CN108990833A - A kind of animal movement behavior method of discrimination and device based on location information - Google Patents
A kind of animal movement behavior method of discrimination and device based on location information Download PDFInfo
- Publication number
- CN108990833A CN108990833A CN201811056886.6A CN201811056886A CN108990833A CN 108990833 A CN108990833 A CN 108990833A CN 201811056886 A CN201811056886 A CN 201811056886A CN 108990833 A CN108990833 A CN 108990833A
- Authority
- CN
- China
- Prior art keywords
- behavior
- animal
- location information
- influent
- standing
- 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
- 241001465754 Metazoa Species 0.000 title claims abstract description 65
- 230000033001 locomotion Effects 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 21
- 230000001133 acceleration Effects 0.000 claims abstract description 37
- 230000006399 behavior Effects 0.000 claims description 134
- 238000012549 training Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 16
- 238000005259 measurement Methods 0.000 claims description 14
- 238000012258 culturing Methods 0.000 claims description 13
- 238000005192 partition Methods 0.000 claims description 11
- 238000012937 correction Methods 0.000 claims description 9
- 239000000203 mixture Substances 0.000 claims description 8
- 230000004927 fusion Effects 0.000 claims description 7
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 238000003786 synthesis reaction Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 2
- 238000003860 storage Methods 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 abstract description 3
- 239000008267 milk Substances 0.000 description 51
- 210000004080 milk Anatomy 0.000 description 51
- 235000013336 milk Nutrition 0.000 description 51
- 241000283690 Bos taurus Species 0.000 description 50
- 230000000694 effects Effects 0.000 description 10
- 230000003542 behavioural effect Effects 0.000 description 8
- 238000012360 testing method Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000004807 localization Effects 0.000 description 3
- 238000007500 overflow downdraw method Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000000571 coke Substances 0.000 description 1
- 230000002301 combined effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000011017 operating method Methods 0.000 description 1
- 230000008560 physiological behavior Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01K—ANIMAL HUSBANDRY; AVICULTURE; APICULTURE; PISCICULTURE; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
- A01K29/00—Other apparatus for animal husbandry
- A01K29/005—Monitoring or measuring activity, e.g. detecting heat or mating
Landscapes
- Life Sciences & Earth Sciences (AREA)
- Environmental Sciences (AREA)
- Biophysics (AREA)
- Animal Husbandry (AREA)
- Biodiversity & Conservation Biology (AREA)
- Position Fixing By Use Of Radio Waves (AREA)
Abstract
The present invention relates to motor behavior identification technology field, especially a kind of animal movement behavior method of discrimination and device based on location information.By obtaining the 3-axis acceleration data of animal movement behavior to be processed, according to 3-axis acceleration data in motor behavior at least standing behavior and influent pH carry out identification classification;The location information of the animal at moment corresponding with 3-axis acceleration data is obtained, and location information is compared with animal feed area;The standing behavior that sorts out of identification and influent pH are differentiated to obtain true standing behavior and true influent pH according to D-S evidence theory algorithm, first classified according to 3-axis acceleration data to motor behavior, data and position data will be accelerated to combine and distinguish to confusing behavior again, the accuracy of behavior classification can be effectively improved.
Description
Technical field
The present invention relates to motor behavior identification technology field, especially a kind of animal movement behavior based on location information is sentenced
Other method and device.
Background technique
Currently, having become the strong tool of pattern-recognition due to the fast development of artificial neural network.Nerve net
The utilization of network expands new field, solves the problems, such as that other pattern-recognitions are indeterminable, classification feature is particularly suitable for mould
The application of formula identification and classification, BP neural network are a kind of most neural network forms of current application, can be applied to object
During motor behavior differentiates.BP neural network is applied in the motor behavior differentiation of object, because its error and function there may be office
Portion's minimum value, the disadvantages of resulting network fault tolerance ability is poor, learning rate is unstable cause to be applied to differentiate motor behavior
When, classifying quality is undesirable.Milk cow behavior is the important indicator of milk cow health and benefit level, and it is existing that milk cow behavior, which accurately differentiates,
For the important research content of animal husbandry.Existing milk cow behavior identification and classification rests on the level of motion state more, can not be effective
Other physiological behaviors by milk cow in motion state and simple motion state carry out the feed of classification processing, especially milk cow
Behavior and standing behavior lead to the motor behavior information inaccuracy identified, influence the judgement to milk cow behavior, and then to milk cow
Raising cause certain negative effect.
Summary of the invention
The object of the present invention is to provide a kind of animal movement behavior method of discrimination and device based on location information, to solve
Certainly existing animal movement behavior method of discrimination can not effectively classify influent pH in motion state with standing state, lead
Cause the problem of the behavioural information inaccuracy identified.
In order to realize accurately identifying for motor behavior, shape will can not effectively be moved by solving existing motor behavior method of discrimination
Influent pH is classified with standing state in state, leads to the problem of the behavioural information identified inaccuracy.The present invention provides one
Animal movement behavior method of discrimination of the kind based on location information, comprising the following steps:
1) the 3-axis acceleration data for obtaining animal movement behavior to be processed, according to 3-axis acceleration data to motor behavior
In at least standing behavior and influent pH carry out identification classification;
2) location information of the animal at moment corresponding with 3-axis acceleration data is obtained, and location information and animal are fed
Region is compared;
3) the standing behavior that sorts out of identification and influent pH are differentiated to obtain according to D-S evidence theory algorithm true
Standing behavior and true influent pH.
Beneficial effect is to be distinguished the feed in standing behavior with true standing behavior according to location information, first
Classified according to 3-axis acceleration data to motor behavior, then data and position data will be accelerated to combine to confusing row
To distinguish, the accuracy of behavior classification can be effectively improved.
Further, in order to solve the problems, such as that similar acceleration information feature is easily sorted out by mistake, feed is accurately identified
With standing, the D-S evidence theory algorithm is as follows:
(1) classification results and location information of standing behavior are as independent criterion and influent pH, standing behavior and not
Determine that behavior as judging result, establishes initial reliability table;
(2) feed area is provided with neck cangue, and the location information of animal includes the back leg position of animal, according to the back leg of animal
The distance of position and neck cangue constructs basic trust partition function to independent criterion respectively;
(3) evidence fusion is obtained as a result, reality to burnt member synthesis each in basic trust partition function according to D-S composition rule
Now true standing behavior and true influent pH are differentiated.
Further, in order to fast implement the corresponding behavior Classification and Identification to 3-axis acceleration data, BP network is made
For Weak Classifier, by the decision value for more classification results that the 3-axis acceleration data repetition training BP network of training set exports,
The strong classifier being made of again AdaBoost algorithm construction multiple BP network Weak Classifiers passes through described strong point in step 1)
Class device in motor behavior at least standing behavior and influent pH carry out identification classification.
Further, accurately location information in order to the foundation of subsequent D-S evidence theory is including in order to obtain
At least six wireless sensors are set on the boundary of the culturing area of the feed area, according to wireless sensor and setting dynamic
Wireless leg label on object back leg and after similarity algorithm is handled positioning obtain the back leg position of animal.
In order to be corrected to location information, make it in error range, further, also by reference to alignment sensor
Each measurement distance for obtaining reference location sensor and each wireless sensor, according to each measurement distance and the reality of corresponding calibration away from
From error correction parameter is obtained, it is corrected according to back leg position of the error correction parameter to animal.
The present invention provide a kind of animal movement behavior discriminating gear based on location information, including memory, processor with
And storage is in memory and the computer program that can run on a processor, when the processor executes described program realization with
Lower step:
1) the 3-axis acceleration data for obtaining animal movement behavior to be processed, according to 3-axis acceleration data to motor behavior
In at least standing behavior and influent pH carry out identification classification;
2) location information of the animal at moment corresponding with 3-axis acceleration data is obtained, and location information and animal are fed
Region is compared;
3) the standing behavior that sorts out of identification and influent pH are differentiated to obtain according to D-S evidence theory algorithm true
Standing behavior and true influent pH, first classified according to 3-axis acceleration data to motor behavior, then number will be accelerated
Confusing behavior is distinguished according to being combined with position data, the accuracy of behavior classification can be effectively improved.
Further, in order to solve the problems, such as that similar acceleration information feature is easily sorted out by mistake, feed is accurately identified
With standing, D-S evidence theory algorithm described in above-mentioned apparatus is as follows:
(1) classification results and location information of standing behavior are as independent criterion and influent pH, standing behavior and not
Determine that behavior as judging result, establishes initial reliability table;
(2) feed area is provided with neck cangue, and the location information of animal includes the back leg position of animal, according to the back leg of animal
The distance of position and neck cangue constructs basic trust partition function to independent criterion respectively;
(3) evidence fusion is obtained as a result, reality to burnt member synthesis each in basic trust partition function according to D-S composition rule
Now true standing behavior and true influent pH are differentiated.
Further, in order to fast implement the corresponding behavior Classification and Identification to 3-axis acceleration data, in above-mentioned apparatus
Using BP network as Weak Classifier, the more classification results exported by the 3-axis acceleration data repetition training BP network of training set
Decision value, then the strong classifier being made of AdaBoost algorithm construction multiple BP network Weak Classifiers leads in step 1)
Cross the strong classifier in motor behavior at least standing behavior and influent pH carry out identification classification.
Further, accurately location information in order to obtain, in order to the foundation of subsequent D-S evidence theory, above-mentioned apparatus
In at least six wireless sensors are set on the boundary of culturing area for including the feed area, according to wireless sensor
It positions with the wireless leg label being arranged on animal back leg and after similarity algorithm is handled and obtains the back leg position of animal.
In order to be corrected to location information, make it in error range, further, the device is also by reference to positioning
Sensor obtains each measurement distance of reference location sensor and each wireless sensor, according to each measurement distance and corresponds to calibration
Actual range obtains error correction parameter, is corrected according to back leg position of the error correction parameter to animal.
Detailed description of the invention
Fig. 1 is a kind of principle flow chart of animal movement behavior method of discrimination based on location information of the invention;
Fig. 2 is the setting schematic diagram of the fixed position of anchor node in milk cattle cultivating region of the invention;
Fig. 3 is the dummy grid figure that milk cattle cultivating region division is identical size in the present invention.
Specific embodiment
The present invention will be further described in detail with reference to the accompanying drawing.
The present invention provides a kind of animal movement behavior method of discrimination based on location information, and this method can be used for multiple types
The differentiation of the motor behavior of type animal, but it is all based on the exercise data of 3-axis acceleration sensor detection, the present invention is with milk cow
For motor behavior, 3-axis acceleration sensor is set in milk cow right rear leg, the feed of milk cow is obtained in real time, hurries up, slowly
Walk, lie low, standing, standing activities from lie down movement 7 kinds of different behaviors.
As shown in Figure 1, comprising the following steps:
1) obtain milk cow motor behavior to be processed 3-axis acceleration data and the moment corresponding with 3-axis acceleration data
The location information of animal.
It is fed, lain low, stood, the movement that lies down, standing activities, is careful, fastly by 3-axis acceleration sensor acquisition milk cow
The 3-axis acceleration data for walking 7 kinds of behavior right rear leg joints or less obtain the corresponding position of 7 kinds of behaviors of milk cow using positioning device
Confidence breath.Preferred orientation device is the wireless leg label being arranged on milk cow back leg.It is directly obtained and is believed by database in Fig. 1
Breath, but the information of database obtains through the above way.
It is positioned to position data by mobile node of wireless sensor network, obtains milk cow position with similarity algorithm
Information, the specific steps are as follows:
As shown in Fig. 2, 6 wireless sensors are evenly arranged around cowshed, that is, culturing area, i.e. 6 anchor sections in Fig. 2
Point, wherein anchor node1It is arranged to aggregation node, then collects information (position, ID sent by other each anchor nodes
Deng), wherein culturing area adjoins the feed area for facing that passageway side is milk cow close to neck cangue position.Sand bed is lain in Fig. 2 for milk cow
Bed rest region.
If cowshed is biserial open-standards cowshed, above-mentioned culturing area indicates biserial open-standards cowshed
The regions such as interior unilateral culturing area, including feed, rest, activity;Anchor node is arranged in up and down the two of the unilateral side culturing area
Side, the positioning for milk cow in the unilateral side culturing area.
As shown in figure 3, according to the distributed intelligence of all anchor nodes, anchor node1By the dummy grid of identical size to feeding
Region is grown to be divided.
Each virtual grid is made of four vertex and four sides, will remove the virtual grid vertex table on culturing area boundary
It is shown as Oj(j=1,2 ..., (S-1)2, the number of S expression virtual grid.
The distance of each wireless leg label to any anchor node is defined as di(1≤i≤6) form distance vector D
(d1,d2,...,d6)。
It is communicated between wireless leg label and anchor node, RSSI ranging data is collected, to constitute vector R (r1,
r2,...,r6), vector R is converted to by distance vector D (d using Lognormal shadowing model1,d2,...,d6), it is more preferable to obtain
Ranging effect.
Distance definition by the virtual grid vertex for removing culturing area boundary to anchor node is hi(1≤i≤6), h1,
h2,...,h6Constitute distance vector H.The approximate journey of wireless leg label with each virtual grid vertex is calculated using similarity function
Degree:
In formula (1), DiIt is the i-th dimension of distance vector D, HiIt is the i-th dimension of distance vector H, uiFor DiAnd HiIt is absolute average
Value.After the similarity for calculating the distance vector H on distance vector D and each vertex by formula (1), constituting corresponding array C is indicated
A certain wireless leg label and all virtual grid vertex degrees of closeness for removing culturing area boundary.Array C is indicated are as follows:
C=[E1,E2,...,E(S-1) 2] (2)
In formula (2), EiFor the similarity value of the distance vector H of the distance vector D and i-th of vertex that are calculated by formula (2).
Then, the coordinate of wireless leg label is obtained by the mass center on three vertex of calculating highest similarity value:
In formula (3), (X, Y) is the coordinate of wireless leg label, (Xj,Yj) it is the apex coordinate with highest similarity.?
In actual working environment, inevitable error is constantly present in the measurement of the distance between wireless leg label and anchor node.
In order to obtain ideal positioning accuracy, need to correct measurement distance by reference to alignment sensor.By comparing actual range and
Distance is measured, accurate error can be obtained.Error correction is as follows:
di'=di(1+δ)(5)
In formula (4) and (5), eiIt is actual distance of the reference location sensor to i-th of anchor node, fiIt is that reference location passes
Measurement distance of the sensor to the i-th anchor node.Actual range between reference location sensor and each anchor node, can pass through and be
Positional relationship after system deployment determines.Measurement distance is that the RSSI value between reference location sensor and each anchor node passes through
Lognormal shadowing model calculates acquisition.δ is error coefficient, reflects the measurement range accuracy of reference location sensor.Parameter
diIt is the measurement distance from a wireless leg label to i-th of anchor node.Parameter di' it is modified distance.
2) according to 3-axis acceleration data in motor behavior at least standing behavior and influent pH carry out identification classification.
It is preferred that according to 3-axis acceleration data by Multi-BP-AdaBoost algorithm at least standing in motor behavior
Behavior and influent pH carry out identification classification.The standing behavior and the identification of influent pH classification can also be existing by other
Mode classification is classified, for example, categorised decision tree-model.
Multi-BP-AdaBoost algorithm is using BP network as Weak Classifier, to the milk cow behavioral data with 7 classifications
Collection, repetition training BP network directly exports the decision value of more classification results, finally by the multiple BP nets of AdaBoost algorithm construction
The strong classifier of network Weak Classifier composition chooses classification corresponding to maximum decision value as affiliated milk cow behavior classification.
If one group of milk cow Behavioral training data set: T={ (p1,q1),…,(pN,qN), wherein pi It indicates
3-axis acceleration data, qiIndicate the determining respective behavior classification of manual video observation, P indicates adding for all milk cow behavior classifications
Speed training data, N indicate the sample size of training set.The realization process of algorithm are as follows:
(1) the total H of BP Weak Classifier and the weight ω of each training sample are initialized1i, whereinI=1 ...,
N。
(2) H Weak Classifier of training.
For m=1,2 ..., H, following steps are executed:
1, the training dataset for Distribution value of having the right is trained, obtains BP Weak Classifier: Gm(p): P → M=1,
2 ..., 7 }, wherein M indicates classified milk cow behavior classification.
2, G is calculatedm(p) error in classification rate,When the condition in bracket meets
When, otherwise it is 0 that I, which returns to 1,.
3, G is calculatedm(p) coefficient,
4, the distribution of training data centralized value is updated
(3) strong classifier is constructedInt is used for the integer part of return parameters.Strong classification
The decision value for the generic that device G (p) outputs test data, and according to the affiliated milk of size estimation test data of output decision value
Ox behavior classification.
3) location information is compared with animal feed area, and identification is sorted out according to D-S evidence theory algorithm
Standing behavior and influent pH differentiated to obtain true standing behavior and true influent pH.
D-S evidence theory algorithm is as follows:
(1) classification results and location information of standing behavior are as independent criterion and influent pH, standing behavior and not
Determine that behavior as judging result, establishes initial reliability table.
The location information that feed/standing behavior classification results and milk cow are arranged is 2 independent criterions, according to different criterions
Milk cow behavior judging result define identification framework Θ={ feed, stand, do not know }.Define belief assignment function mij(i=
1,2, j=1,2,3), andWherein i respectively corresponds 2 criterions, the corresponding j=1 of coke first " feed " in identification framework,
" standing " corresponds to j=2, " uncertain " corresponding j=3.
(2) feed area is provided with neck cangue, and the location information of animal includes the back leg position of animal, according to the back leg of animal
The distance of position and neck cangue constructs basic trust partition function to independent criterion respectively.
For 2 independent criterions, basic trust partition function is constructed respectively using interval method, as shown in table 1, a in table 1
Indicate the horizontal distance between milk cow back leg and neck cangue;B indicates that the average body of milk cow (not including head) is long;C indicates each milk cow
The maximum value of average localization error;D indicates the width in milk cattle cultivating region, and neck cangue position is as shown in Figure 2.Using leg fixation side
When formula disposes 3-axis acceleration sensor, easily obscure between influent pH and standing behavior, for feed/standing behavior classification knot
This criterion of fruit, the two probability can be considered as approximately equal, therefore behavior probability estimated by classification results is set as 0.5, another
The probability of kind behavior is set as 0.4, and probabilistic probability is set as 0.1.When the horizontal distance between milk cow back leg and neck cangue is close to L
When (the average body that milk cow does not include head is long), milk cow is most probably in fed conditions;Water between milk cow back leg and neck cangue
When flat distance is close to 0, milk cow is in parallel or back to neck cangue state, is more likely to as standing behavior, furthermore milk cow position is away from neck
Cangue is remoter, and the probability of standing behavior is bigger.
Table 1
(3) evidence fusion is obtained as a result, reality to burnt member synthesis each in basic trust partition function according to D-S composition rule
Now true standing behavior and true influent pH are differentiated.
According to D-S composition rule to each burnt member synthesis, the Mass function under 2 groups of evidence combined effects is derived.
In formula (6), m' is evidence fusion results, m'(feed), m'(stands), m'(it is uncertain) be respectively evidence fusion after
Feed is stood, uncertain probability value.K,k1、k2And k3For the intermediate variable of fusion process.m1(feed), m1(standing), m1
Feed that (uncertain) is determined by algorithm classification result, standing, uncertain probability value, m2(feed), m2(standing), m2(no
Determine) it is that the feed obtained, standing, uncertain probability value are calculated by milk cow position.Entire evidence is being obtained for identifying frame
In frame Θ after the probability assignments of each state, can be determined as the rule of formula (7) and formula (8) behavior state locating for milk cow (feed or
It stands).
In formula (7) and (8), ε1、ε2ε is generally set for previously given threshold value for the reliability of milk cow behavior classification1
Value be much larger than ε2Value, debugged in research repeatedly, ε be set1、ε2Respectively 0.2 and 0.03.Such as the item of formula (7) and formula (8)
Part is unable to satisfy, then milk cow behavior classification results are uncertain, and corresponding behavioral data is removed from milk cow Activity recognition.
Milk cow is fed in specific experiment, is stood, is lain low, is careful, is hurried up, standing activities, lie down 7 kinds of motor behaviors of movement
Acceleration and position data be acquired, every ox acquires 4h data daily.Because test during data packetloss and other because
Element influences, and test acquires 25921 groups of initial data altogether, wherein the motor behavior duration is more than that the valid data of 5s have 18030
Group data.The behavior conversion time for movement and the standing activities of lying down is usually more than 5s, can using the data that the duration is more than 4s
Effectively include the whole process of milk cow behavioral activity, guarantees the integrality and distinguishability of behavioral data.7 kinds of movements row of acquisition
It is constituted in detail for data as shown in table 2.
Table 2
It chooses 10818 datas in 13057 valid data (accounting for total 60% ratio of valid data) and is used as training dataset,
For constructing Binary decision tree-model;7212 datas (accounting for total 40% ratio of valid data) are used as test data set, to classification
The training of device model progress accelerometer measures (input) and respective behavior classification (output).
Milk cow behavior pattern classify as shown in table 3 using Multi-BP-AdaBoost algorithm, a is classification in table 3
Used in test sample sum;B is the behavior sum that Multi-BP-AdaBoost algorithm predicts each behavior pattern.
Classification results show to lie low, the movement that lies down, standing activities, be careful, the identification for behavior of hurrying up can reach desired water
It is flat.But there is also be mistaken for other behaviors, wherein feed and standing behavior are easy mutually to obscure (respectively 38% He
38.2%).In general, as shown in table 4, the overall performance of Multi-BP-AdaBoost algorithm is preferable.7 kinds of milk cow
The recognition accuracy of behavioural characteristic is higher, and in addition to feed and standing behavior, the sensitivity of other behaviors and precision are good.
Table 3
Table 4
The actual location data of milk cow be using existing image processing software obtain, by with based on similarity function
The position data that location algorithm obtains directly relatively realize.In a static condition, the positioning performance of system is than movement shape
Positioning performance under state is high.Table 5 shows the maximum value of five wireless leg labels and reference position sensor positioning performance, puts down
Mean value and minimum value.The position error of reference position sensor is significantly lower than the position error of wireless leg label.This is because
During test, the position coordinates of reference position sensor are to maintain constant.
After location algorithm of the application based on RSSI similarity, the average localization error of five wireless leg labels is 1.16
Rice, maximum average localization error is 1.30 meters, long lower than the average body of milk cow (not including head).The result of acquirement confirms position
Confidence breath can be further used for distinguishing feed and stand.This error will not influence the respective behavior for not needing altitude location precision
The analysis of index.In addition, reference position sensor has higher positioning accuracy, it can be effective by formula (4) and formula (5)
The range error of wireless leg label is modified, to guarantee positioning performance.
Table 5
Data sample is predicted by the data fusion method based on D-S evidence theory, and using this method to it
It is reanalysed, wherein source of evidence is the classification results as provided by Multi-BP-AdaBoost algorithm and milk cow position
Composition.Table 6 shows that classification results again, the recognition effect of two kinds of behaviors significantly improve, and a expression makes in reclassifying in table 6
Data sample sum;B indicates data fusion method prediction feed, standing and uncertain behavior based on D-S evidence theory
Behavior sum.It is this improved the result shows that, it is a kind of feasible operating method that evidence, which is associated,.But simultaneously as
It is unable to satisfy the condition of classification standard, has given up 77 data samples.Table 7 illustrates the statistic property of data fusion method.Two
20 percentage points and 18.5 percentage points have been respectively increased in the sensitivity of kind behavior and precision.Further, since in reclassifying
The ratio of true negative has declined in used data sample, and the accuracy fed and stood inevitably reduces.
Table 6
Table 7
Specific embodiment of the present invention is presented above, but the present invention is not limited to described embodiment.
Under the thinking that the present invention provides, to the skill in above-described embodiment by the way of being readily apparent that those skilled in the art
Art means are converted, are replaced, are modified, and play the role of with the present invention in relevant art means it is essentially identical, realize
Goal of the invention it is also essentially identical, the technical solution formed in this way is to be finely adjusted to be formed to above-described embodiment, this technology
Scheme is still fallen in protection scope of the present invention.
Claims (10)
1. a kind of animal movement behavior method of discrimination based on location information, which comprises the following steps:
1) the 3-axis acceleration data for obtaining animal movement behavior to be processed, according to 3-axis acceleration data in motor behavior
At least standing behavior and influent pH carries out identification classification;
2) location information of the animal at corresponding with 3-axis acceleration data moment is obtained, and by location information and animal feed area
It is compared;
3) the standing behavior that sorts out of identification and influent pH are differentiated according to D-S evidence theory algorithm and is really stood
Vertical behavior and true influent pH.
2. the animal movement behavior method of discrimination according to claim 1 based on location information, which is characterized in that the D-
S evidence theory algorithm is as follows:
(1) classification results and location information of standing behavior as independent criterion and influent pH, standing behavior and are not known
Behavior establishes initial reliability table as judging result;
(2) feed area is provided with neck cangue, and the location information of animal includes the back leg position of animal, according to the back leg position of animal
Basic trust partition function is constructed respectively to independent criterion with the distance of neck cangue;
(3) evidence fusion is obtained as a result, realizing true to burnt member synthesis each in basic trust partition function according to D-S composition rule
Real standing behavior and true influent pH is differentiated.
3. the animal movement behavior method of discrimination according to claim 2 based on location information, which is characterized in that by BP net
Network passes through the decision for more classification results that the 3-axis acceleration data repetition training BP network of training set exports as Weak Classifier
It is worth, then the strong classifier being made of AdaBoost algorithm construction multiple BP network Weak Classifiers, by described in step 1)
Strong classifier in motor behavior at least standing behavior and influent pH carry out identification classification.
4. the animal movement behavior method of discrimination according to claim 2 based on location information, which is characterized in that including
At least six wireless sensors are set on the boundary for having the culturing area of the feed area, are existed according to wireless sensor and setting
Wireless leg label on animal back leg and after similarity algorithm is handled positioning obtain the back leg position of animal.
5. the animal movement behavior method of discrimination according to claim 4 based on location information, which is characterized in that also pass through
Reference location sensor obtains each measurement distance of reference location sensor and each wireless sensor, according to each measurement distance and right
The actual range that should be demarcated obtains error correction parameter, carries out school according to back leg position of the error correction parameter to animal
Just.
6. a kind of animal movement behavior discriminating gear based on location information, including memory, processor and it is stored in storage
In device and the computer program that can run on a processor, which is characterized in that the processor realized when executing described program with
Lower step:
1) the 3-axis acceleration data for obtaining animal movement behavior to be processed, according to 3-axis acceleration data in motor behavior
At least standing behavior and influent pH carries out identification classification;
2) location information of the animal at corresponding with 3-axis acceleration data moment is obtained, and by location information and animal feed area
It is compared;
3) the standing behavior that sorts out of identification and influent pH are differentiated according to D-S evidence theory algorithm and is really stood
Vertical behavior and true influent pH.
7. the animal movement behavior discriminating gear according to claim 6 based on location information, which is characterized in that the D-
S evidence theory algorithm is as follows:
(1) classification results and location information of standing behavior as independent criterion and influent pH, standing behavior and are not known
Behavior establishes initial reliability table as judging result;
(2) feed area is provided with neck cangue, and the location information of animal includes the back leg position of animal, according to the back leg position of animal
Basic trust partition function is constructed respectively to independent criterion with the distance of neck cangue;
(3) evidence fusion is obtained as a result, realizing true to burnt member synthesis each in basic trust partition function according to D-S composition rule
Real standing behavior and true influent pH is differentiated.
8. the animal movement behavior discriminating gear according to claim 7 based on location information, which is characterized in that by BP net
Network passes through the decision for more classification results that the 3-axis acceleration data repetition training BP network of training set exports as Weak Classifier
It is worth, then the strong classifier being made of AdaBoost algorithm construction multiple BP network Weak Classifiers, by described in step 1)
Strong classifier in motor behavior at least standing behavior and influent pH carry out identification classification.
9. the animal movement behavior discriminating gear according to claim 7 based on location information, which is characterized in that including
At least six wireless sensors are set on the boundary for having the culturing area of the feed area, are existed according to wireless sensor and setting
Wireless leg label on animal back leg and after similarity algorithm is handled positioning obtain the back leg position of animal.
10. the animal movement behavior discriminating gear according to claim 9 based on location information, which is characterized in that also logical
Cross reference location sensor obtain reference location sensor and each wireless sensor each measurement distance, according to each measurement distance with
The actual range of corresponding calibration obtains error correction parameter, carries out school according to back leg position of the error correction parameter to animal
Just.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811056886.6A CN108990833A (en) | 2018-09-11 | 2018-09-11 | A kind of animal movement behavior method of discrimination and device based on location information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811056886.6A CN108990833A (en) | 2018-09-11 | 2018-09-11 | A kind of animal movement behavior method of discrimination and device based on location information |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108990833A true CN108990833A (en) | 2018-12-14 |
Family
ID=64590883
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811056886.6A Pending CN108990833A (en) | 2018-09-11 | 2018-09-11 | A kind of animal movement behavior method of discrimination and device based on location information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108990833A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114731967A (en) * | 2022-04-08 | 2022-07-12 | 内蒙古慧云科技有限公司 | Detection device and detection method for optimal hybridization time of sows |
CN115868428A (en) * | 2022-12-15 | 2023-03-31 | 四川农业大学 | Intelligent cattle necklace based on six-axis inertial sensor and monitoring method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070258649A1 (en) * | 2005-03-04 | 2007-11-08 | Bach Klaus J | Method for implementing n-dimensional object recognition using dynamic adaptive recognition layers |
CN103118333A (en) * | 2013-01-25 | 2013-05-22 | 河南科技大学 | Similarity based wireless sensor network mobile node positioning method |
CN105104291A (en) * | 2015-07-27 | 2015-12-02 | 河南科技大学 | Dairy cow motion state judging method and corresponding intelligent feeding method |
CN105307264A (en) * | 2015-07-27 | 2016-02-03 | 河南科技大学 | Mobile node positioning method for wireless sensor network |
-
2018
- 2018-09-11 CN CN201811056886.6A patent/CN108990833A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070258649A1 (en) * | 2005-03-04 | 2007-11-08 | Bach Klaus J | Method for implementing n-dimensional object recognition using dynamic adaptive recognition layers |
CN103118333A (en) * | 2013-01-25 | 2013-05-22 | 河南科技大学 | Similarity based wireless sensor network mobile node positioning method |
CN105104291A (en) * | 2015-07-27 | 2015-12-02 | 河南科技大学 | Dairy cow motion state judging method and corresponding intelligent feeding method |
CN105307264A (en) * | 2015-07-27 | 2016-02-03 | 河南科技大学 | Mobile node positioning method for wireless sensor network |
Non-Patent Citations (2)
Title |
---|
JORGE A. VÁZQUEZ DIOSDADO等: ""classification of behavior in housed dairy cows using an accelerometer-based activity monitoring system"", 《ANIMAL BIOTELEMETRY》 * |
JUN WANG: "Development and validation of an ensemble classifier for real-time recognition of cow behavior patterns from accelerometer data and location data", 《PLOS ONE》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114731967A (en) * | 2022-04-08 | 2022-07-12 | 内蒙古慧云科技有限公司 | Detection device and detection method for optimal hybridization time of sows |
CN114731967B (en) * | 2022-04-08 | 2023-04-25 | 内蒙古慧云科技有限公司 | Device and method for detecting optimal sow mating time |
CN115868428A (en) * | 2022-12-15 | 2023-03-31 | 四川农业大学 | Intelligent cattle necklace based on six-axis inertial sensor and monitoring method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bao et al. | Artificial intelligence in animal farming: A systematic literature review | |
Alvarenga et al. | Using a three-axis accelerometer to identify and classify sheep behaviour at pasture | |
Escalante et al. | Sow-activity classification from acceleration patterns: a machine learning approach | |
CN107256017B (en) | Route planning method and system | |
Xu et al. | Automatic scoring of postures in grouped pigs using depth image and CNN-SVM | |
CN109344872A (en) | A kind of recognition methods of national costume image | |
Yin et al. | Estrus detection in dairy cows from acceleration data using self-learning classification models. | |
CN107491734A (en) | Semi-supervised Classification of Polarimetric SAR Image method based on multi-core integration Yu space W ishart LapSVM | |
Tian et al. | Real-time behavioral recognition in dairy cows based on geomagnetism and acceleration information | |
Feiyang et al. | Monitoring behavior of poultry based on RFID radio frequency network | |
Lu et al. | An automatic splitting method for the adhesive piglets’ gray scale image based on the ellipse shape feature | |
CN112101333A (en) | Smart cattle farm monitoring and identifying method and device based on deep learning | |
CN108990833A (en) | A kind of animal movement behavior method of discrimination and device based on location information | |
CN109446931A (en) | Animal movement behavior method of discrimination and device based on sequential correlation analysis | |
CN110414369B (en) | Cow face training method and device | |
CN111199187A (en) | Animal behavior identification method based on algorithm, corresponding storage medium and electronic device | |
Liu et al. | Classification of cow behavior patterns using inertial measurement units and a fully convolutional network model | |
Zhang et al. | Early lameness detection in dairy cattle based on wearable gait analysis using semi-supervised LSTM-autoencoder | |
CN116543462A (en) | Method for identifying and judging dairy cow health condition based on dairy cow behaviors of video bones | |
Wang et al. | Development of a wireless measurement system for classifying cow behavior using accelerometer data and location data | |
Qin et al. | Identification of body size characteristic points based on the Mask R-CNN and correlation with body weight in Ujumqin sheep | |
Gao et al. | Adaptive cow movement detection using evolving spiking neural network models | |
Nigade et al. | Review Paper on IOT based Cattle Health Monitoring System | |
Yeşil et al. | Recognition of Hereford and Simmental cattle breeds via computer vision. | |
CN112801118B (en) | Pork pig marketing benefit evaluation system and method based on artificial intelligence and big data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181214 |
|
RJ01 | Rejection of invention patent application after publication |