CN103090946A - Method and system for measuring single fruit tree yield - Google Patents

Method and system for measuring single fruit tree yield Download PDF

Info

Publication number
CN103090946A
CN103090946A CN2013100146865A CN201310014686A CN103090946A CN 103090946 A CN103090946 A CN 103090946A CN 2013100146865 A CN2013100146865 A CN 2013100146865A CN 201310014686 A CN201310014686 A CN 201310014686A CN 103090946 A CN103090946 A CN 103090946A
Authority
CN
China
Prior art keywords
fruit
point cloud
dimensional point
tree
radius
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.)
Granted
Application number
CN2013100146865A
Other languages
Chinese (zh)
Other versions
CN103090946B (en
Inventor
赵春江
陆声链
郭新宇
王传宇
温维亮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Research Center for Information Technology in Agriculture
Original Assignee
Beijing Research Center for Information Technology in Agriculture
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Research Center for Information Technology in Agriculture filed Critical Beijing Research Center for Information Technology in Agriculture
Priority to CN201310014686.5A priority Critical patent/CN103090946B/en
Publication of CN103090946A publication Critical patent/CN103090946A/en
Application granted granted Critical
Publication of CN103090946B publication Critical patent/CN103090946B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention discloses a method and a system for measuring a single fruit tree yield. Three-dimensional point cloud data of a fruit tree are obtained and are pre-processed to obtain a three-dimensional point cloud data set of the fruit tree; the three-dimensional point cloud data set of the fruit tree is divided to obtain a fruit three-dimensional point cloud data set which only comprises fruit information; statistics is carried on the fruit three-dimensional point cloud data set to obtain N fruit three-dimensional point cloud data subsets and the number of fruits N; the fruit radius of each corresponding fruit is calculated according to the fruit three-dimensional point cloud data subsets to obtain a fruit radius digit group; fruit weight corresponding to each fruit radius in the fruit radius digit group is calculated one by one according to basic parameters of a relation model of fruit radius and fruit weight, and the yield of the whole fruit tree is calculated through accumulation. The method and the system for measuring the single fruit tree yield are capable of measuring the single fruit tree yield accurately, quickly and harmlessly.

Description

The method and system of the single tree of fruit tree yield monitoring
Technical field
The present invention relates to output of the fruit tree intelligent measure field, relate in particular to the method and system of the single tree of fruit tree yield monitoring.
Background technology
The measurement of output of the fruit tree all has important application demand in Agriculture production and management and agronomy research.The harmless major part of accurately measuring of output of the fruit tree is the method that adopts complicate statistics fruit number and estimate the single tree of fruit tree total production traditionally, this method wastes time and energy, and because fruit in fruit tree is often more, be difficult to remember which fruit is to add up, so the method for complicate statistics fruit number is easy to cause larger error.
In recent years, along with reaching its maturity of the technology such as the fast development of infotech, particularly machine vision, sonic detection, laser measurement, for the intelligent measure of output of the fruit tree provides new approach.
Designed a real-time oranges and tangerines system for measuring yield based on machine vision in " estimating in real time the citrus production colorized vision system " literary composition, this system by one with the robot of digital camera in fruit tree limit walking, and the digital picture of acquisition fruit tree canopy, by utilizing fruit and leaf and the difference of other background informations on color in digital picture, oranges and tangerines are separated from digital picture, then oranges and tangerines are identified number, and according to each oranges and tangerines the weight of these oranges and tangerines of magnitude estimation in image, realize on this basis the measurement statistics of whole orchard output of the fruit tree.
Adopt digital picture to carry out the measurement of wild blueberry output in " estimating the wild blueberry fruit yield based on color digital image " literary composition, the method utilizes digital camera to take a digital pictures directly over blueberry plant colony, then utilizing in image fruit and the difference of other partial informations on color to carry out background rejects, make the pixel information that only keeps fruit in digital picture, and realize the output of captured blueberry colony by the quantity of adding up these pixels.
similarly, utilize machine vision technique to carry out the nondestructive measurement of mandarin tree output in " surveying the product method based on the oranges and tangerines that image is processed " literary composition, the method is at first by the digital pictures from an angle shot individual plant oranges and tangerines fruit tree, then based on different other the oranges and tangerines Image Segmentation Usings of RGB color model to obtaining of predefined mandarin tree, extract the overall circumference of fruit number and each fruit from cut apart the citrusfruit image that obtains, the characteristic parameters such as the fruit total area, calculate at last the total production of fruit tree, thereby realized the measurement of individual plant output of the fruit tree.
" based on the apple garden recovery prediction of image processing " utilizes Digital image technology, by obtain the digital picture in orchard florescence in fruit tree, fruit tree by the prior foundation relation between density and output of the fruit tree of blooming is simultaneously extracted the bloom fruit tree total production in density prediction orchard of the fruit tree that obtains from the digital picture of obtaining.
In existing output of the fruit tree measuring method based on machine vision, only a side from fruit tree obtains the fruit tree digital picture that comprises fruit information, because the fruit tree canopy is with luxuriant foliage and spreading branches in leafy profusion, serious shielding, therefore no matter from which side photographic images, also can't comprise all fruits on fruit tree in image, this will cause finally having larger error by the fruit quantity of extracting in captured digital picture with real quantity, and cause the larger error of fruit yield that calculates from fruit quantity.Simultaneously, all can there are differences owing to taking from different perspectives the fruit that comprises in the fruit tree digital picture obtain, there are the characteristics that measurement result is not unique, have manual operation randomness in the method for therefore only measuring output of the fruit tree by the method for taking an image.In addition, this method calculates the parameters such as the girth of fruit and radius by detecting the fruit pixel size obtain in captured digital picture, all have the problem of distortion due to the image of digital camera shooting, this can cause that also the output of the fruit tree that finally calculates produces another kind of error.
In output of the fruit tree evaluation method described in " based on the apple garden recovery prediction of image processing ", it is the relation by bloom density and the actual measurement output of the fruit tree of prior foundation, again by obtaining the digital picture that fruit tree blooms and calculate the density of blooming, thereby derive the possible output of these fruit trees.This method is not analyzed by the digital picture of directly obtaining the fruit tree fruiting period, the error of result of calculation may be larger, because the actual output of fruit tree also can be subject to the impact of multiple external condition, comprise disease and pest, rain, snow etc., the existing density of blooming of fruit tree might not be brought expected yield.Therefore this method relatively is suitable for the output estimation is carried out in large-scale orchard, and is not suitable for the measurement of individual plant output of the fruit tree.So said method all can't be accomplished the output of individual plant fruit tree is measured accurately.
Summary of the invention
The technical matters that (one) will solve
How accurately and fast, nondestructively the output of individual plant fruit tree is measured the technical problem to be solved in the present invention is, for defects.
(2) technical scheme
For addressing the above problem, the invention provides the method and system of the single tree of fruit tree yield monitoring, described method comprises:
A: obtain the three dimensional point cloud of fruit tree, and described three dimensional point cloud is carried out pre-service obtain fruit tree three dimensional point cloud collection;
B: described fruit tree three dimensional point cloud collection is cut apart the fruit three dimensional point cloud collection that obtains only comprising fruit information;
C: described fruit three dimensional point cloud collection is added up, obtained N fruit three dimensional point cloud subset and fruit number N, N is positive integer;
D: calculate the fruit radius of each corresponding fruit according to described fruit three dimensional point cloud subset, obtain fruit radius array;
E: calculate one by one fruit quality corresponding to each fruit radius in fruit radius array according to the underlying parameter in the relational model of fruit radius and fruit quality, accumulation calculating goes out the output of whole strain fruit tree;
Also comprise before described step e:
S: measure fruit radius and the fruit quality of sample fruit, set up the relational model of fruit radius and fruit quality, obtain described underlying parameter.
Preferably, described steps A specifically comprises:
A1: the fruit tree to band fruit state obtains the three dimensional point cloud of different angles, and forms fruit tree three-dimensional point cloud raw data set, comprises colouring information in wherein said three dimensional point cloud;
A2: the noise spot that described fruit tree three-dimensional point cloud raw data is concentrated is tentatively rejected;
A3: the noise spot of again the fruit tree three-dimensional point cloud raw data after preliminary rejecting being concentrated carries out secondary rejects, and obtains only comprising the fruit tree three dimensional point cloud collection of fruit tree information.
Preferably, described step B specifically comprises:
B1: the color characteristic that obtains data point corresponding to fruit organ;
B2: according to the color value of each concentrated data point of described fruit tree three dimensional point cloud and the distance of described color characteristic, be partitioned into fruit organ data point in addition, obtain described fruit three dimensional point cloud collection.
Preferably, in described step C, described fruit three dimensional point cloud collection is added up and be specially: described fruit three dimensional point cloud collection is carried out cluster, the data point that will belong to same fruit is divided in a fruit three dimensional point cloud subset, and the subset number that obtains is exactly the fruit number.
Preferably, described step D specifically comprises:
D1: calculate the length and width parameter in each fruit three dimensional point cloud subset;
D2: the center point coordinate that calculates each fruit three dimensional point cloud subset;
D3: calculate all data points in each fruit three dimensional point cloud subset and the distance of described center point coordinate, and calculate mean distance;
D4: calculate fruit radius corresponding to each fruit three dimensional point cloud subset according to described length and width parameter and described mean distance, all fruit radiuses form described fruit radius array.
For addressing the above problem, the present invention also provides the system of the single tree of fruit tree yield monitoring, and described system comprises:
Pretreatment module, three dimensional point cloud are cut apart module, fruit counting module, fruit radius calculation module and output of the fruit tree computing module;
Described pretreatment module is obtained the three dimensional point cloud of fruit tree, and described three dimensional point cloud is carried out pre-service obtains fruit tree three dimensional point cloud collection;
Described three dimensional point cloud is cut apart module, described fruit tree three dimensional point cloud collection is cut apart the fruit three dimensional point cloud collection that obtains only comprising fruit information;
Described fruit counting module is added up described fruit three dimensional point cloud collection, obtains the fruit number;
Described fruit radius calculation module calculates the fruit radius of each corresponding fruit according to described fruit three dimensional point cloud subset, obtain fruit radius array;
Described output of the fruit tree computing module calculates fruit quality corresponding to each fruit radius in fruit radius array one by one according to the underlying parameter in the relational model of fruit radius and fruit quality, and accumulation calculating goes out the output of whole strain fruit tree;
Described system also comprises: the underlying parameter module, measure fruit radius and the fruit quality of sample fruit, and set up the relational model of fruit radius and fruit quality, obtain described underlying parameter.
Preferably, described pretreatment module comprises: the raw data set acquisition module, tentatively reject module and secondary and put forward the rejecting module;
Described raw data set acquisition module obtains the three dimensional point cloud of different angles to the fruit tree of being with the fruit state, and forms fruit tree three-dimensional point cloud raw data set, comprises colouring information in wherein said three dimensional point cloud;
Described preliminary rejecting module is tentatively rejected the noise spot that described fruit tree three-dimensional point cloud raw data is concentrated;
Described secondary is put forward the rejecting module, and the noise spot that the fruit tree three-dimensional point cloud raw data after preliminary rejecting is concentrated carries out the secondary rejecting, obtains only comprising the fruit tree three dimensional point cloud collection of fruit tree information.
Preferably, described three dimensional point cloud is cut apart module and is comprised: color characteristic acquisition module and fruit three dimensional point cloud collection acquisition module;
Described color characteristic acquisition module obtains the color characteristic of data point corresponding to fruit organ;
Described fruit three dimensional point cloud collection acquisition module according to the color value of each concentrated data point of described fruit tree three dimensional point cloud and the distance of described color characteristic, is partitioned into fruit organ data point in addition, obtains described fruit three dimensional point cloud collection.
Preferably, described fruit radius calculation module comprises: length and width acquisition module, center point coordinate computing module, mean distance computing module and fruit radius array acquisition module;
Described length and width acquisition module calculates the length and width parameter in each fruit three dimensional point cloud subset;
Described center point coordinate computing module calculates the center point coordinate of each fruit three dimensional point cloud subset;
Described mean distance computing module calculates all data points in each fruit three dimensional point cloud subset and the distance of described center point coordinate, and calculates mean distance;
Described fruit radius array acquisition module calculates fruit radius corresponding to each fruit three dimensional point cloud subset according to described length and width parameter and described mean distance, and all fruit radiuses form described fruit radius array.
(3) beneficial effect
The present invention proposes the method and system of the single tree of fruit tree yield monitoring, by obtaining the three dimensional point cloud of fruit tree, and carry out pre-service and obtain fruit tree three dimensional point cloud collection; Utilize fruit to obtain fruit three dimensional point cloud collection cut apart different from the color characteristic of other organs, in the employing method of birdsing of the same feather flock together, fruit three dimensional point cloud collection is cut apart, guarantee treatment effeciency and the accuracy of fruit counting; Calculate the fruit radius of each corresponding fruit according to fruit three dimensional point cloud subset, obtain fruit radius array; Calculate one by one fruit quality corresponding to each fruit radius in fruit radius array according to the underlying parameter in the relational model of fruit radius and fruit quality, accumulation calculating goes out the output of whole strain fruit tree, can also calculate the weight of fruit mean diameter and average each fruit, can be accurately and fast, nondestructively the single tree of fruit tree output is measured.
Description of drawings
Fig. 1 is the flow chart of steps of the method for the single tree of fruit tree yield monitoring;
Fig. 2 is the particular flow sheet of steps A in the method for the single tree of fruit tree yield monitoring;
Fig. 3 is the particular flow sheet of step B in the method for the single tree of fruit tree yield monitoring;
Fig. 4 is the particular flow sheet of step D in the method for the single tree of fruit tree yield monitoring;
Fig. 5 is the composition schematic diagram of the system of the single tree of fruit tree yield monitoring.
Embodiment
Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.Following examples are used for explanation the present invention, but are not used for limiting the scope of the invention.
Embodiment one
The method of the single tree of fruit tree yield monitoring is provided in the embodiment of the present invention one, and steps flow chart specifically comprises the following steps as shown in Figure 1:
Steps A: obtain the three dimensional point cloud of fruit tree, and three dimensional point cloud is carried out pre-service obtain fruit tree three dimensional point cloud collection D.
Utilize the large laser spatial digitizer directly to obtain the three dimensional point cloud of fruit tree from the field, and the three dimensional point cloud that obtains is carried out the pre-service such as noise points deleting, steps flow chart specifically comprises the following steps as shown in Figure 2:
Steps A 1: the fruit tree to band fruit state obtains the three dimensional point cloud of different angles, and forms fruit tree three-dimensional point cloud raw data set, wherein comprises colouring information in three dimensional point cloud.
The achievement phase fruit tree that needs is carried out yield monitoring in the orchard utilizes laser 3 d scanner (for example FARO focus3D120) to carry out 3-D scanning, at least carry out multistation scanning from 3 angles of fruit tree during scanning, collect the fruit tree three-dimensional point cloud raw data set with colouring information.But utilize three-dimensional laser scanner under the environment of orchard, fruit tree to be carried out in the three dimensional point cloud acquisition process, due to blocking between different fruit trees and between organ, and the impact of other external conditions (as wind), the three dimensional point cloud that obtains inevitably can with noise spot, therefore need to carry out the noise spot in following two steps and remove operation.
Steps A 2: the noise spot that fruit tree three-dimensional point cloud raw data is concentrated is tentatively rejected, concrete grammar is as follows: at first adopt minimum neighbours' method to carry out the preliminary rejecting of noise spot, each data point of namely fruit tree three-dimensional point cloud raw data being concentrated is carried out minimum neighbours and is checked, if the neighbours of certain data point (data point that is not more than 0.5cm apart from the air line distance of this point is the neighbours of this point) number is about to this data point and concentrates deletion from the three-dimensional point cloud raw data less than 30.
Steps A 3: the noise spot of again the fruit tree three-dimensional point cloud raw data after preliminary rejecting being concentrated carries out secondary rejects, and obtains only comprising the fruit tree three dimensional point cloud collection D of fruit tree information.The noise spot that fruit tree three dimensional point cloud after preliminary rejecting is concentrated carries out quadratic noise rejecting processing, use minimum nearest neighbour method can't reject and don't belong to the data point of fruit tree its data in order to remove those, method is to utilize general three-dimensional point cloud process software to import the above fruit tree three-dimensional point cloud raw data set that obtains, and does not belong to the data point of the organs such as fruit tree leaf, fruit, limb by those in the method rejecting three-dimensional point cloud of choosing alternately data point and deletion.Noise points deleting through above two steps is processed, and concentrates the noise spot that does not belong to fruit tree partly to remove the three-dimensional point cloud raw data of fruit tree, obtains only comprising the fruit tree three dimensional point cloud collection D of fruit tree information.
Step B: fruit tree three dimensional point cloud collection D is cut apart the fruit three dimensional point cloud collection D that obtains only comprising fruit information f, this step is mainly previous step to be processed the fruit tree three dimensional point cloud collection D that obtains cut apart, and the data point that does not wherein belong to fruit is rejected, specific implementation process specifically comprises the following steps as shown in Figure 3:
Step B1: the color characteristic that obtains data point corresponding to fruit organ.Color characteristic according to the data point that in fruit tree three dimensional point cloud collection D Calculation of Three Dimensional cloud data, the fruit organ is corresponding, concrete grammar is: at first choose 50 of data point corresponding to fruit organ by interactive mode from fruit tree three dimensional point cloud collection D, then calculate the color average of these 50 data points.The color of supposing certain data point i in 50 selected data points is c i(r, g, b), wherein the i value is the natural number between 1 to 50, the computing formula of the color characteristic of fruit organ is:
OC ( r , g , b ) = ( 1 50 Σ i = 1 50 c ir , 1 50 Σ i = 1 50 c ig , 1 50 Σ i = 1 50 c ib )
C wherein irBe the r component value of the color of i data point, c igBe the g component value of the color of i data point, c ibIt is the b component value of the color of i data point.
Step B2: according to the color value of each data point in fruit tree three dimensional point cloud collection D and color characteristic apart from d f, be partitioned into fruit organ data point in addition, obtain fruit three dimensional point cloud collection D fTo each the data point p in the three dimensional point cloud collection D that obtains in steps A 3, calculate respectively the color c of this data point pThe color characteristic of (r, g, b) and fruit apart from d f=| c pr-OC r|+| c pg-OC g|+| c pb-OC b|, c wherein pr, c pgAnd c pbBe respectively r component value, g component value and the b component value of the color of data point p, OC r, OC gAnd OC bBe respectively r component value, g component value and the b component value of the color characteristic of fruit organ.If d f>50, data point p is rejected from data set D.Obtain at last only keeping the fruit three dimensional point cloud collection D of fruit information f
Step C: to fruit three dimensional point cloud collection D fAdd up, obtain N fruit three dimensional point cloud subset D fiWith fruit number N, concrete grammar is: to fruit three dimensional point cloud collection D fAnalyze, utilize the clustering method of distance-based to carry out cluster to data point, the data point that is about to belong to same fruit is divided in a subset, thereby with whole fruit 3-D data set D fBe divided into N little fruit three dimensional point cloud subset (remembered that every subset is D fi, wherein the i value is the natural number between 1 to N), N is the fruit number that calculates.
Step D: according to fruit three dimensional point cloud subset D fiCalculate the fruit radius r of each corresponding fruit i, obtaining fruit radius array, steps flow chart specifically comprises the following steps as shown in Figure 4:
Step D1: calculate each fruit three dimensional point cloud subset D fiInterior length and width parameter.At first calculate fruit three dimensional point cloud subset D fiMinimum external rectangular parallelepiped, then from rectangular length l and the width w of the minimum external rectangular parallelepiped that the calculates intercepting in the xoy plane.
Step D2: calculate each fruit three dimensional point cloud subset D fiCenter point coordinate v fi
Step D3: calculate each fruit three dimensional point cloud subset D fiIn all data points and center point coordinate v fiDistance, and calculate fruit three dimensional point cloud subset D fiMiddle have a few and center point coordinate v fiMean distance d fi
Step D4: calculate each fruit three dimensional point cloud subset D according to length and width parameter and mean distance fiCorresponding fruit radius r i=((l+w)/2+d fi/ 2)/2, all fruit radius r iForm fruit radius array.
E: calculate one by one fruit quality corresponding to each fruit radius in fruit radius array according to the underlying parameter in the relational model of fruit radius and fruit quality, accumulation calculating goes out the output of whole strain fruit tree.
Also comprise before step e:
Step S: measure fruit radius and the fruit quality of sample fruit, set up the relational model of fruit radius and fruit quality, obtain underlying parameter.This step is carried out before steps A again, also can be after step D, carry out before step e, carried out as example before steps A take step S in the process flow diagram of Fig. 1.
The Main Function of step S is to set up the fruit radius of fruit tree and the relational model of fruit quality by actual measurement, and method is:
In the orchard, needs are carried out the achievement phase fruit tree of yield monitoring, from setting fruit 8-20 that gathers different sizes, measure radius and the weight of each fruit, and utilize the Mathematical Fitting methods such as regretional analysis to set up the relational model g=r * λ of fruit radius and weight, wherein g is fruit weight, r is the fruit radius, and λ is the underlying parameter that match obtains.
The fruit radius and the relational model g=r * λ of fruit weight that adopt step S to set up, the radius of each fruit of taking-up and calculate the weight of each fruit from the fruit radius array that step D calculates, accumulation calculating can obtain the output of whole strain fruit tree.
Can also calculate other parameters by the result of calculation of utilizing above several step to obtain, the mean radius that calculates fruit as fruit radius and fruit number according to whole all fruits of strain fruit tree calculates mean fruit weight amount etc. according to fruit quality and the fruit number of whole all fruits of strain fruit tree.
By said method, in the situation that fruit tree and fruit all be can't harm, can measure the single tree of fruit tree output accurately and fast.
Embodiment two
For achieving the above object, the system of the single tree of fruit tree yield monitoring also is provided in embodiments of the invention two, form schematic diagram as shown in Figure 5, specifically comprise:
Pretreatment module 510, three dimensional point cloud are cut apart module 520, fruit counting module 530, fruit radius calculation module 540 and output of the fruit tree computing module 550.
Pretreatment module 510 is used for obtaining the three dimensional point cloud of fruit tree, and three dimensional point cloud is carried out pre-service obtains fruit tree three dimensional point cloud collection D.
Pretreatment module 510 specifically comprises: raw data set acquisition module 511, tentatively reject module 512 and secondary and carry and reject module 513.
Raw data set acquisition module 511 is used for the fruit tree of band fruit state is obtained the three dimensional point cloud of different angles, and forms fruit tree three-dimensional point cloud raw data set, wherein comprises colouring information in three dimensional point cloud.The achievement phase fruit tree that needs is carried out yield monitoring in the orchard utilizes laser 3 d scanner (for example FARO focus3D120) to carry out 3-D scanning, at least carry out multistation scanning from 3 angles of fruit tree during scanning, collect the fruit tree three-dimensional point cloud raw data set with colouring information.But utilize three-dimensional laser scanner under the environment of orchard, fruit tree to be carried out in the three dimensional point cloud acquisition process, due to blocking between different fruit trees and between organ, and the impact of other external conditions (as wind), the three dimensional point cloud that obtains inevitably can with noise spot, therefore need to carry out the noise spot in following two steps and remove operation.
The preliminary module 512 of rejecting, be used for the noise spot that fruit tree three-dimensional point cloud raw data is concentrated is tentatively rejected, concrete grammar is as follows: at first adopt minimum neighbours' method to carry out the preliminary rejecting of noise spot, each data point of namely fruit tree three-dimensional point cloud raw data being concentrated is carried out minimum neighbours and is checked, if the neighbours of certain data point (data point that is not more than 0.5cm apart from the air line distance of this point is the neighbours of this point) number is about to this data point and concentrates deletion from the three-dimensional point cloud raw data less than 30.
Secondary is carried and is rejected module 513, is used for that the noise spot that the fruit tree three-dimensional point cloud raw data after preliminary rejecting is concentrated is carried out secondary and rejects, and obtains only comprising the fruit tree three dimensional point cloud collection D of fruit tree information.Carrying out secondary rejects in order to remove the data point that the fruit tree its data could be rejected and don't belong to the minimum nearest neighbour method of those uses, method is to utilize general three-dimensional point cloud process software to import the above fruit tree three-dimensional point cloud raw data set that obtains, and does not belong to the data point of the organs such as fruit tree leaf, fruit, limb by those in the method rejecting three-dimensional point cloud of choosing alternately data point and deletion.Noise points deleting through above two steps is processed, and concentrates the noise spot that does not belong to fruit tree partly to remove the three-dimensional point cloud raw data of fruit tree, namely obtains only comprising the fruit tree three dimensional point cloud collection D of fruit tree information.
Three dimensional point cloud is cut apart module 520, is used for fruit tree three dimensional point cloud collection is cut apart the fruit three dimensional point cloud collection D that obtains only comprising fruit information f
Three dimensional point cloud is cut apart module 520 and is specifically comprised: color characteristic acquisition module 521 and fruit three dimensional point cloud collection acquisition module 522.
Color characteristic acquisition module 521 obtains the color characteristic of data point corresponding to fruit organ.At first choose 50 of data point corresponding to fruit organ by interactive mode from fruit tree three dimensional point cloud collection D, then calculate the color average of these 50 data points.The color of supposing certain data point i in 50 selected data points is c i(r, g, b), wherein the i value is the natural number between 1 to 50, the computing formula of the color characteristic of fruit organ is:
OC ( r , g , b ) = ( 1 50 Σ i = 1 50 c ir , 1 50 Σ i = 1 50 c ig , 1 50 Σ i = 1 50 c ib )
C wherein irBe the r component value of the color of i data point, c igBe the g component value of the color of i data point, c ibIt is the b component value of the color of i data point.
Fruit three dimensional point cloud collection acquisition module 522 according to the color value of each data point in fruit tree three dimensional point cloud collection D and the distance of color characteristic, is partitioned into fruit organ data point in addition, obtains fruit three dimensional point cloud collection D fSecondary is carried each the data point p that rejects in the three dimensional point cloud collection D that obtains in module 513, calculated respectively the color c of this data point pThe color characteristic of (r, g, b) and fruit apart from d f=| c pr-OC r|+| c pg-OC g|+| c pb-OC b|, c wherein pr, c pgAnd c pbBe respectively r component value, g component value and the b component value of the color of data point p, OC r, OC gAnd OC bBe respectively r component value, g component value and the b component value of the color characteristic of fruit organ.If d f>50, data point p is rejected from data set D.Obtain at last only keeping the fruit three dimensional point cloud collection D of fruit information f
Fruit counting module 530 is used for fruit three dimensional point cloud collection D fAdd up, obtain fruit number N.The method of statistics is: to fruit three dimensional point cloud collection D fAnalyze, utilize the clustering method of distance-based to carry out cluster to data point, the data point that is about to belong to same fruit is divided in a subset, thereby with whole fruit 3-D data set D fBe divided into N little fruit three dimensional point cloud subset (remember that every subset is, wherein the i value is the natural number between 1 to N), N is the fruit number that calculates.
Fruit radius calculation module 540 is used for according to fruit three dimensional point cloud subset D fiCalculate the fruit radius r of each corresponding fruit i, obtain fruit radius array.
Fruit radius calculation module 540 specifically comprises: length and width acquisition module 541, centre coordinate computing module 542, mean distance computing module 543 and fruit radius array acquisition module 544.
Length and width acquisition module 541 calculates each fruit three dimensional point cloud subset D fiInterior length and width parameter, the length and width parameter comprises length l and width w.
Centre coordinate computing module 542 calculates each fruit three dimensional point cloud subset D fiCenter point coordinate v fi
Mean distance computing module 543 calculates each fruit three dimensional point cloud subset D fiIn all data points and center point coordinate v fiDistance, and calculate fruit three dimensional point cloud subset D fiMiddle have a few and center point coordinate v fiMean distance d fi
Fruit radius array acquisition module 544, the mean distance d that length l, width w and the average distance calculation module 543 that obtains according to length and width acquisition module 541 obtains fiCalculate fruit radius r corresponding to each fruit three dimensional point cloud subset i=((l+w)/2+d fi/ 2)/2, all fruit radius r iForm fruit radius array.
Output of the fruit tree computing module 550 calculates fruit quality corresponding to each fruit radius in fruit radius array one by one according to the underlying parameter in the relational model of fruit radius and fruit quality, and accumulation calculating goes out the output of whole strain fruit tree.
This system also comprises: underlying parameter module 560, measure fruit radius and the fruit quality of sample fruit, and set up the relational model of fruit radius and fruit quality, obtain underlying parameter.Set up the fruit radius of fruit tree and the relational model of fruit quality by actual measurement, method is: the achievement phase fruit tree that in the orchard, needs is carried out yield monitoring, from setting fruit 8-20 that gathers different sizes, measure radius and the weight of each fruit, and utilize the Mathematical Fitting methods such as regretional analysis to set up the relational model g=r * λ of fruit radius and weight, wherein g is fruit weight, and r is the fruit radius, and λ is the underlying parameter that match obtains.
By said apparatus, in the situation that fruit tree and fruit all be can't harm, can measure the single tree of fruit tree output accurately and fast.
Above embodiment only is used for explanation the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; without departing from the spirit and scope of the present invention; can also make a variety of changes and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (9)

1. the method for the single tree of fruit tree yield monitoring, is characterized in that, described method specifically comprises:
A: obtain the three dimensional point cloud of fruit tree, and described three dimensional point cloud is carried out pre-service obtain fruit tree three dimensional point cloud collection;
B: described fruit tree three dimensional point cloud collection is cut apart the fruit three dimensional point cloud collection that obtains only comprising fruit information;
C: described fruit three dimensional point cloud collection is added up, obtained N fruit three dimensional point cloud subset and fruit number N, N is positive integer;
D: calculate the fruit radius of each corresponding fruit according to described fruit three dimensional point cloud subset, obtain fruit radius array;
E: calculate one by one fruit quality corresponding to each fruit radius in fruit radius array according to the underlying parameter in the relational model of fruit radius and fruit quality, accumulation calculating goes out the output of whole strain fruit tree;
Also comprise before described step e:
S: measure fruit radius and the fruit quality of sample fruit, set up the relational model of fruit radius and fruit quality, obtain described underlying parameter.
2. the method for claim 1, is characterized in that, described steps A specifically comprises:
A1: the fruit tree to band fruit state obtains the three dimensional point cloud of different angles, and forms fruit tree three-dimensional point cloud raw data set, comprises colouring information in wherein said three dimensional point cloud;
A2: the noise spot that described fruit tree three-dimensional point cloud raw data is concentrated is tentatively rejected;
A3: the noise spot of again the fruit tree three-dimensional point cloud raw data after preliminary rejecting being concentrated carries out secondary rejects, and obtains only comprising the fruit tree three dimensional point cloud collection of fruit tree information.
3. the method for claim 1, is characterized in that, described step B specifically comprises:
B1: the color characteristic that obtains data point corresponding to fruit organ;
B2: according to the color value of each concentrated data point of described fruit tree three dimensional point cloud and the distance of described color characteristic, be partitioned into fruit organ data point in addition, obtain described fruit three dimensional point cloud collection.
4. the method for claim 1, it is characterized in that, in described step C, described fruit three dimensional point cloud collection is added up and be specially: described fruit three dimensional point cloud collection is carried out cluster, the data point that will belong to same fruit is divided in a fruit three dimensional point cloud subset, and the subset number that obtains is exactly the fruit number.
5. the method for claim 1, is characterized in that, described step D specifically comprises:
D1: calculate the length and width parameter in each fruit three dimensional point cloud subset;
D2; Calculate the center point coordinate of each fruit three dimensional point cloud subset;
D3: calculate all data points in each fruit three dimensional point cloud subset and the distance between described center point coordinate, and calculate mean distance;
D4: calculate fruit radius corresponding to each fruit three dimensional point cloud subset according to described length and width parameter and described mean distance, all fruit radiuses form described fruit radius array.
6. the system of the single tree of fruit tree yield monitoring, is characterized in that, described system specifically comprises:
Pretreatment module, three dimensional point cloud are cut apart module, fruit counting module, fruit radius calculation module and output of the fruit tree computing module;
Described pretreatment module is obtained the three dimensional point cloud of fruit tree, and described three dimensional point cloud is carried out pre-service obtains fruit tree three dimensional point cloud collection;
Described three dimensional point cloud is cut apart module, described fruit tree three dimensional point cloud collection is cut apart the fruit three dimensional point cloud collection that obtains only comprising fruit information;
Described fruit counting module is added up described fruit three dimensional point cloud collection, obtains the fruit number;
Described fruit radius calculation module calculates the fruit radius of each corresponding fruit according to described fruit three dimensional point cloud subset, obtain fruit radius array;
Described output of the fruit tree computing module calculates fruit quality corresponding to each fruit radius in fruit radius array one by one according to the underlying parameter in the relational model of fruit radius and fruit quality, and accumulation calculating goes out the output of whole strain fruit tree;
Described system also comprises: the underlying parameter module, measure fruit radius and the fruit quality of sample fruit, and set up the relational model of fruit radius and fruit quality, obtain described underlying parameter.
7. system as claimed in claim 6, is characterized in that, described pretreatment module comprises: the raw data set acquisition module, tentatively reject module and secondary and put forward the rejecting module;
Described raw data set acquisition module obtains the three dimensional point cloud of different angles to the fruit tree of being with the fruit state, and forms fruit tree three-dimensional point cloud raw data set, comprises colouring information in wherein said three dimensional point cloud;
Described preliminary rejecting module is tentatively rejected the noise spot that described fruit tree three-dimensional point cloud raw data is concentrated;
Described secondary is put forward the rejecting module, and the noise spot that the fruit tree three-dimensional point cloud raw data after preliminary rejecting is concentrated carries out the secondary rejecting, obtains only comprising the fruit tree three dimensional point cloud collection of fruit tree information.
8. system as claimed in claim 6, is characterized in that, described three dimensional point cloud is cut apart module and comprised: color characteristic acquisition module and fruit three dimensional point cloud collection acquisition module;
Described color characteristic acquisition module obtains the color characteristic of data point corresponding to fruit organ;
Described fruit three dimensional point cloud collection acquisition module according to the color value of each concentrated data point of described fruit tree three dimensional point cloud and the distance of described color characteristic, is partitioned into fruit organ data point in addition, obtains described fruit three dimensional point cloud collection.
9. system as claimed in claim 6, is characterized in that, described fruit radius calculation module comprises: length and width acquisition module, centre coordinate computing module, mean distance computing module and fruit radius array acquisition module;
Described length and width acquisition module calculates the length and width parameter in each fruit three dimensional point cloud subset;
Described centre coordinate computing module calculates the center point coordinate of each fruit three dimensional point cloud subset;
Described mean distance computing module calculates all data points in each fruit three dimensional point cloud subset and the distance of described center point coordinate, and calculates mean distance;
Described fruit radius array acquisition module calculates fruit radius corresponding to each fruit three dimensional point cloud subset according to described length and width parameter and described mean distance, and all fruit radiuses form described fruit radius array.
CN201310014686.5A 2013-01-15 2013-01-15 Method and system for measuring single fruit tree yield Active CN103090946B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310014686.5A CN103090946B (en) 2013-01-15 2013-01-15 Method and system for measuring single fruit tree yield

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310014686.5A CN103090946B (en) 2013-01-15 2013-01-15 Method and system for measuring single fruit tree yield

Publications (2)

Publication Number Publication Date
CN103090946A true CN103090946A (en) 2013-05-08
CN103090946B CN103090946B (en) 2014-12-17

Family

ID=48203800

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310014686.5A Active CN103090946B (en) 2013-01-15 2013-01-15 Method and system for measuring single fruit tree yield

Country Status (1)

Country Link
CN (1) CN103090946B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103518516A (en) * 2013-10-21 2014-01-22 贵州省辣椒研究所 Method for measuring field yield of chilies
CN104677342A (en) * 2015-03-08 2015-06-03 无锡桑尼安科技有限公司 Apple orchard yield measuring method
CN104881626A (en) * 2015-01-19 2015-09-02 新疆农业大学 Recognition method for fruit of fruit tree
CN112215184A (en) * 2020-10-21 2021-01-12 安徽农业大学 Camellia oleifera fruit tree yield detection method based on three-dimensional laser scanner
CN112556606A (en) * 2020-12-24 2021-03-26 宁夏农林科学院农业经济与信息技术研究所(宁夏农业科技图书馆) Self-propelled wolfberry fruit actual measurement method and device based on binocular vision
CN115388993A (en) * 2022-10-08 2022-11-25 山东省果树研究所 Fruit tree yield measuring device and measuring method
CN115581142A (en) * 2022-11-25 2023-01-10 海南大学三亚研究院 Online detection device for number and peel breaking rate of sweet potatoes

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002008014A (en) * 2000-06-20 2002-01-11 Ricoh Co Ltd Method and device for extracting three-dimensional shape, and recording medium
CN101085442A (en) * 2006-06-06 2007-12-12 中国农业大学 Method for treating and classifying orange image based on RGB composite model
CN101493313A (en) * 2009-02-27 2009-07-29 中国农业大学 Image processing process for ripe fruit identification and positioning
WO2012020696A1 (en) * 2010-08-11 2012-02-16 株式会社トプコン Device for processing point group position data, system for processing point group position data, method for processing point group position data and program for processing point group position data
CN102682286A (en) * 2012-04-16 2012-09-19 中国农业大学 Fruit identification method of picking robots based on laser vision systems

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002008014A (en) * 2000-06-20 2002-01-11 Ricoh Co Ltd Method and device for extracting three-dimensional shape, and recording medium
CN101085442A (en) * 2006-06-06 2007-12-12 中国农业大学 Method for treating and classifying orange image based on RGB composite model
CN101493313A (en) * 2009-02-27 2009-07-29 中国农业大学 Image processing process for ripe fruit identification and positioning
WO2012020696A1 (en) * 2010-08-11 2012-02-16 株式会社トプコン Device for processing point group position data, system for processing point group position data, method for processing point group position data and program for processing point group position data
CN102682286A (en) * 2012-04-16 2012-09-19 中国农业大学 Fruit identification method of picking robots based on laser vision systems

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
STAJNKO D等: "Estimation of number and diameter of apple fruits in an orchard during the growing season bythermal imaging", 《COMPUTERS AND ELECTRONICS IN AGRICULTURE》 *
张亚静等: "基于图像处理的柑橘测产方法", 《农业机械学报》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103518516A (en) * 2013-10-21 2014-01-22 贵州省辣椒研究所 Method for measuring field yield of chilies
CN103518516B (en) * 2013-10-21 2015-07-22 贵州省辣椒研究所 Method for measuring field yield of chilies
CN104881626A (en) * 2015-01-19 2015-09-02 新疆农业大学 Recognition method for fruit of fruit tree
CN104881626B (en) * 2015-01-19 2017-12-29 新疆农业大学 The recognition methods of Tree Fruit
CN104677342A (en) * 2015-03-08 2015-06-03 无锡桑尼安科技有限公司 Apple orchard yield measuring method
CN112215184A (en) * 2020-10-21 2021-01-12 安徽农业大学 Camellia oleifera fruit tree yield detection method based on three-dimensional laser scanner
CN112556606A (en) * 2020-12-24 2021-03-26 宁夏农林科学院农业经济与信息技术研究所(宁夏农业科技图书馆) Self-propelled wolfberry fruit actual measurement method and device based on binocular vision
CN115388993A (en) * 2022-10-08 2022-11-25 山东省果树研究所 Fruit tree yield measuring device and measuring method
CN115581142A (en) * 2022-11-25 2023-01-10 海南大学三亚研究院 Online detection device for number and peel breaking rate of sweet potatoes
CN115581142B (en) * 2022-11-25 2024-01-30 海南大学三亚研究院 Online detection device for sweet potato quantity and peel breaking rate

Also Published As

Publication number Publication date
CN103090946B (en) 2014-12-17

Similar Documents

Publication Publication Date Title
CN103090946B (en) Method and system for measuring single fruit tree yield
CN106356757B (en) A kind of power circuit unmanned plane method for inspecting based on human-eye visual characteristic
Wang et al. Deep learning approach for apple edge detection to remotely monitor apple growth in orchards
CN104850836A (en) Automatic insect image identification method based on depth convolutional neural network
CN112164062A (en) Wasteland information extraction method and device based on remote sensing time sequence analysis
CN109544501A (en) A kind of transmission facility defect inspection method based on unmanned plane multi-source image characteristic matching
CN108197583A (en) The building change detecting method of optimization and image structure feature is cut based on figure
CN112418188A (en) Crop growth whole-course digital assessment method based on unmanned aerial vehicle vision
CN109241941A (en) A method of the farm based on deep learning analysis monitors poultry quantity
CN112749627A (en) Method and device for dynamically monitoring tobacco based on multi-source remote sensing image
CN107767364B (en) Method for accurately extracting temperature of tree canopy based on infrared thermal image
CN111178177A (en) Cucumber disease identification method based on convolutional neural network
Grocholsky et al. A camera and laser system for automatic vine balance assessment
CN109523528A (en) A kind of transmission line of electricity extracting method based on unmanned plane binocular vision SGC algorithm
CN111462058A (en) Method for quickly detecting effective ears of rice
CN116071424A (en) Fruit space coordinate positioning method based on monocular vision
Tang et al. Corn tassel detection based on image processing
CN106067173A (en) The Complexity Measurement lossless detection method of citrusfruit pol
CN101916439A (en) Space debris hyperspectral sequence detection method based on Hilbert-Huang transform
CN109344769A (en) A kind of photovoltaic plant detection method and system based on remote sensing image
KR20180096966A (en) Automatic Counting Method of Rice Plant by Centroid of Closed Rice Plant Contour Image
CN115631421A (en) Intelligent protection method and system for cultivated land
CN115457001A (en) Photovoltaic panel foreign matter detection method, system, device and medium based on VGG network
CN110554406B (en) Method for inverting secondary forest structure parameters based on unmanned aerial vehicle stereo photogrammetry point cloud
CN107967714A (en) A kind of method that forest canopy density is automatically extracted by unmanned plane digital elevation model

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant