CN106096219B - A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance - Google Patents
A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance Download PDFInfo
- Publication number
- CN106096219B CN106096219B CN201610087127.0A CN201610087127A CN106096219B CN 106096219 B CN106096219 B CN 106096219B CN 201610087127 A CN201610087127 A CN 201610087127A CN 106096219 B CN106096219 B CN 106096219B
- Authority
- CN
- China
- Prior art keywords
- fruit
- vegetable recognition
- vegetable
- statistician
- recognition
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/68—Food, e.g. fruit or vegetables
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Sorting Of Articles (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a kind of Data Quality Analysis methods for the evaluation of fruit and vegetable recognition algorithm performance, wherein it is used to evaluate the data of fruit and vegetable recognition algorithm performance as to the positive exact figures of fruit and vegetable recognition in the fruit and vegetable recognition result images obtained after fruits and vegetables image application fruit and vegetable recognition algorithm, using the statistical system for obtaining the positive exact figures of fruit and vegetable recognition in fruit and vegetable recognition result images as a kind of hard measurement system, the quality of data of the positive exact figures of fruit and vegetable recognition is analyzed using a kind of hard measurement systematic analytic method based on attribute agreement.Using present invention may determine that can truly reflect the degree of fruit and vegetable recognition algorithm performance, and then the effectiveness of the determining fruit and vegetable recognition algorithm performance evaluation conclusion obtained based on data used for evaluating the data of fruit and vegetable recognition algorithm performance.
Description
Technical field
The present invention relates to a kind of Data Quality Analysis methods for the evaluation of fruit and vegetable recognition algorithm performance.
Background technology
Fruits and vegetables production is time-consuming and laborious, and agricultural robot is a kind of extraordinary solution party for realizing fruits and vegetables production automation
Case.As the important component of agricultural robot, one of main task of vision system be realize fruits and vegetables machine vision from
Dynamic identification.Therefore, the fruit and vegetable recognition algorithm based on machine vision is the hot spot of current agricultural robot research field.In current base
In the fruit and vegetable recognition algorithm research of machine vision, fruit and vegetable recognition accuracy is for evaluating the main of fruit and vegetable recognition algorithm performance
Evaluation index.Fruit and vegetable recognition accuracy is the ratio that the positive exact figures of fruit and vegetable recognition account for identified fruits and vegetables sum.So fruit and vegetable recognition is just
The calculating of true rate need to be premised on obtaining the positive exact figures of fruit and vegetable recognition.In existing research, the positive exact figures of fruit and vegetable recognition rely primarily on naked eyes
Artificial statistics acquisition is carried out to having been subjected to the fruit and vegetable recognition result images obtained after fruit and vegetable recognition algorithm process.Fruit and vegetable recognition is correct
Several statistical results is easy to be influenced by many factors such as statistician, statistical method, statistical environments, therefore has one
Fixed variation.In fact, different statisticians repeatedly unite to same fruit and vegetable recognition result images using same statistical method
The statistical result of meter, the positive exact figures of fruit and vegetable recognition is deteriorated there may be certain.Even if same statistician, to same fruit and vegetable recognition knot
Fruit image carries out multiplicating statistics using same statistical method, and the statistical result of the positive exact figures of fruit and vegetable recognition is it is equally possible that there are one
It is fixed to be deteriorated.If counting, the positive exact figures variation of obtained fruit and vegetable recognition is excessive, and the positive exact figures of fruit and vegetable recognition for counting gained can not be true
Reflect the performance of fruit and vegetable recognition algorithm on the spot, i.e. the quality of data of the positive exact figures of fruit and vegetable recognition is poor, therefore is based on such fruits and vegetables
Identify that the result for the fruit and vegetable recognition algorithm performance evaluation that positive exact figures obtain is also insecure.However, being calculated in current fruit and vegetable recognition
In the research of method, after statistics obtains the positive exact figures of fruit and vegetable recognition, the positive exact figures of fruit and vegetable recognition are used directly to evaluation fruit and vegetable recognition and calculate
The performance of method, without the Data Quality Analysis by the positive exact figures of fruit and vegetable recognition.In addition, at present also without being calculated for fruit and vegetable recognition
The Data Quality Analysis method of method performance evaluation.Therefore, in fruit and vegetable recognition algorithm research, it is also very desirable to which one kind can be used in fruit
The Data Quality Analysis method of vegetable recognizer performance evaluation, to ensure that the data using the high quality of data carry out fruit and vegetable recognition calculation
The correct evaluation of method performance.
Invention content
The purpose of the present invention is to provide it is a kind of for fruit and vegetable recognition algorithm performance evaluation Data Quality Analysis method, with
Determine that the data for the evaluation of fruits and vegetables algorithm performance can truly reflect the degree of fruit and vegetable recognition algorithm performance and be based on institute
With the effectiveness for the fruit and vegetable recognition algorithm performance evaluation conclusion that data obtain.
The technical solution adopted by the present invention is：
1. a kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance, wherein for evaluating fruit and vegetable recognition
The data of algorithm performance is to the fruits and vegetables knowledges in the fruit and vegetable recognition result images obtained after fruits and vegetables image application fruit and vegetable recognition algorithm
Not positive exact figures；Using the statistical system for obtaining the positive exact figures of fruit and vegetable recognition in fruit and vegetable recognition result images as a kind of hard measurement
System, the method for obtaining the positive exact figures of fruit and vegetable recognition in fruit and vegetable recognition result images are：The every width of statistician's naked eyes qualitative observation
The recognition result of all fruits and vegetables in fruit and vegetable recognition result images, i.e., by observe fruit and vegetable recognition algorithm obtain fruits and vegetables fitting circle with
The tightness degree of fruits and vegetables edge fitting judges the correctness of recognition result；If being fitted closely, the fruit and vegetable recognition is correct,
The positive exact figures of fruit and vegetable recognition add 1；If fruits and vegetables fitting circle is bigger than normal or less than normal or fruits and vegetables fitting circle position is there are deviation, which knows
Not mistake, the positive exact figures of fruit and vegetable recognition remain unchanged；Statistics obtains the positive exact figures of fruit and vegetable recognition in every width fruit and vegetable recognition result images；Make
The quality of data of the positive exact figures of fruit and vegetable recognition is divided with a kind of hard measurement systematic analytic method based on attribute agreement
Analysis, comprises the following steps：1. the value that variable p is arranged is 0；2. determining the fruit and vegetable recognition obtained after m width application fruit and vegetable recognition algorithms
Result images, and to this m width fruit and vegetable recognition result images from 1 to m serial numbers, wherein m is to be actually used in fruit recognizer
The fruit and vegetable recognition result figure film size number of performance test；3. selecting n statisticians, number is respectively 1 to n, and wherein n is by formula (1)
It is calculated：
4. by statistician n by the number random alignment of m width fruit and vegetable recognition result images, then by statistician n by m width
Fruit and vegetable recognition result images give statistician 1 by the number order after random alignment by width, and statistician 1 uses fruit and vegetable recognition
Positive exact figures statistical method carries out the statistics of the positive exact figures of fruit and vegetable recognition by width, and records statistical result, statistics by statistician n
Personnel 1 do not know the statistics sequence and statistical result of m width fruit and vegetable recognition result images；5. by statistician 1 by m width fruit and vegetable recognitions
Then m width fruit and vegetable recognition result images are pressed the number after random alignment by the number random alignment of result images by statistician 1
Sequence gives statistician 2 by width, and statistician 2 carries out a fruit and vegetable recognition using the positive exact figures statistical method of fruit and vegetable recognition by width
The statistics of positive exact figures, and statistical result is recorded by statistician 1, statistician 2 does not know the system of m width fruit and vegetable recognition result images
Meter sequence and statistical result；6. if n is more than 2, step is jumped to 7.；If n adds 1 equal to 2, p, step is then branched to 8.；
7. repeating step after the statistician 2 during then 5. statistician 3 distinguishes step of replacing to statistician n 5. one time, p adds 1；⑧
Judge whether p is equal to 2；If p is equal to 2, step is jumped to 9.；If p is not equal to 2,4. interval jumps to step after 7 days；⑨
Computational methods counting statistics personnel 1 using the Fleiss Kappa values in attribute agreement in measurement System Analysis and system
The repeatability and reproducibility of the quality of data for the corresponding characterization positive exact figures of fruit and vegetable recognition of whole statistical results that meter personnel n is recorded
Fleiss Kappa values；10. the repeatability and reproducibility of the quality of data of the positive exact figures of characterization fruit and vegetable recognition 9. obtained to step
Fleiss Kappa value application measurement System Analysis in the evaluation based on Fleiss Kappa values in attribute agreement
Criterion finally obtains the Data Quality Analysis of the positive exact figures of fruit and vegetable recognition as a result, obtaining the data of the characterization positive exact figures of fruit and vegetable recognition
The analysis result of the repeatability and reproducibility of quality.
2. a kind of composition of hard measurement system described in is as follows：It will be to obtaining after fruits and vegetables image application fruit and vegetable recognition algorithm
Fruit and vegetable recognition result images are as the measurand in hard measurement system；The fruit and vegetable recognition in fruit and vegetable recognition result images will be counted
The statistical method of positive exact figures is as the soft measurer in hard measurement system；Using statistician as the measurement people in hard measurement system
Member；Using the working environment of statistician as the measuring environment in hard measurement system.
3. a kind of soft measurer described in is different from the measurer with physical entity, is a kind of statistics fruit and vegetable recognition result figure
The measurement object of the statistical method of the positive exact figures of fruit and vegetable recognition as in, this soft measurer is fruit and vegetable recognition result images, measures knot
Fruit is the positive exact figures of fruit and vegetable recognition of fruit and vegetable recognition result images, and wherein fruit and vegetable recognition result images are to fruits and vegetables image application fruits and vegetables
The result obtained after recognizer.
4. survey crew described in is correctly to grasp the statistics people of the positive exact figures statistical method of fruit and vegetable recognition by training
Member.
The invention has the advantages that：
The present invention is analyzed by the quality of data to the data for evaluating fruit and vegetable recognition algorithm, is determined for evaluating
The quality of data of the data of fruit and vegetable recognition algorithm, can be true for evaluating the data of fruit and vegetable recognition algorithm performance so as to determination
The degree of reflection fruit and vegetable recognition algorithm performance and the fruit and vegetable recognition algorithm performance evaluation conclusion obtained based on data used on the spot
Effectiveness, and then can by using the data of the high quality of data carry out fruit and vegetable recognition algorithm actual performance correct evaluation,
Ensure the validity of fruit and vegetable recognition algorithm performance evaluation conclusion.
Description of the drawings
Fig. 1 is hard measurement system composition schematic diagram.
Fig. 2 is the hard measurement systematic analytic method figure based on attribute agreement.
Fig. 3 is soft gage measuring principle schematic.
Fig. 4 is soft measurer input and output schematic diagram.
Specific implementation mode
The invention will be further described with reference to the accompanying drawings and examples.
A kind of Data Quality Analysis method for the tomato recognizer performance evaluation that is blocked is implemented as follows：
1. a kind of Data Quality Analysis method for the tomato recognizer performance evaluation that is blocked is blocked for evaluating
The data of tomato recognizer performance are being blocked of being obtained after tomato recognizer of being blocked to the Tomato Image application that is blocked
The tomato that is blocked in tomato recognition result image identifies positive exact figures；It will be blocked in tomato recognition result image for obtaining
The tomato that is blocked identifies the statistical system of positive exact figures as a kind of hard measurement system, and acquisition is blocked in tomato recognition result image
The tomato that the is blocked method that identifies positive exact figures be：The every width of statistician's naked eyes qualitative observation is blocked tomato recognition result image
In all tomatoes that are blocked recognition result, i.e., by observation be blocked tomato recognizer acquisition the tomato fitting circle that is blocked
The correctness of recognition result is judged with the tightness degree of tomato edge fitting；If being fitted closely (as shown in 1 in Fig. 3),
Then the tomato identification that is blocked is correct, and the tomato that is blocked identifies that positive exact figures add 1；If the tomato fitting circle that is blocked is less than normal (in such as Fig. 3
Shown in 2) or there are deviation (as shown in 4 in Fig. 3) (as shown in 3 in Fig. 3) bigger than normal or the tomato fitting circle position that is blocked, then should
The tomato that is blocked identifies that mistake, the tomato that is blocked identify that positive exact figures remain unchanged；Statistics obtain every width be blocked tomato identification knot
The tomato that is blocked in fruit image identifies positive exact figures；Use a kind of hard measurement systematic analytic method pair based on attribute agreement
The tomato that is blocked identifies that the quality of data of positive exact figures is analyzed, and comprises the following steps：1. the value that variable p is arranged is 0；2. determining
M=300 width applications are blocked the tomato recognition result image that is blocked obtained after tomato recognizer, and are hidden to this 300 width
Tomato recognition result image is kept off from 1 to 300 serial number；3. selecting n=3 statisticians, number is respectively 1 to 3；Wherein n
It is calculated by formula (1)：
4. the number random alignment for tomato recognition result image that 300 width are blocked by statistician 3, then by counting people
Member 3 gives the 300 width tomato recognition result image that is blocked to statistician 1 by the number order after random alignment by width, counts
Personnel 1 identify that positive exact figures statistical method carries out the statistics that the primary tomato that is blocked identifies positive exact figures by width using the tomato that is blocked,
And by statistician 3 record statistical result, statistician 1 do not know 300 width be blocked tomato recognition result image statistics it is suitable
Sequence and statistical result；5. the number random alignment for tomato recognition result image that 300 width are blocked by statistician 1, then by
Statistician 1 gives the 300 width tomato recognition result image that is blocked to statistician by the number order after random alignment by width
2, statistician 2 identifies that positive exact figures statistical method carries out the primary tomato that is blocked by width and identifies positive exact figures using the tomato that is blocked
Statistics, and record statistical result by statistician 1, statistician 2 do not know that 300 width are blocked the system of tomato recognition result image
Meter sequence and statistical result；6. because n is more than 2, step is jumped to 7.；7. 300 width are blocked tomato by statistician 1
Then 300 width are blocked tomato recognition result image by random by the number random alignment of recognition result image by statistician 1
Number order after arrangement gives statistician 3 by width, 3 use of statistician be blocked tomato identify positive exact figures statistical method by
Width carries out the statistics that the primary tomato that is blocked identifies positive exact figures, and records statistical result by statistician 1, and statistician 3 does not know
300 width of road be blocked tomato recognition result image statistics sequence and statistical result；P adds 1；8. judging whether p is equal to 2；If p etc.
In 2, then step is jumped to 9.；If p is not equal to 2,4. interval jumps to step after 7 days；9. using belonging in measurement System Analysis
The computational methods of Fleiss Kappa values in property consistency analysis, as shown in formula (2), counting statistics personnel 1 and statistician 3
The corresponding repeatability and reproducibility for characterizing the tomato that is blocked and identifying the quality of data of positive exact figures of whole statistical results recorded
Fleiss Kappa values；Wherein repeatability is that every statistician of characterization is respectively blocked tomato recognition result figure to 300 width
As repeating statistics 2 times, the consistency between 2 statistical results；The Fleiss Kappa values for calculating repeatability, need to calculate each statistics
The repeated Fleiss Kappa values of personnel；It is respectively that every statistician is respectively blocked kind to 300 width to calculate data used
Eggplant recognition result image repeats to count 2 statistical results；Reproducibility is that 3 300 width of statistician couple of characterization are blocked tomato
After recognition result image is counted, the consistency of the statistical result between 3 statisticians；Calculate the Fleiss of reproducibility
Kappa values need to use the statistical result of 3 statisticians；
In formula：kF tIt is Fleiss Kappa；N is the tomato recognition result amount of images that is blocked；M is that every width is blocked tomato
Recognition result image is by statistics total degree；A is that single width is blocked the tomato number that is blocked occurred in tomato recognition result image
The quantity of probable value；xikIt is the i-th width and the tomato number that is wherein blocked is being united for a tomato recognition result images that are blocked of k
Metering number；xkBe be blocked the tomato recognition result image that is blocked that tomato number is k by statistics number；10. 9. to step
Obtained characterization be blocked tomato identify positive exact figures the quality of data repeatability and reproducibility Fleiss Kappa value applications
The interpretational criteria based on Fleiss Kappa values in measurement System Analysis in attribute agreement is finally obtained and is blocked kind
Eggplant identifies the Data Quality Analysis of positive exact figures as a result, obtaining characterizing the repetition for the quality of data that the tomato that is blocked identifies positive exact figures
The analysis result of property and reproducibility；Wherein, the interpretational criteria based on Fleiss Kappa values is：If Fleiss Kappa values are less than
0.4, statistical result consistency is poor；If Fleiss Kappa values are greater than or equal to 0.4 and are less than or equal to 0.75, statistical result one
Cause property is preferable；If Fleiss Kappa values are more than 0.75, statistical result consistency is very good.
2. a kind of composition of hard measurement system described in is as shown in Figure 1, as follows：The Tomato Image that is blocked will be applied and be hidden
The tomato recognition result image that is blocked obtained after gear tomato recognizer is as the measurand in hard measurement system；It will statistics
The tomato that is blocked being blocked in tomato recognition result image identifies the statistical method of positive exact figures as soft in hard measurement system
Measurer；Using statistician as the survey crew in hard measurement system；Using the working environment of statistician as hard measurement system
In measuring environment.
A kind of soft measurer described in 3., be different from physical entity measurer, be it is a kind of statistics be blocked tomato identification
The tomato that is blocked in result images identifies that the statistical method of positive exact figures, soft measurer are output and input as shown in Figure 4：It is this soft
Input, that is, measurement object of measurer is the tomato recognition result image that is blocked, and it is to be blocked that this soft measurer output, which is measurement result,
The tomato that is blocked of tomato recognition result image identifies positive exact figures, for carrying out Data Quality Analysis, wherein the tomato knowledge that is blocked
Other result images are to be blocked the result obtained after tomato recognizer to the Tomato Image application that is blocked.
4. survey crew described in is correctly to have grasped the system that the tomato that is blocked identifies positive exact figures statistical method by training
Meter personnel.
Test result is shown：Statistician 1,2,3 respective repeatability Fleiss Kappa values are respectively 0.7084,
0.8743 and 0.8359, show that the repeatability of statistician 1 and statistician 2 are very good, the repeatability of statistician 3 is preferably.
The Fleiss Kappa values of reproducibility are 0.5915, show that reproducibility is preferable.That is 3 statisticians are corresponding all to be blocked kind
Eggplant identifies that the reproducibility data quality of positive exact figures statistical result is preferable, the tomato identification that is blocked of statistician 1 and statistician 2
The repeated quality of data of positive exact figures statistical result is very good, and the tomato that is blocked of statistician 3 identifies positive exact figures statistical result
The repeated quality of data it is preferable.The tomato that is blocked to ensure to be counted identifies that positive exact figures height reflects the tomato identification that is blocked
The actual performance of algorithm, and then ensure to identify that the tomato identification that is blocked that positive exact figures obtain is calculated based on the tomato that is blocked counted
The validity of method Evaluation results, it is proposed that there is extraordinary number using obtained by statistician 2 or the statistics of statistician 3
The tomato recognizer performance evaluation that is blocked is carried out according to the tomato identification number that is blocked of quality.
Claims (4)
1. a kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance, which is characterized in that for evaluating fruits and vegetables
The data of recognizer performance is to the fruits in the fruit and vegetable recognition result images obtained after fruits and vegetables image application fruit and vegetable recognition algorithm
Vegetable identifies positive exact figures；
Using for obtaining the statistical systems of the positive exact figures of fruit and vegetable recognition in fruit and vegetable recognition result images as a kind of hard measurement system,
The method of the positive exact figures of fruit and vegetable recognition obtained in fruit and vegetable recognition result images is：The every width fruits and vegetables of statistician's naked eyes qualitative observation are known
The recognition result of all fruits and vegetables in other result images, i.e., by observing the fruits and vegetables fitting circle and fruits and vegetables side that fruit and vegetable recognition algorithm obtains
The tightness degree of edge fitting judges the correctness of recognition result；If being fitted closely, the fruit and vegetable recognition is correct, and fruits and vegetables are known
Not positive exact figures add 1；If fruits and vegetables fitting circle is bigger than normal or less than normal or fruits and vegetables fitting circle position is there are deviation, the fruit and vegetable recognition is wrong
Accidentally, the positive exact figures of fruit and vegetable recognition remain unchanged；Statistics obtains the positive exact figures of fruit and vegetable recognition in every width fruit and vegetable recognition result images；
Using a kind of hard measurement systematic analytic method based on attribute agreement to the quality of data of the positive exact figures of fruit and vegetable recognition
It is analyzed, is comprised the following steps：
1. the value that variable p is arranged is 0；
2. determining the fruit and vegetable recognition result images obtained after m width application fruit and vegetable recognition algorithms, and to this m width fruit and vegetable recognition result figure
As from 1 to m serial numbers, wherein m is the fruit and vegetable recognition result figure film size number for being actually used in fruit recognizer performance test；
3. selecting n statisticians, number is respectively 1 to n, and wherein n is calculated by formula (1)：
4. by statistician n by the number random alignment of m width fruit and vegetable recognition result images, then by statistician n by m width fruits and vegetables
Recognition result image gives statistician 1 by the number order after random alignment by width, and statistician 1 is correct using fruit and vegetable recognition
Number statistical method carries out the statistics of the positive exact figures of fruit and vegetable recognition by width, and records statistical result, statistician by statistician n
1 does not know the statistics sequence and statistical result of m width fruit and vegetable recognition result images；
5. by statistician 1 by the number random alignment of m width fruit and vegetable recognition result images, then by statistician 1 by m width fruits and vegetables
Recognition result image gives statistician 2 by the number order after random alignment by width, and statistician 2 is correct using fruit and vegetable recognition
Number statistical method carries out the statistics of the positive exact figures of fruit and vegetable recognition by width, and records statistical result, statistician by statistician 1
2 do not know the statistics sequence and statistical result of m width fruit and vegetable recognition result images；
6. if n is more than 2, step is jumped to 7.；If n adds 1 equal to 2, p, step is then branched to 8.；
7. repeating step after the statistician 2 during then 5. statistician 3 distinguishes step of replacing to statistician n 5. one time, p adds
1；
8. judging whether p is equal to 2；If p is equal to 2, step is jumped to 9.；If p is not equal to 2, interval jumps to step after 7 days
Suddenly 4.；
9. using the computational methods counting statistics people of the Fleiss Kappa values in attribute agreement in measurement System Analysis
The repeatability of the quality of data of the 1 characterization positive exact figures of fruit and vegetable recognition corresponding with whole statistical results that statistician n is recorded of member
With the Fleiss Kappa values of reproducibility；
10. the Fleiss of the repeatability and reproducibility of the quality of data of the positive exact figures of characterization fruit and vegetable recognition 9. obtained to step
The interpretational criteria based on Fleiss Kappa values in Kappa value application measurement System Analysis in attribute agreement, finally
The Data Quality Analysis of the positive exact figures of fruit and vegetable recognition is obtained as a result, obtaining the repetition of the quality of data of the characterization positive exact figures of fruit and vegetable recognition
The analysis result of property and reproducibility.
2. a kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance as described in claim 1, feature
It is, a kind of composition of hard measurement system is as follows：By the fruits and vegetables to being obtained after fruits and vegetables image application fruit and vegetable recognition algorithm
Recognition result image is as the measurand in hard measurement system；By counting, the fruit and vegetable recognition in fruit and vegetable recognition result images is correct
Several statistical methods is as the soft measurer in hard measurement system；Using statistician as the survey crew in hard measurement system；It will
The working environment of statistician is as the measuring environment in hard measurement system.
3. a kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance as claimed in claim 2, feature
It is, a kind of soft measurer, is different from the measurer with physical entity, is in a kind of statistics fruit and vegetable recognition result images
The measurement object of the statistical method of the positive exact figures of fruit and vegetable recognition, this soft measurer is fruit and vegetable recognition result images, and measurement result is fruit
The positive exact figures of fruit and vegetable recognition of vegetable recognition result image, wherein fruit and vegetable recognition result images are to calculate fruits and vegetables image application fruit and vegetable recognition
The result obtained after method.
4. a kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance as claimed in claim 2, feature
It is, the survey crew is correctly to grasp the statistician of the positive exact figures statistical method of fruit and vegetable recognition by training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610087127.0A CN106096219B (en) | 2016-01-29 | 2016-01-29 | A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610087127.0A CN106096219B (en) | 2016-01-29 | 2016-01-29 | A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106096219A CN106096219A (en) | 2016-11-09 |
| CN106096219B true CN106096219B (en) | 2018-09-21 |
Family
ID=58702343
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610087127.0A Expired - Fee Related CN106096219B (en) | 2016-01-29 | 2016-01-29 | A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106096219B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109255795B (en) * | 2018-09-11 | 2021-04-06 | 中国计量大学 | Tomato plant edge sorting method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177445A (en) * | 2013-03-13 | 2013-06-26 | 浙江大学 | Outdoor tomato identification method based on subsection threshold image segmentation and light spot identification |
| CN103336946A (en) * | 2013-06-17 | 2013-10-02 | 浙江大学 | Binocular stereoscopic vision based clustered tomato identification method |
-
2016
- 2016-01-29 CN CN201610087127.0A patent/CN106096219B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103177445A (en) * | 2013-03-13 | 2013-06-26 | 浙江大学 | Outdoor tomato identification method based on subsection threshold image segmentation and light spot identification |
| CN103336946A (en) * | 2013-06-17 | 2013-10-02 | 浙江大学 | Binocular stereoscopic vision based clustered tomato identification method |
Non-Patent Citations (3)
| Title |
|---|
| Research on image segmentation methods of tomato in nature conditions;Rong Xiang等;《In 2011 4th International congress on image and signal processing》;20111231;1268-1272 * |
| 属性值测量一致性分析及其对质量判定的影响分析;戚鹏;《中国优秀硕士学位论文全文数据库经济与管理科学辑》;20111215;J145-642 * |
| 开放环境中番茄的双目立体视觉识别与定位;项荣;《中国博士学位论文全文数据库农业科技辑》;20140715;D048-1 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106096219A (en) | 2016-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100489870C (en) | Method and multidimensional system for statistical process control | |
| CN105528288B (en) | A kind of method for testing software and device | |
| CN105374367B (en) | Abnormal frame detection method and device | |
| US20090093999A1 (en) | Method and system for quantifying damage in a structure | |
| White et al. | Methods for assessing reliability and validity for a measurement tool: a case study and critique using the WHO haemoglobin colour scale | |
| Seifi et al. | Variable sampling plan for resubmitted lots based on process capability index and Bayesian approach | |
| CN108270636A (en) | Link-quality-evaluating method and device | |
| CN103424420A (en) | Fitting-based nuclear magnetic resonance signal processing method | |
| Sokolov | Sensitivity of goodness of fit indices to lack of measurement invariance with categorical indicators and many groups | |
| CN105045220B (en) | A kind of method of quality control based on multivariable Z score quality control chart for being used for laboratory diagnosis field or field of industrial production | |
| Cervantes | DFIT: An R package for Raju's differential functioning of items and tests framework | |
| CN106096219B (en) | A kind of Data Quality Analysis method for the evaluation of fruit and vegetable recognition algorithm performance | |
| AU2020100135A4 (en) | Method, system and apparatus for evaluating sensory assessors’ concentration ability | |
| CN101592692B (en) | Evaluation method of measuring machines | |
| Haindl et al. | Texture fidelity benchmark | |
| CN115841049B (en) | Equipment life reliability assessment method and system based on multi-source information fusion | |
| Vanacore et al. | Robustness of κ‐type coefficients for clinical agreement | |
| CN110348094A (en) | Petroleum pipeline leakage detection method and system based on influence network | |
| Bashkansky et al. | Some metrological aspects of the comparison between two ordinal measuring systems | |
| CN114489760A (en) | Code quality evaluation method and code quality evaluation device | |
| US6748279B2 (en) | Method for improving a manufacturing process by conducting a full factorial experiment to optimize process variable settings | |
| CN110338748B (en) | Method for quickly positioning vision value, storage medium, terminal and vision detector | |
| Spagnolo et al. | Forensic metrology: uncertainty of measurements in forensic analysis | |
| Lemeshko et al. | Extending the application of Grubbs-type tests in rejecting anomalous measurements | |
| CN110514796A (en) | A kind of sample collection detection method and device based on experience curve |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180921 Termination date: 20190129 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |