN 102830072A
- Authority
- CN
- China
- Prior art keywords
- rice
- rice leaf
- wavelet
- neural network
- near infrared
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses an identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy. The method comprises the following steps: (1) collecting four kinds of rice leaf samples, emitting a near infrared spectrum with a wave number in a range of 4000 to 12000/cm to back sides of the rice leaf samples and collecting diffuse reflection spectrum information of all the rice leaf samples; (2) respectively processing the diffuse reflection spectrum information of all the rice leaf samples by using a wavelet transformation method so as obtain corresponding wavelet features; (3) establishing a neural network model with the wavelet features of the rice leaf samples as input and set values of pollution types corresponding to the rice leaf samples as output; and (4) substituting the wavelet features obtained after the step (1) and step (2) into the neural network model obtained in step (3) so as to obtain pollution types of rice to be identified. The identification method provided by the invention has the advantages of high precision, simple operation, low cost and capacity of realizing rapid and nondestructive identification of heavy metal pollution in rice.
Description
Technical field
The invention belongs to the heavy metal pollution detection range of plant, relate in particular to a kind of method of differentiating heavy metal pollution paddy rice blade based near infrared spectrum.
Background technology
Since the progress of industrial society, mining and industry " three wastes ", and using in a large number of chemical fertilizer and agricultural chemicals risen farmland soil heavy metals content, growth of serious threat plant and environmental quality.Paddy rice is China's important crops, and the security relationship of Rice Production is to national economy.
At present, the existing a large amount of rice fields of China receive the pollution of toxic heavy metal, and this not only influences the g and D of paddy rice, reduces yield and quality, and gets into human body through the heavy metal of paddy rice absorption by food chain, has a strong impact on human health.Lead in the heavy metal, cadmium, mercury are the leading indicators in the soil environment quality standard evaluation, also are the main heavy metal elements that paddy rice pollutes, and endanger bigger.Like cadmium, it is one of the strongest heavy metal of a kind of toxicity, compares with other non-essential element; Cadmium has stronger from the ability of soil to plant migration, and activity is stronger in environment, and therefore coefficient of concentration is bigger in green plants; Often under the situation that does not influence plant normal growth; Plant can accumulate the cadmium of higher concentration, gets into human body through food chain then, causes health problem.Mercury also is a kind of to human body and the bigger heavy metal of plant hazard, can show poor growth, symptom that leaf is withered and yellow after the general plant mercury poisoning, but not obvious usually, is difficult for discovering, and after the people eats mercury-contaminated crop by mistake, may cause mercury poisoning.Lead also is the non-essential element of plant, after it and plant contact, will produce certain toxic action to plant, and is light then the metabolic process in the plant is got muddled, and growing is suppressed, heavy then cause plant death.Therefore, based on pollution of heavy metals in rice be difficult for perceptibility and potential hazard property, differentiate that the heavy metal pollution paddy rice has great importance.
At present, the analytical approach that is used for heavy metal has: ultraviolet spectrophotometry, atomic absorption method, atomic fluorescence method, inductively coupled plasma method, X fluorescence spectrum, inductively coupled plasma mass spectroscopy etc.These method major parts are high to instrument requirement, and it is loaded down with trivial details to exist pretreatment operation, detect shortcomings such as cost height.And under a lot of situation, whether people only need to differentiate paddy rice fast contaminated by heavy metals, and need not to learn the concrete content value of heavy metal.
Summary of the invention
The invention provides a kind of method based near infrared spectrum discriminating heavy metal pollution paddy rice blade, this method identification precision is high, easy to operate.
A kind of method based near infrared spectrum discriminating heavy metal pollution paddy rice blade may further comprise the steps:
(1) gather four types of rice leaf samples, the emission wave-number range is 4000～12000cm to the rice leaf sample back side
-1Near infrared spectrum, and gather the spectrum information that diffuses of all rice leaf samples, in said four types of rice leaf samples, one type is normal rice leaf sample, its excess-three class is respectively mercury, cadmium, lead contamination rice leaf sample;
(2) utilize the wavelet conversion method respectively the spectrum information that diffuses of rice leaf sample to be handled, obtain corresponding wavelet character;
(3) wavelet character with the rice leaf sample is input, and the blade pollution type setting value corresponding with the rice leaf sample is output, sets up neural network model;
(4) obtain the wavelet character of paddy rice to be measured according to step (1)～(2), carry it into the neural network model described in the step (3), obtain the pollution type of paddy rice to be measured.
Heavy metal mainly is through changing macromolecular conformation, destroying the rice tissue structure to the influence of paddy growth, and is especially comparatively remarkable to the plant chloroplast effect on structure.Visible basal granule disintegration in the chloroplast of crosscut, thylakoid decreased number in the plant leaf blade of mercury processing back; The thylakoid lamella is irregular alignment; Marked inflation appears in granum-thylakoid and stroma-thylakoid lamella, also visible stroma-thylakoid sheet fault rupture, the little phenomenon of the loose expansion of granum-thylakoid; Chloroplast structure is more complete in the plant leaf blade of cadmium processing back, but torsional deformation, the fuzzy and slight swelling of thylakoid lamella; Plumbous handle chloroplast swelling in the plant leaf blade of back, chloroplast membranes is impaired, lamella distortion, phenomenon at random.
Because the variation of the blade construction component that the heavy metal stress such as variation of chloroplast structure cause can cause the difference of near-infrared absorption spectrum, near infrared spectrum capable of using differentiates that whether contaminated by heavy metals rice leaf is; And utilize the different of different heavy metals and chlorophyll, the different plant pathologies that cause of protein and other binding site, can realize that the kind of heavy metal is differentiated.
Rice leaf and normal rice leaf with the mercury that obtains, cadmium, lead contamination are modeling object, through obtaining its spectral information, set up neural network model, and the model after the domestication promptly can be used for differentiating heavy metal pollution paddy rice blade.
In the step (1), four types of rice leaf samples can obtain through following steps: mercurous 1.5mgKg is set respectively
-1, contain cadmium 1.0mgKg
-1, leaded 500mgKg
-1, do not contain four blocks of soil of heavy metal, rice cultivation on said four blocks of soil when paddy rice to five leaf during the phase, is chosen blade respectively as mercury pollution, cadmium pollution, lead contamination, normal rice leaf sample from four blocks of soil respectively.
To rice leaf sample back side emission near infrared spectrum, can better gather spectral information.
In the step (2); Wavelet transformation can be used for the noise reduction of spectrum, the compression of data; Wavelet transformation all has the ability of characterization signal in time domain and frequency field; And have the characteristics of multiresolution analysis, through selecting suitable wavelet function and decomposition level, utilize wavelet transformation can obtain comparatively comprehensively near infrared spectrum characteristic information of rice leaf.
The Daubechies small echo has advantages such as orthogonality, tight supportive; Through Daubechies small echo series (its exponent number is respectively 2,4,6,8,10,12,16) is carried out Analysis and Screening, female small echo that wavelet conversion adopts is preferably the female small echo (Db2) of 2 rank Daubechies.
The selection of wavelet decomposition level influences obtaining of spectral signature information, and decomposition level is high more, and high frequency noise is not grudged many more, and the spectral information amount of obtaining is few more, and the wavelet decomposition level is preferably 1～5, and more preferably 2 or 3.
In the step (3), said blade pollution type is mercury pollution, cadmium pollution, lead contamination and normal, in neural network model, can the output setting value of mercury pollution, cadmium pollution, lead contamination and normal four kinds of blade pollution types be decided to be 1,2,3,4 respectively.
Original spectrum obtains the wavelet character of different frequency and yardstick after wavelet conversion; Wavelet character quantity as input value is big; And concern relative complex between the output valve, and artificial neural network has self-learning capability, can improve analysis precision through study; Big at sample size, concern that complicated model more has superiority aspect setting up, that the while artificial neural network algorithm has is anti-interference, noise resisting ability and the strong advantage of non-linear conversion ability.
Said neural network model is preferably radial base neural net model or back transfer neural network model.
When mother wavelet function is the female small echos of 2 rank Daubechies, and the wavelet decomposition level is 3 o'clock, and said neural network model is preferably the radial base neural net model, and the radial base neural net model that set up this moment is good to the recognition effect of different heavy metal pollution blades.
When mother wavelet function is the female small echos of 2 rank Daubechies, and the wavelet decomposition level is 2 o'clock, and said neural network model is the back transfer neural network model, and the back transfer neural network model that set up this moment is good to the recognition effect of different heavy metal pollution blades.
With respect to prior art, beneficial effect of the present invention is:
(1) the present invention only need obtain the near infrared spectrum of rice leaf, can carry out the discriminating of pollution of heavy metals in rice, and is simple to operate, with low cost, can save time and the labour;
(2) high, the reliable results of identification precision of the present invention can realize quick, the Non-Destructive Testing of pollution of heavy metals in rice.
Description of drawings
Fig. 1 is the original spectrogram that diffuses of rice leaf sample among the embodiment 1;
The diffuse spectrogram of Fig. 2 after for the wavelet reconstruction of rice leaf sample among the embodiment 1;
Fig. 3 is the original spectrogram that diffuses of rice leaf sample among the embodiment 2;
The diffuse spectrogram of Fig. 4 after for the wavelet reconstruction of rice leaf sample among the embodiment 2;
Fig. 5 is the original spectrogram that diffuses of rice leaf sample among the embodiment 3;
The diffuse spectrogram of Fig. 6 after for the wavelet reconstruction of rice leaf sample among the embodiment 3;
Fig. 7 is the original spectrogram that diffuses of rice leaf sample among the embodiment 4;
The diffuse spectrogram of Fig. 8 after for the wavelet reconstruction of rice leaf sample among the embodiment 4.
Embodiment
Embodiment 1
1, sets up model
(1) mercurous 1.5mgKg is set respectively
-1, contain cadmium 1.0mgKg
-1, leaded 500mgKg
-1, do not contain four blocks of soil of heavy metal, rice cultivation on four blocks of soil; Spend 11 seeds during paddy rice is selected for use, seedling raising is after 15 days, and the rice transplanting seedling uses sufficient base manure in three blocks of soil that set content of beary metal, in time irrigates, and regularly applies nitrogenous fertilizer and composite fertilizer.
(2) when paddy rice to five leaf during the phase; From four blocks of soil, gather 20 rice leaf samples respectively; As calibration set, utilizing Nicolet Nexus870 (Thermo Corporation USA) Fourier transform near infrared spectrometer is 4000～12000cm to the back side of rice leaf sample emission wave-number range
-1Near infrared spectrum, and gather the spectrum information that diffuses of all rice leaf samples; Near infrared spectrometer is provided with scanning times 32 times, resolution 4cm
-1, room temperature is controlled at about 25 ℃ during the rice leaf sample collection, and it is stable that humidity keeps.
(3) utilize wavelet function Daubechies 2 (Db2), it is 3 that decomposition level is set, and the spectrum information that diffuses of all rice leaf samples of gathering is handled, and obtains 521 corresponding wavelet characters, and the wavelet character value of rice leaf sample is seen table 1; Diffuse spectrogram such as Fig. 1, shown in Figure 2 behind original the diffuse spectrogram and the wavelet reconstruction of all rice leaf samples are by knowing that the spectrogram after the processing has kept basic spectral information among the figure.
(4) wavelet character with the rice leaf sample of gained in the step (3) is input; The blade pollution type setting value corresponding with the rice leaf sample is output; The output setting value that mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types are set is respectively 1,2,3,4, sets up the RBF neural network model; The RBF of RBF network is the Gaussian function, and the input layer number is 521, and the hidden layer node number is 115, and output layer node number is 1; Target error is 1*10
-4, the expansion rate coefficient is 3, the data-switching mode is the standardization conversion, obtains the mapping relations like table 1.
Table 1 is used for the partial database of modelling
| NO | X 1 | X 2 | X 3 | X 4 | X 5 | X 6 | X 7 | X 8 | X 9 | X 10 | … | Y |
| 1 | 0.4491 | 0.4489 | 0.4485 | 0.4474 | 0.4460 | 0.4444 | 0.4426 | 0.4414 | 0.4396 | 0.4381 | … | 1 |
| 2 | 0.4469 | 0.4469 | 0.4470 | 0.4461 | 0.4444 | 0.4424 | 0.4409 | 0.4399 | 0.4371 | 0.4355 | … | 1 |
| 3 | 0.4480 | 0.4480 | 0.4480 | 0.4470 | 0.4451 | 0.4435 | 0.4417 | 0.4410 | 0.4382 | 0.4368 | … | 1 |
| 4 | 0.4489 | 0.4490 | 0.4496 | 0.4487 | 0.4468 | 0.4454 | 0.4434 | 0.4429 | 0.4411 | 0.4391 | … | 1 |
| 5 | 0.4503 | 0.4503 | 0.4506 | 0.4493 | 0.4488 | 0.4465 | 0.4451 | 0.4441 | 0.4425 | 0.4406 | … | 1 |
| 6 | 0.4829 | 0.4828 | 0.4819 | 0.4816 | 0.4812 | 0.4813 | 0.4803 | 0.4799 | 0.4793 | 0.4792 | … | 2 |
| 7 | 0.4785 | 0.4783 | 0.4784 | 0.4779 | 0.4767 | 0.4770 | 0.4758 | 0.4744 | 0.4735 | 0.4737 | … | 2 |
| 8 | 0.4736 | 0.4734 | 0.4726 | 0.4713 | 0.4709 | 0.4702 | 0.4688 | 0.4670 | 0.4659 | 0.4653 | … | 2 |
| 9 | 0.4707 | 0.4706 | 0.4702 | 0.4695 | 0.4684 | 0.4675 | 0.4657 | 0.4644 | 0.4628 | 0.4620 | … | 2 |
| 10 | 0.4652 | 0.4652 | 0.4655 | 0.4642 | 0.4635 | 0.4621 | 0.4604 | 0.4581 | 0.4567 | 0.4554 | … | 2 |
| 11 | 0.4617 | 0.4617 | 0.4606 | 0.4593 | 0.4581 | 0.4566 | 0.4550 | 0.4522 | 0.4502 | 0.4482 | … | 3 |
| 12 | 0.4629 | 0.4629 | 0.4616 | 0.4601 | 0.4590 | 0.4579 | 0.4561 | 0.4536 | 0.4520 | 0.4495 | … | 3 |
| 13 | 0.4690 | 0.4689 | 0.4682 | 0.4674 | 0.4665 | 0.4648 | 0.4641 | 0.4622 | 0.4607 | 0.4593 | … | 3 |
| 14 | 0.4697 | 0.4696 | 0.4686 | 0.4679 | 0.4670 | 0.4657 | 0.4649 | 0.4623 | 0.4611 | 0.4599 | … | 3 |
| 15 | 0.4693 | 0.4691 | 0.4689 | 0.4674 | 0.4671 | 0.4662 | 0.4650 | 0.4627 | 0.4616 | 0.4602 | … | 3 |
| 16 | 0.4857 | 0.4857 | 0.4864 | 0.4865 | 0.4858 | 0.4858 | 0.4855 | 0.4851 | 0.4847 | 0.4847 | … | 4 |
| 17 | 0.4871 | 0.4872 | 0.4882 | 0.4883 | 0.4880 | 0.4880 | 0.4873 | 0.4873 | 0.4866 | 0.4873 | … | 4 |
| 18 | 0.4819 | 0.4819 | 0.4821 | 0.4819 | 0.4812 | 0.4806 | 0.4800 | 0.4794 | 0.4790 | 0.4787 | … | 4 |
| 19 | 0.4746 | 0.4747 | 0.4754 | 0.4752 | 0.4742 | 0.4732 | 0.4720 | 0.4710 | 0.4700 | 0.4691 | … | 4 |
| 20 | 0.4706 | 0.4707 | 0.4713 | 0.4702 | 0.4687 | 0.4676 | 0.4668 | 0.4648 | 0.4636 | 0.4627 | … | 4 |
Wherein, NO is meant the sequence number of paddy rice sample, X
1, X
2, X
3, X
4Represent the wavelet character value; Y is an output valve.
2, utilize the pollution type of model prediction calibration set rice leaf sample
After accomplishing the domestication of RBF neural network model, with in 1 according to the wavelet character of rice leaf sample in the calibration set that step (2)～(3) obtain, carry it into and set up good RBF neural network model in 1, obtain output valve (as shown in table 2); Through output valve, confirm rice leaf sample contamination type according to following principle: when output valve greater than 0.5 less than 1.5 the time, be mercury pollution; When output valve greater than 1.5 less than 2.5 the time, be cadmium pollution; When output valve greater than 2.5 less than 3.5 the time, be lead contamination; When output valve greater than 3.5 less than 4.5 the time, for normally.
Rice leaf sample prediction pollution type and actual pollution type in table 2 calibration set
| NO | X 1 | X 2 | X 3 | X 4 | X 5 | X 6 | X 7 | X 8 | … | Y | S 1 | S 2 |
| 1 | 0.4491 | 0.4489 | 0.4485 | 0.4474 | 0.4460 | 0.4444 | 0.4426 | 0.4414 | … | 0.9925 | 1 | 1 |
| 2 | 0.4469 | 0.4469 | 0.4470 | 0.4461 | 0.4444 | 0.4424 | 0.4409 | 0.4399 | … | 1.0086 | 1 | 1 |
| 3 | 0.4480 | 0.4480 | 0.4480 | 0.4470 | 0.4451 | 0.4435 | 0.4417 | 0.4410 | … | 1.0030 | 1 | 1 |
| 4 | 0.4489 | 0.4490 | 0.4496 | 0.4487 | 0.4468 | 0.4454 | 0.4434 | 0.4429 | … | 1.0036 | 1 | 1 |
| 5 | 0.4503 | 0.4503 | 0.4506 | 0.4493 | 0.4488 | 0.4465 | 0.4451 | 0.4441 | … | 0.9915 | 1 | 1 |
| 6 | 0.4829 | 0.4828 | 0.4819 | 0.4816 | 0.4812 | 0.4813 | 0.4803 | 0.4799 | … | 1.9970 | 2 | 2 |
| 7 | 0.4785 | 0.4783 | 0.4784 | 0.4779 | 0.4767 | 0.4770 | 0.4758 | 0.4744 | … | 2.0030 | 2 | 2 |
| 8 | 0.4736 | 0.4734 | 0.4726 | 0.4713 | 0.4709 | 0.4702 | 0.4688 | 0.4670 | … | 1.9964 | 2 | 2 |
| 9 | 0.4707 | 0.4706 | 0.4702 | 0.4695 | 0.4684 | 0.4675 | 0.4657 | 0.4644 | … | 2.0001 | 2 | 2 |
| 10 | 0.4652 | 0.4652 | 0.4655 | 0.4642 | 0.4635 | 0.4621 | 0.4604 | 0.4581 | … | 2.0024 | 2 | 2 |
| 11 | 0.4617 | 0.4617 | 0.4606 | 0.4593 | 0.4581 | 0.4566 | 0.4550 | 0.4522 | … | 3.0044 | 3 | 3 |
| 12 | 0.4629 | 0.4629 | 0.4616 | 0.4601 | 0.4590 | 0.4579 | 0.4561 | 0.4536 | … | 2.9913 | 3 | 3 |
| 13 | 0.4690 | 0.4689 | 0.4682 | 0.4674 | 0.4665 | 0.4648 | 0.4641 | 0.4622 | … | 3.0027 | 3 | 3 |
| 14 | 0.4697 | 0.4696 | 0.4686 | 0.4679 | 0.4670 | 0.4657 | 0.4649 | 0.4623 | … | 3.0057 | 3 | 3 |
| 15 | 0.4693 | 0.4691 | 0.4689 | 0.4674 | 0.4671 | 0.4662 | 0.4650 | 0.4627 | … | 2.9918 | 3 | 3 |
| 16 | 0.4857 | 0.4857 | 0.4864 | 0.4865 | 0.4858 | 0.4858 | 0.4855 | 0.4851 | … | 3.9809 | 4 | 4 |
| 17 | 0.4871 | 0.4872 | 0.4882 | 0.4883 | 0.4880 | 0.4880 | 0.4873 | 0.4873 | … | 4.0051 | 4 | 4 |
| 18 | 0.4819 | 0.4819 | 0.4821 | 0.4819 | 0.4812 | 0.4806 | 0.4800 | 0.4794 | … | 4.0037 | 4 | 4 |
| 19 | 0.4746 | 0.4747 | 0.4754 | 0.4752 | 0.4742 | 0.4732 | 0.4720 | 0.4710 | … | 3.9920 | 4 | 4 |
| 20 | 0.4706 | 0.4707 | 0.4713 | 0.4702 | 0.4687 | 0.4676 | 0.4668 | 0.4648 | … | 4.0089 | 4 | 4 |
Wherein, NO is meant the sequence number of paddy rice sample, X
1, X
2, X
3, X
4Represent the wavelet character value; Y is an output valve, S
1Be model prediction paddy rice leaf samples pollution type, S
2Be actual pollution type.
Can know through data analysis: calibration set coefficient of determination R
2 cBe 0.9999, calibration set root-mean-square error RMSECV is 0.0089, calibration set correlated error REC is 0.0029, the RBF neural network model of foundation reaches 100% to the predictablity rate of calibration set paddy rice sample.
3, utilize the pollution type of model prediction checking collection rice leaf to be measured
89 rice leafs to be measured of random acquisition on the heavy-metal contaminated soil of step from 1 (1) configuration; As the checking collection; Obtain the wavelet character value (as shown in table 3) of all rice leafs to be measured based on step (2)～(3) in 1; And carry it into the good RBF neural network model of foundation, obtain model output valve Y; Through output valve Y, confirm rice leaf pollution type to be measured based on decision principle in 2.As space is limited, only 20 rice leaf data to be measured are listed in this, see table 3.
Rice leaf prediction heavy metal pollution type to be measured and actual pollution type are concentrated in table 3 checking
| NO | X 1 | X 2 | X 3 | X 4 | X 5 | X 6 | X 7 | X 8 | … | Y | S 1 | S 2 |
| 1 | 0.4460 | 0.4459 | 0.4452 | 0.4441 | 0.4427 | 0.4406 | 0.4388 | 0.4376 | … | 0.7924 | 1 | 1 |
| 2 | 0.4639 | 0.4639 | 0.4646 | 0.4642 | 0.4636 | 0.4628 | 0.4624 | 0.4625 | … | 1.2469 | 1 | 1 |
| 3 | 0.4484 | 0.4483 | 0.4482 | 0.4473 | 0.4454 | 0.4443 | 0.4423 | 0.4410 | … | 0.8271 | 1 | 1 |
| 4 | 0.4522 | 0.4524 | 0.4520 | 0.4517 | 0.4503 | 0.4494 | 0.4477 | 0.4464 | … | 0.4331 | 0 | 1 |
| 5 | 0.4546 | 0.4545 | 0.4556 | 0.4545 | 0.4531 | 0.4520 | 0.4507 | 0.4503 | … | 1.1707 | 1 | 1 |
| 6 | 0.4692 | 0.4690 | 0.4680 | 0.4668 | 0.4660 | 0.4646 | 0.4637 | 0.4615 | … | 1.9747 | 2 | 2 |
| 7 | 0.4615 | 0.4614 | 0.4608 | 0.4594 | 0.4577 | 0.4567 | 0.4548 | 0.4523 | … | 2.2003 | 2 | 2 |
| 8 | 0.4602 | 0.4600 | 0.4588 | 0.4577 | 0.4564 | 0.4550 | 0.4528 | 0.4502 | … | 2.0439 | 2 | 2 |
| 9 | 0.4584 | 0.4583 | 0.4569 | 0.4557 | 0.4541 | 0.4521 | 0.4500 | 0.4471 | … | 2.5038 | 0 | 2 |
| 10 | 0.4831 | 0.4830 | 0.4828 | 0.4824 | 0.4825 | 0.4829 | 0.4817 | 0.4813 | … | 1.9024 | 2 | 2 |
| 11 | 0.4727 | 0.4727 | 0.4726 | 0.4717 | 0.4715 | 0.4702 | 0.4692 | 0.4673 | … | 2.9721 | 3 | 3 |
| 12 | 0.4716 | 0.4714 | 0.4713 | 0.4698 | 0.4691 | 0.4685 | 0.4680 | 0.4658 | … | 3.1033 | 3 | 3 |
| 13 | 0.4724 | 0.4723 | 0.4722 | 0.4712 | 0.4704 | 0.4693 | 0.4683 | 0.4665 | … | 2.9816 | 3 | 3 |
| 14 | 0.4605 | 0.4604 | 0.4598 | 0.4582 | 0.4574 | 0.4557 | 0.4542 | 0.4516 | … | 3.0803 | 3 | 3 |
| 15 | 0.4708 | 0.4708 | 0.4709 | 0.4694 | 0.4695 | 0.4684 | 0.4670 | 0.4651 | … | 2.7845 | 3 | 3 |
| 16 | 0.4854 | 0.4854 | 0.4864 | 0.4865 | 0.4859 | 0.4860 | 0.4855 | 0.4851 | … | 3.9135 | 4 | 4 |
| 17 | 0.4669 | 0.4668 | 0.4670 | 0.4663 | 0.4653 | 0.4637 | 0.4625 | 0.4608 | … | 3.8701 | 4 | 4 |
| 18 | 0.4663 | 0.4664 | 0.4664 | 0.4661 | 0.4643 | 0.4635 | 0.4618 | 0.4602 | … | 3.9802 | 4 | 4 |
| 19 | 0.4643 | 0.4645 | 0.4643 | 0.4631 | 0.4615 | 0.4597 | 0.4583 | 0.4562 | … | 4.2497 | 4 | 4 |
| 20 | 0.4640 | 0.4640 | 0.4644 | 0.4634 | 0.4620 | 0.4599 | 0.4584 | 0.4566 | … | 4.3355 | 4 | 4 |
Wherein, NO is meant the sequence number of paddy rice sample, X
1, X
2, X
3, X
4Represent the wavelet character value; The Y output valve, S
1Be model prediction paddy rice leaf samples pollution type, S
2Be actual pollution type; S
1And S
2In: 1,2,3,4 represent mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types respectively, 0 is false judgment.
Can know through data analysis: checking collection coefficient of determination R
2 vBe 0.9460, verify that collection root-mean-square error RMSEP is 0.2587, verifies that collection correlated error REP is 0.1087, the RBF neural network model is respectively 95.5%, 81.8%, 91.3% and 100% to the predictablity rate of mercury, cadmium, lead contamination paddy rice and normal rice leaf.
Embodiment 2
1, the foundation of model
(1)～(2) with embodiment 1;
(3) utilize wavelet function Daubechies 2 (Db2), it is 2 that decomposition level is set, and the spectrogram that diffuses of all paddy rice samples of gathering is handled, and obtains 1039 corresponding wavelet characters, and the wavelet character value of paddy rice sample is seen table 4; Diffuse spectrogram such as Fig. 3, shown in Figure 4 behind original the diffuse spectrogram and the wavelet reconstruction of all paddy rice leaf samples can know that the spectrogram after the processing has kept basic spectral information from figure.
(4) wavelet character with the rice leaf sample of gained in the step (3) is input; The blade pollution type setting value corresponding with the rice leaf sample is output; The output setting value that mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types are set is respectively 1,2,3,4, sets up the BP neural network model; The input layer number of BP network is 1039, and the hidden layer node number is 100, and output layer node number is 1; The minimum training speed of setting network is 0.01, and the data-switching mode is the standardization conversion, and maximum iteration time is 1000 times, obtains the mapping relations like table 3.
Table 4 is used for the partial database of modelling
| NO | X 1 | X 2 | X 3 | X 4 | X 5 | X 6 | X 7 | X 8 | X 9 | X 10 | … | Y |
| 1 | 0.3175 | 0.3176 | 0.3165 | 0.3174 | 0.3166 | 0.3160 | 0.3163 | 0.3147 | 0.3150 | 0.3138 | … | 1 |
| 2 | 0.3159 | 0.3161 | 0.3159 | 0.3162 | 0.3153 | 0.3154 | 0.3149 | 0.3138 | 0.3134 | 0.3123 | … | 1 |
| 3 | 0.3168 | 0.3169 | 0.3164 | 0.3170 | 0.3163 | 0.3160 | 0.3150 | 0.3145 | 0.3140 | 0.3132 | … | 1 |
| 4 | 0.3174 | 0.3175 | 0.3179 | 0.3180 | 0.3172 | 0.3175 | 0.3160 | 0.3159 | 0.3154 | 0.3146 | … | 1 |
| 5 | 0.3184 | 0.3184 | 0.3186 | 0.3187 | 0.3179 | 0.3176 | 0.3174 | 0.3172 | 0.3165 | 0.3152 | … | 1 |
| 6 | 0.3414 | 0.3416 | 0.3404 | 0.3410 | 0.3406 | 0.3404 | 0.3410 | 0.3400 | 0.3401 | 0.3405 | … | 2 |
| 7 | 0.3384 | 0.3384 | 0.3376 | 0.3386 | 0.3381 | 0.3376 | 0.3380 | 0.3366 | 0.3373 | 0.3373 | … | 2 |
| 8 | 0.3349 | 0.3348 | 0.3340 | 0.3344 | 0.3332 | 0.3331 | 0.3339 | 0.3325 | 0.3326 | 0.3324 | … | 2 |
| 9 | 0.3328 | 0.3328 | 0.3325 | 0.3325 | 0.3321 | 0.3317 | 0.3320 | 0.3308 | 0.3307 | 0.3306 | … | 2 |
| 10 | 0.3289 | 0.3290 | 0.3292 | 0.3291 | 0.3285 | 0.3279 | 0.3285 | 0.3274 | 0.3267 | 0.3268 | … | 2 |
| 11 | 0.3265 | 0.3266 | 0.3257 | 0.3258 | 0.3249 | 0.3245 | 0.3247 | 0.3233 | 0.3236 | 0.3226 | … | 3 |
| 12 | 0.3273 | 0.3274 | 0.3264 | 0.3265 | 0.3256 | 0.3250 | 0.3252 | 0.3241 | 0.3244 | 0.3234 | … | 3 |
| 13 | 0.3316 | 0.3316 | 0.3312 | 0.3310 | 0.3308 | 0.3302 | 0.3304 | 0.3294 | 0.3296 | 0.3282 | … | 3 |
| 14 | 0.3321 | 0.3322 | 0.3311 | 0.3315 | 0.3308 | 0.3307 | 0.3309 | 0.3296 | 0.3303 | 0.3289 | … | 3 |
| 15 | 0.3318 | 0.3319 | 0.3310 | 0.3319 | 0.3308 | 0.3303 | 0.3305 | 0.3300 | 0.3304 | 0.3294 | … | 3 |
| 16 | 0.3434 | 0.3436 | 0.3435 | 0.3443 | 0.3437 | 0.3441 | 0.3441 | 0.3430 | 0.3438 | 0.3433 | … | 4 |
| 17 | 0.3444 | 0.3447 | 0.3445 | 0.3456 | 0.3452 | 0.3453 | 0.3453 | 0.3448 | 0.3453 | 0.3450 | … | 4 |
| 18 | 0.3407 | 0.3409 | 0.3404 | 0.3413 | 0.3406 | 0.3408 | 0.3405 | 0.3399 | 0.3406 | 0.3394 | … | 4 |
| 19 | 0.3355 | 0.3358 | 0.3357 | 0.3365 | 0.3357 | 0.3360 | 0.3358 | 0.3348 | 0.3353 | 0.3343 | … | 4 |
| 20 | 0.3327 | 0.3330 | 0.3329 | 0.3335 | 0.3325 | 0.3325 | 0.3318 | 0.3311 | 0.3312 | 0.3304 | … | 4 |
Wherein, NO is meant the sequence number of paddy rice sample, X
1, X
2, X
3, X
4Represent the wavelet character value; Y is an output valve.
2, utilize the pollution type of model prediction calibration set paddy rice leaf samples
After accomplishing the domestication of BP neural network model, with in 1 according to the wavelet character value of rice leaf sample in the calibration set that step (1)～(2) obtain, bring into and set up good BP neural network model, obtain output valve (as shown in table 5); Through output valve, confirm rice leaf sample contamination type according to decision principle among the embodiment 1.
Rice leaf sample prediction pollution type and actual pollution type in table 5 calibration set
| NO | X 1 | X 2 | X 3 | X 4 | X 5 | X 6 | X 7 | X 8 | … | Y | S 1 | S 2 |
| 1 | 0.3175 | 0.3176 | 0.3165 | 0.3174 | 0.3166 | 0.3160 | 0.3163 | 0.3147 | … | 0.9494 | 1 | 1 |
| 2 | 0.3159 | 0.3161 | 0.3159 | 0.3162 | 0.3153 | 0.3154 | 0.3149 | 0.3138 | … | 0.9651 | 1 | 1 |
| 3 | 0.3168 | 0.3169 | 0.3164 | 0.3170 | 0.3163 | 0.3160 | 0.3150 | 0.3145 | … | 0.9651 | 1 | 1 |
| 4 | 0.3174 | 0.3175 | 0.3179 | 0.3180 | 0.3172 | 0.3175 | 0.3160 | 0.3159 | … | 0.9521 | 1 | 1 |
| 5 | 0.3184 | 0.3184 | 0.3186 | 0.3187 | 0.3179 | 0.3176 | 0.3174 | 0.3172 | … | 0.9549 | 1 | 1 |
| 6 | 0.3414 | 0.3416 | 0.3404 | 0.3410 | 0.3406 | 0.3404 | 0.3410 | 0.3400 | … | 1.8721 | 2 | 2 |
| 7 | 0.3384 | 0.3384 | 0.3376 | 0.3386 | 0.3381 | 0.3376 | 0.3380 | 0.3366 | … | 1.9788 | 2 | 2 |
| 8 | 0.3349 | 0.3348 | 0.3340 | 0.3344 | 0.3332 | 0.3331 | 0.3339 | 0.3325 | … | 2.0295 | 2 | 2 |
| 9 | 0.3328 | 0.3328 | 0.3325 | 0.3325 | 0.3321 | 0.3317 | 0.3320 | 0.3308 | … | 1.9993 | 2 | 2 |
| 10 | 0.3289 | 0.3290 | 0.3292 | 0.3291 | 0.3285 | 0.3279 | 0.3285 | 0.3274 | … | 1.9919 | 2 | 2 |
| 11 | 0.3265 | 0.3266 | 0.3257 | 0.3258 | 0.3249 | 0.3245 | 0.3247 | 0.3233 | … | 2.7586 | 3 | 3 |
| 12 | 0.3273 | 0.3274 | 0.3264 | 0.3265 | 0.3256 | 0.3250 | 0.3252 | 0.3241 | … | 3.0769 | 3 | 3 |
| 13 | 0.3316 | 0.3316 | 0.3312 | 0.3310 | 0.3308 | 0.3302 | 0.3304 | 0.3294 | … | 3.0237 | 3 | 3 |
| 14 | 0.3321 | 0.3322 | 0.3311 | 0.3315 | 0.3308 | 0.3307 | 0.3309 | 0.3296 | … | 2.9666 | 3 | 3 |
| 15 | 0.3318 | 0.3319 | 0.3310 | 0.3319 | 0.3308 | 0.3303 | 0.3305 | 0.3300 | … | 3.4123 | 3 | 3 |
| 16 | 0.3434 | 0.3436 | 0.3435 | 0.3443 | 0.3437 | 0.3441 | 0.3441 | 0.3430 | … | 3.9442 | 4 | 4 |
| 17 | 0.3444 | 0.3447 | 0.3445 | 0.3456 | 0.3452 | 0.3453 | 0.3453 | 0.3448 | … | 3.9443 | 4 | 4 |
| 18 | 0.3407 | 0.3409 | 0.3404 | 0.3413 | 0.3406 | 0.3408 | 0.3405 | 0.3399 | … | 3.9441 | 4 | 4 |
| 19 | 0.3355 | 0.3358 | 0.3357 | 0.3365 | 0.3357 | 0.3360 | 0.3358 | 0.3348 | … | 4.2537 | 4 | 4 |
| 20 | 0.3327 | 0.3330 | 0.3329 | 0.3335 | 0.3325 | 0.3325 | 0.3318 | 0.3311 | … | 2.6975 | 0 | 4 |
Wherein, NO is meant the sequence number of paddy rice sample, X
1, X
2, X
3, X
4Represent the wavelet character value; Y is an output valve, S
1Be model prediction paddy rice leaf samples pollution type, S
2Be actual pollution type; S
1And S
2In: 1,2,3,4 represent mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types respectively, 0 is false judgment.
Can know through data analysis: calibration set coefficient of determination R
2 cBe 0.9695, calibration set root-mean-square error RMSECV is 0.1954, calibration set correlated error REC is 0.0562, the BP neural network model of foundation reaches 95.8% to the predictablity rate of calibration set paddy rice sample.
3, utilize the pollution type of model prediction checking collection rice leaf to be measured
89 rice leafs to be measured of random acquisition on the heavy-metal contaminated soil of step from 1 (1) configuration; As the checking collection; Obtain the wavelet character value (as shown in table 6) of all rice leafs to be measured based on step (1)～(2) in 1; And carry it into the good BP neural network model of foundation, obtain the model output valve; Through output valve, confirm rice leaf pollution type to be measured based on decision principle among the embodiment 1.As space is limited, only wherein 20 rice leaf data to be measured are listed in this, see table 6.
Rice leaf prediction heavy metal pollution type to be measured is concentrated in table 6 checking
| NO | X 1 | X 2 | X 3 | X 4 | X 5 | X 6 | X 7 | X 8 | … | Y | S 1 | S 2 |
| 1 | 0.3153 | 0.3153 | 0.3148 | 0.3147 | 0.3145 | 0.3137 | 0.3136 | 0.3127 | … | 0.9560 | 1 | 1 |
| 2 | 0.3280 | 0.3281 | 0.3283 | 0.3285 | 0.3286 | 0.3279 | 0.3282 | 0.3276 | … | 1.2105 | 1 | 1 |
| 3 | 0.3170 | 0.3170 | 0.3168 | 0.3169 | 0.3166 | 0.3160 | 0.3154 | 0.3145 | … | 0.9550 | 1 | 1 |
| 4 | 0.3198 | 0.3198 | 0.3201 | 0.3194 | 0.3194 | 0.3193 | 0.3191 | 0.3182 | … | 0.9473 | 1 | 1 |
| 5 | 0.3214 | 0.3214 | 0.3217 | 0.3223 | 0.3216 | 0.3211 | 0.3210 | 0.3199 | … | 0.9461 | 1 | 1 |
| 6 | 0.3318 | 0.3318 | 0.3305 | 0.3310 | 0.3304 | 0.3296 | 0.3306 | 0.3289 | … | 2.6573 | 0 | 2 |
| 7 | 0.3263 | 0.3263 | 0.3256 | 0.3260 | 0.3251 | 0.3244 | 0.3250 | 0.3230 | … | 2.0107 | 2 | 2 |
| 8 | 0.3254 | 0.3254 | 0.3242 | 0.3245 | 0.3238 | 0.3233 | 0.3236 | 0.3224 | … | 1.9000 | 2 | 2 |
| 9 | 0.3241 | 0.3241 | 0.3233 | 0.3230 | 0.3221 | 0.3220 | 0.3223 | 0.3205 | … | 1.9235 | 2 | 2 |
| 10 | 0.3417 | 0.3416 | 0.3411 | 0.3416 | 0.3412 | 0.3408 | 0.3419 | 0.3409 | … | 1.8858 | 2 | 2 |
| 11 | 0.3342 | 0.3343 | 0.3340 | 0.3344 | 0.3333 | 0.3334 | 0.3340 | 0.3329 | … | 2.0107 | 2 | 3 |
| 12 | 0.3335 | 0.3334 | 0.3327 | 0.3334 | 0.3328 | 0.3316 | 0.3325 | 0.3312 | … | 2.1499 | 2 | 3 |
| 13 | 0.3340 | 0.3341 | 0.3337 | 0.3340 | 0.3334 | 0.3327 | 0.3335 | 0.3318 | … | 3.3455 | 3 | 3 |
| 14 | 0.3257 | 0.3255 | 0.3252 | 0.3249 | 0.3247 | 0.3234 | 0.3242 | 0.3229 | … | 3.0775 | 3 | 3 |
| 15 | 0.3329 | 0.3329 | 0.3329 | 0.3330 | 0.3322 | 0.3315 | 0.3327 | 0.3314 | … | 2.7612 | 3 | 3 |
| 16 | 0.3431 | 0.3435 | 0.3433 | 0.3443 | 0.3440 | 0.3440 | 0.3440 | 0.3433 | … | 3.9438 | 4 | 4 |
| 17 | 0.3301 | 0.3302 | 0.3297 | 0.3306 | 0.3296 | 0.3297 | 0.3296 | 0.3286 | … | 3.7082 | 4 | 4 |
| 18 | 0.3297 | 0.3299 | 0.3296 | 0.3301 | 0.3291 | 0.3297 | 0.3289 | 0.3278 | … | 3.3673 | 0 | 4 |
| 19 | 0.3282 | 0.3286 | 0.3283 | 0.3285 | 0.3274 | 0.3275 | 0.3268 | 0.3259 | … | 3.1124 | 0 | 4 |
| 20 | 0.3280 | 0.3284 | 0.3278 | 0.3288 | 0.3277 | 0.3276 | 0.3269 | 0.3263 | … | 3.4766 | 0 | 4 |
Wherein, NO is meant the sequence number of paddy rice sample, X
1, X
2, X
3, X
4Represent the wavelet character value; Y is an output valve, S
1Be model prediction paddy rice leaf samples pollution type, S
2Be actual pollution type; S
1And S
2In: 1,2,3,4 represent mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types respectively, 0 is false judgment.
Can know that through data analysis the predictablity rate of BP neural network model is 83.1%, checking collection coefficient of determination R
2 vBe 0.8752, verify that collection root-mean-square error RMSEP is 0.3932, verifies that collection correlated error REP is 0.1164.
Embodiment 3
1, sets up model
(1)～(2) step is with embodiment 1;
(3) utilize wavelet function Daubechies 2 (Db2), it is 4 that decomposition level is set, and the spectrum information that diffuses of all rice leaf samples of gathering is handled, and obtains 262 corresponding wavelet characters; Diffuse spectrogram such as Fig. 5, shown in Figure 6 behind original the diffuse spectrogram and the wavelet reconstruction of all rice leaf samples are by knowing that the spectrogram after the processing has kept basic spectral information among the figure;
(4) wavelet character with the rice leaf sample of gained in the step (3) is input; The blade pollution type setting value corresponding with the rice leaf sample is output; The output setting value that mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types are set is respectively 1,2,3,4, sets up the RBF neural network model.
2, utilize the pollution type of model prediction calibration set paddy rice leaf samples
After accomplishing the domestication of RBF neural network model, with in 1 according to the wavelet character value of paddy rice leaf samples in the calibration set that step (1)～(2) obtain, bring into and set up good RBF neural network model, obtain output valve; Through output valve, confirm paddy rice leaf samples pollution type according to decision principle among the embodiment 1.
Can know through data analysis: calibration set coefficient of determination R
2 cBe 0.9999, calibration set root-mean-square error RMSECV is 0.0090, calibration set correlated error REC is 0.0035, the RBF neural network model of foundation reaches 100% to the predictablity rate of calibration set paddy rice sample.
3, utilize the pollution type of model prediction checking collection rice leaf to be measured
89 rice leafs to be measured of random acquisition on the heavy-metal contaminated soil of step from 1 (1) configuration; As the checking collection; Obtain the wavelet character value of all rice leafs to be measured based on step (1)～(2) in 1, and carry it into and set up good RBF neural network model, obtain model output valve Y; Through output valve Y, confirm rice leaf pollution type to be measured based on decision principle among the embodiment 1.
Can know through data analysis: checking collection coefficient of determination R
2 vBe 0.9228, verify that collection root-mean-square error RMSEP is 0.3092, verifies that collection correlated error REP is 0.1258, the RBF neural network model is respectively 95.5%, 72.7%, 82.6% and 90.9% to the predictablity rate of mercury, cadmium, lead contamination paddy rice and normal rice leaf.
Embodiment 4
1, sets up model
(1)～(2) step is with embodiment 1;
(3) utilize wavelet function Daubechies 2 (Db2), it is 5 that decomposition level is set, and the spectrogram that diffuses of all rice leaf samples of gathering is handled, and obtains 132 corresponding wavelet characters; Diffuse spectrogram such as Fig. 7, shown in Figure 8 behind original the diffuse spectrogram and the wavelet reconstruction of all rice leaf samples are by knowing that the spectrogram after the processing has kept basic spectral information among the figure;
(4) wavelet character with the rice leaf sample of gained in the step (3) is input; The blade pollution type setting value corresponding with the rice leaf sample is output; The output setting value that mercury pollution, cadmium pollution, lead contamination and normal four kinds of pollution types are set is respectively 1,2,3,4, sets up the RBF neural network model.
2, utilize the pollution type of model prediction calibration set paddy rice leaf samples
After accomplishing the domestication of RBF neural network model, with in 1 according to the wavelet character value of rice leaf sample in the calibration set that step (1)～(2) obtain, bring into and set up good RBF neural network model, obtain output valve; Through output valve, confirm rice leaf sample contamination type according to decision principle among the embodiment 1.
Can know through data analysis: calibration set coefficient of determination R
2 cBe 0.9999, calibration set root-mean-square error RMSECV is 0.0091, calibration set correlated error REC is 0.0032, the RBF neural network model of foundation reaches 100% to the predictablity rate of calibration set rice leaf sample.
3, utilize the pollution type of model prediction checking collection rice leaf to be measured
89 rice leafs to be measured of random acquisition on the heavy-metal contaminated soil of step from 1 (1) configuration; As the checking collection; Obtain the wavelet character value of all rice leafs to be measured based on step (1)～(2) in 1, and carry it into and set up good RBF neural network model, obtain model output valve Y; Through output valve Y, confirm rice leaf pollution type to be measured based on decision principle among the embodiment 1.
Can know through data analysis: checking collection coefficient of determination R
2 vBe 0.8987, verify that collection root-mean-square error RMSEP is 0.3542, verifies that collection correlated error REP is 0.1502, the RBF neural network model is respectively 72.7%, 86.4%, 100% and 86.4% to the predictablity rate of mercury, cadmium, lead contamination rice leaf and normal rice leaf.
Claims (7)
1. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade is characterized in that, may further comprise the steps:
(1) gather four types of rice leaf samples, the emission wave-number range is 4000～12000cm to the rice leaf sample back side
-1Near infrared spectrum, and gather the spectrum information that diffuses of all rice leaf samples, in said four types of rice leaf samples, one type is normal rice leaf sample, its excess-three class is respectively mercury, cadmium, lead contamination rice leaf sample;
(2) utilize the wavelet conversion method respectively the spectrum information that diffuses of rice leaf sample to be handled, obtain corresponding wavelet character;
(3) wavelet character with the rice leaf sample is input, and the blade pollution type setting value corresponding with the rice leaf sample is output, sets up neural network model;
(4) obtain the wavelet character of paddy rice to be measured according to step (1)～(2), carry it into the neural network model described in the step (3), obtain the pollution type of paddy rice to be measured.
2. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade as claimed in claim 1 is characterized in that in the step (2), female small echo that wavelet conversion adopts is preferably the female small echo of 2 rank Daubechies.
3. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade as claimed in claim 1 is characterized in that in the step (2), the wavelet decomposition level that wavelet conversion adopts is 1～5.
4. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade as claimed in claim 3 is characterized in that in the step (2), the wavelet decomposition level that wavelet conversion adopts is 2 or 3.
5. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade as claimed in claim 1 is characterized in that in the step (3), said neural network model is radial base neural net model or back transfer neural network model.
6. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade as claimed in claim 2 is characterized in that in the step (3), the wavelet decomposition level that wavelet conversion adopts is at 3 o'clock, and said neural network model is the radial base neural net model.
7. the method based near infrared spectrum discriminating heavy metal pollution paddy rice blade as claimed in claim 2 is characterized in that in the step (3), the wavelet decomposition level that wavelet conversion adopts is at 2 o'clock, and said neural network model is the back transfer neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012102864555A CN102830072A (en) | 2012-08-13 | 2012-08-13 | Identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012102864555A CN102830072A (en) | 2012-08-13 | 2012-08-13 | Identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102830072A true CN102830072A (en) | 2012-12-19 |
Family
ID=47333284
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012102864555A Pending CN102830072A (en) | 2012-08-13 | 2012-08-13 | Identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102830072A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104359882A (en) * | 2014-11-12 | 2015-02-18 | 江南大学 | Method for simultaneously measuring hybrid pigment by synchronous fluorescence spectroscopy with RBF (Radial Basis Function) neural network |
| CN104374731A (en) * | 2014-11-18 | 2015-02-25 | 湖南省食品测试分析中心 | MIR spectral analysis model and method for identifying excessive cadmium content of rice. |
| CN105092522A (en) * | 2015-08-13 | 2015-11-25 | 浙江大学 | Determination method for chromium content in cotton-seed meal |
| CN105738306A (en) * | 2016-04-13 | 2016-07-06 | 天津大学 | Portable fire control emergency rescue multi-gas rapid remote sensor and detecting method |
| CN106872383A (en) * | 2017-03-31 | 2017-06-20 | 南京农业大学 | A kind of paddy rice Reflectance position extracting method based on continuous wavelet analysis |
| CN109916845A (en) * | 2019-01-18 | 2019-06-21 | 武汉大学 | Rice Cd stress intensity diagnostic blade clamp device based on near-infrared specific wavelength |
| CN109961035A (en) * | 2019-03-19 | 2019-07-02 | 广东省农业科学院水稻研究所 | A kind of rice plant of tillering stage growing way wisdom monitoring adjusting method |
| CN110907387A (en) * | 2019-11-22 | 2020-03-24 | 南开大学 | Near-infrared rapid detection method for content of camellia saponin in camellia chrysantha leaves |
| CN112748085A (en) * | 2020-12-22 | 2021-05-04 | 湖南省水稻研究所 | Method for establishing near-infrared model for predicting cadmium content in rice and method for predicting cadmium content in rice |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004212392A (en) * | 2002-12-17 | 2004-07-29 | Okayama Prefecture | Surface inspection method |
| CN101520412A (en) * | 2009-03-23 | 2009-09-02 | 中国计量学院 | Near infrared spectrum analyzing method based on isolated component analysis and genetic neural network |
| US20090281738A1 (en) * | 2006-06-14 | 2009-11-12 | Bridgestone Corporation | Tire inspection device |
| CN101706443A (en) * | 2009-03-27 | 2010-05-12 | 上海工程技术大学 | Smoothness evaluation method of seams of clothing fabrics |
| CN101738383A (en) * | 2008-11-07 | 2010-06-16 | 中国农业科学院农业环境与可持续发展研究所 | Mid-infrared spectrum-based method for quickly detecting contents of heavy metal elements in soil |
| CN102507516A (en) * | 2011-09-28 | 2012-06-20 | 江南大学 | Method for detecting food pigment by combination of fluorescence spectroscopy and artificial neural network |
-
2012
- 2012-08-13 CN CN2012102864555A patent/CN102830072A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004212392A (en) * | 2002-12-17 | 2004-07-29 | Okayama Prefecture | Surface inspection method |
| US20090281738A1 (en) * | 2006-06-14 | 2009-11-12 | Bridgestone Corporation | Tire inspection device |
| CN101738383A (en) * | 2008-11-07 | 2010-06-16 | 中国农业科学院农业环境与可持续发展研究所 | Mid-infrared spectrum-based method for quickly detecting contents of heavy metal elements in soil |
| CN101520412A (en) * | 2009-03-23 | 2009-09-02 | 中国计量学院 | Near infrared spectrum analyzing method based on isolated component analysis and genetic neural network |
| CN101706443A (en) * | 2009-03-27 | 2010-05-12 | 上海工程技术大学 | Smoothness evaluation method of seams of clothing fabrics |
| CN102507516A (en) * | 2011-09-28 | 2012-06-20 | 江南大学 | Method for detecting food pigment by combination of fluorescence spectroscopy and artificial neural network |
Non-Patent Citations (8)
| Title |
|---|
| YONG HE ET AL: "Discrimination of varieties of tea using near infrared spectroscopy by principal component analysis and BP model", 《JOURNAL OF FOOD ENGINEERING》, vol. 79, no. 4, 30 April 2007 (2007-04-30) * |
| 修丽娜 等: "镉污染水稻高光谱诊断分析与建模", 《光谱学与光谱分析》, vol. 31, no. 1, 15 December 2011 (2011-12-15) * |
| 刘美玲: "融合多元环境因子的水稻重金属污染水平遥感评估模型", 《中国博士学位论文全文数据库》, no. 8, 15 August 2011 (2011-08-15) * |
| 单杨 等: "近红外光谱结合小波变换-径向基神经网络用于奶粉蛋白质与脂肪含量的测定", 《红外与毫米波学报》, vol. 29, no. 2, 15 April 2010 (2010-04-15) * |
| 姚霞 等: "水稻叶片色素含量近红外光谱估测模型研究", 《分析化学》, vol. 40, no. 4, 15 April 2012 (2012-04-15), pages 589 - 591 * |
| 方美红 等: "小波分析用于水稻叶片氮含量高光谱反演", 《应用科学学报》, vol. 28, no. 4, 30 July 2010 (2010-07-30) * |
| 王平: "农作物重金属污染胁迫遥感弱信息增强与计算", 《中国博士学位论文全文数据库》, no. 11, 15 November 2010 (2010-11-15) * |
| 陈志德 等: "不同水稻品种在Cd、As和Hg胁迫下吸收积累特性", 《中国农学通报》, vol. 24, no. 2, 15 February 2008 (2008-02-15) * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104359882A (en) * | 2014-11-12 | 2015-02-18 | 江南大学 | Method for simultaneously measuring hybrid pigment by synchronous fluorescence spectroscopy with RBF (Radial Basis Function) neural network |
| CN104374731A (en) * | 2014-11-18 | 2015-02-25 | 湖南省食品测试分析中心 | MIR spectral analysis model and method for identifying excessive cadmium content of rice. |
| CN105092522A (en) * | 2015-08-13 | 2015-11-25 | 浙江大学 | Determination method for chromium content in cotton-seed meal |
| CN105738306A (en) * | 2016-04-13 | 2016-07-06 | 天津大学 | Portable fire control emergency rescue multi-gas rapid remote sensor and detecting method |
| CN106872383A (en) * | 2017-03-31 | 2017-06-20 | 南京农业大学 | A kind of paddy rice Reflectance position extracting method based on continuous wavelet analysis |
| CN106872383B (en) * | 2017-03-31 | 2019-05-24 | 南京农业大学 | A kind of rice Reflectance position extracting method based on continuous wavelet analysis |
| CN109916845A (en) * | 2019-01-18 | 2019-06-21 | 武汉大学 | Rice Cd stress intensity diagnostic blade clamp device based on near-infrared specific wavelength |
| CN109961035A (en) * | 2019-03-19 | 2019-07-02 | 广东省农业科学院水稻研究所 | A kind of rice plant of tillering stage growing way wisdom monitoring adjusting method |
| CN110907387A (en) * | 2019-11-22 | 2020-03-24 | 南开大学 | Near-infrared rapid detection method for content of camellia saponin in camellia chrysantha leaves |
| CN112748085A (en) * | 2020-12-22 | 2021-05-04 | 湖南省水稻研究所 | Method for establishing near-infrared model for predicting cadmium content in rice and method for predicting cadmium content in rice |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102830072A (en) | Identification method for rice leaves contaminated by heavy metals based on near infrared spectroscopy | |
| Yitbarek et al. | Impacts of land use on selected physicochemical properties of soils of Abobo area, western Ethiopia | |
| Yang et al. | Growth of soybean seedlings in relay strip intercropping systems in relation to light quantity and red: far-red ratio | |
| Shepherd et al. | Infrared spectroscopy—enabling an evidence-based diagnostic surveillance approach to agricultural and environmental management in developing countries | |
| Zhang et al. | Converting paddy fields to Lei bamboo (Phyllostachys praecox) stands affected soil nutrient concentrations, labile organic carbon pools, and organic carbon chemical compositions | |
| Wöhling et al. | Bayesian model averaging to explore the worth of data for soil‐plant model selection and prediction | |
| Durdevic et al. | Spatial variability of soil organic matter content in Eastern Croatia assessed using different interpolation methods | |
| Hou et al. | Comprehensive assessment of fertilization, spatial variability of soil chemical properties, and relationships among nutrients, apple yield and orchard age: A case study in Luochuan County, China | |
| Hagemann et al. | Moss and lichen decomposition in old-growth and harvested high-boreal forests estimated using the litterbag and minicontainer methods | |
| Zhao et al. | Relationships of leaf nitrogen concentration and canopy nitrogen density with spectral features parameters and narrow-band spectral indices calculated from field winter wheat (Triticum aestivum L.) spectra | |
| Rad et al. | The distribution of ecological species groups in Fagetum communities of Caspian forests: determination of effective environmental factors | |
| CN109214635A (en) | A kind of evaluation method of compost maturity | |
| Idowu et al. | Soil quality of semi-arid conservation reserve program lands in Eastern New Mexico | |
| CN105608296B (en) | A kind of blade potassium concn inversion method based on lichee canopy spectra | |
| Behnke et al. | Agronomic assessment of cover cropping and tillage practices across environments | |
| Wang et al. | Estimating cadmium concentration in the edible part of Capsicum annuum using hyperspectral models | |
| Brown et al. | Topsoil characteristics from a paired farm survey of organic versus conventional farming in southern England | |
| Reddy et al. | Assessing climate suitability for sustainable vegetable Roselle (Hibiscus sabdariffa var. sabdariffa L.) cultivation in India using MaxEnt model | |
| Luo et al. | Foliar nutrient diagnosis norms for litchi (Litchi chinensis Sonn. cv. Feizixiao) in South China | |
| Millner et al. | Foliar nutrients in Eucalyptus species in New Zealand | |
| Sheppard et al. | Variation among chlorine concentration ratios for native and agronomic plants | |
| Li et al. | Spatial variation of winegrape yield and berry composition and their relationships to spatiotemporal distribution of soil water content | |
| Hayman et al. | Erratum to: The impact of El Niño Southern Oscillation on seasonal drought in the southern Australian grainbelt | |
| Valeeva et al. | Monitoring of urban environmental quality using chlorophyll delayed fluorescence. | |
| Parsons et al. | Genetic variation for adaptive traits in bottlebrush squirreltail in the northern Intermountain West, United States |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20121219 |