Chapter 14 Chapter 14
Accuracy Assessment Accuracy Assessment Accuracy Assessment Accuracy Assessment
Introduction to Remote Sensing, Introduction to Remote Sensing,
James B. Campbell
James B. Campbell pp
O tline O tline Outline Outline
Definisi Definisi
Accuracy & PrecissionAccuracy & Precission
SignificanceSignificance
SignificanceSignificance
Source of Classification Error Source of Classification Error
Error Characteristics Error Characteristics
Measurement of Map Accuracy Measurement of Map Accuracy
Error MatrixError Matrix
Omission & CommissionOmission & Commission
User & Producer AccuracyUser & Producer Accuracy
Interpretation of the Error matrix Interpretation of the Error matrix
Percentage CorrectPercentage Correct
Percentage CorrectPercentage Correct
Quantitative Assessment of Error matrixQuantitative Assessment of Error matrix
Definition Definition Definition Definition
Accuracy : correctness, mengukur “kecocokan” antara Accuracy : correctness, mengukur “kecocokan” antara suatu image yg tidak diketahui kualitasnya dengan
suatu image yg tidak diketahui kualitasnya dengan sebuah standar image
sebuah standar image
Precission : detail, “The distinction is important because Precission : detail, “The distinction is important because one may be able to increase accuracy by decreasing
one may be able to increase accuracy by decreasing precission”,
precission”,
Meningkatkan detail = menambah ragam kategori . Misal Meningkatkan detail = menambah ragam kategori . Misal
: forest = caniferous, pine, shortleaf pine atau mature : forest = caniferous, pine, shortleaf pine atau mature shortleaf pine
shortleaf pineÎ Î akan menambah peluang klasifikasi akan menambah peluang klasifikasi error
error error error
Statistical context : high accuracy = low bias, (estimated Statistical context : high accuracy = low bias, (estimated value is consistenrly close to an accepted reference
value is consistenrly close to an accepted reference value)
value)
value)
value)
Definition
Definition
Definition
Definition
Definition Definition Definition Definition
Significance Significance
Accuracy has many practical implication : Accuracy has many practical implication : Accuracy has many practical implication : Accuracy has many practical implication :
effect legal standing, operational usefulness, effect legal standing, operational usefulness, validity for scientific research.
validity for scientific research.
So ce of Classification E o So ce of Classification E o Source of Classification Error Source of Classification Error
Manual Interpretation : Misidentification, Manual Interpretation : Misidentification, Excessive generalization, Error registration, Excessive generalization, Error registration, Variation in detail of interpretation etc.
Variation in detail of interpretation etc.
Character of landscape : parcel size, variation in Character of landscape : parcel size, variation in
l i l id i i b f
l i l id i i b f
parcel size, parcel identities number of parcel size, parcel identities number of
categories, arrangement of categories, number categories, arrangement of categories, number of parcel per category shapes of parcel
of parcel per category shapes of parcel of parcel per category, shapes of parcel, of parcel per category, shapes of parcel, radiometric and spectral contrast with radiometric and spectral contrast with surrounding parcel
surrounding parcel
surrounding parcel
surrounding parcel
So ce of Classification E o So ce of Classification E o Source of Classification Error Source of Classification Error
Three error types dominate:
Three error types dominate:
Data Acquisition Errors: These include sensor performance, stability of Data Acquisition Errors: These include sensor performance, stability of the platform, and conditions of viewing. We can reduce them or
the platform, and conditions of viewing. We can reduce them or compensate for them by making systematic corrections (e g by compensate for them by making systematic corrections (e g by compensate for them by making systematic corrections (e.g., by compensate for them by making systematic corrections (e.g., by calibrating detector response with on
calibrating detector response with on--board light sources generating board light sources generating known radiances). We can make corrections, often modified by
known radiances). We can make corrections, often modified by
ancillary data such as known atmospheric conditions, during the initial ancillary data such as known atmospheric conditions, during the initial processing of the raw data.
processing of the raw data.
p g
p g
Data Processing Errors: An example is misregistration of equivalent Data Processing Errors: An example is misregistration of equivalent pixels in the different bands of the Landsat Thematic Mapper. The goal pixels in the different bands of the Landsat Thematic Mapper. The goal in geometric correction is to hold the mismatch to a displacement of no in geometric correction is to hold the mismatch to a displacement of no more than one pixel. Under ideal conditions, and with as many as 25 more than one pixel. Under ideal conditions, and with as many as 25
d l (GC ) d d l h
d l (GC ) d d l h
ground control points (GCP) spread around a scene, we can realize this ground control points (GCP) spread around a scene, we can realize this goal. Misregistrations of several pixels significantly compromise
goal. Misregistrations of several pixels significantly compromise accuracy.
accuracy.
SceneScene--dependent Errors: As alluded to in the previous page, one such dependent Errors: As alluded to in the previous page, one such error relates to how we define and establish the class which in turn is error relates to how we define and establish the class which in turn is error relates to how we define and establish the class, which, in turn, is error relates to how we define and establish the class, which, in turn, is sensitive to the resolution of the observing system and the reference sensitive to the resolution of the observing system and the reference map or photo. Mixed pixels fall into this category.
map or photo. Mixed pixels fall into this category.
So ce of Classification E o
So ce of Classification E o
Source of Classification Error
Source of Classification Error
E o Cha acte istics E o Cha acte istics Error Characteristics Error Characteristics
Classification error : assignment pixel to one category Classification error : assignment pixel to one category that different from true category ( as determined
that different from true category ( as determined ground observation/ground
ground observation/ground--truth ). truth ).
Error Characteristic : Error Characteristic :
Error are not distributed over the image at random, display a Error are not distributed over the image at random, display a gg ,, p yp y degree of systematic, ordered occurrence in space.
degree of systematic, ordered occurrence in space.
Often erroneously assigned pixels are not spatially isolated but Often erroneously assigned pixels are not spatially isolated but occur grouped in areas of varied size and shape (Campbell 1981) occur grouped in areas of varied size and shape (Campbell 1981)
Errors may have specific spatial relationships to the parcels toErrors may have specific spatial relationships to the parcels to
Errors may have specific spatial relationships to the parcels to Errors may have specific spatial relationships to the parcels to which they pertain, for example, they may tend to occur at which they pertain, for example, they may tend to occur at edges or in the interiors of the parcels
edges or in the interiors of the parcels
E o Cha acte istics E o Cha acte istics Error Characteristics Error Characteristics
Tiga macam eror patern dari Landsat, Cogalton (1984). Dark area = error clasification, white area = correct.
Meas ement of Map Acc ac Meas ement of Map Acc ac Measurement of Map Accuracy Measurement of Map Accuracy
Compare the “true map”/ reference map, Compare the “true map”/ reference map, (asumsi lebih akurat) with image to be
(asumsi lebih akurat) with image to be
( ) g
( ) g
evaluated.
evaluated.
Jika pembandingan tanpa memperhatikan Jika pembandingan tanpa memperhatikan
Jika pembandingan tanpa memperhatikan Jika pembandingan tanpa memperhatikan posisi pixel, klasifikasi total bisa dianggap posisi pixel, klasifikasi total bisa dianggap sama meskipun sebenarnya posisi dengan sama meskipun sebenarnya posisi dengan sama meskipun sebenarnya posisi dengan sama meskipun sebenarnya posisi dengan image reference tidak sesuai
image reference tidak sesuaiÎ Î site site-- spesific accuracy
spesific accuracy
spesific accuracy
spesific accuracy
Meas ement of Map Acc ac Meas ement of Map Acc ac Measurement of Map Accuracy Measurement of Map Accuracy
Site & Non Site Specific Error Site & Non Site Specific Error
Meas ement of Map Acc ac Meas ement of Map Acc ac Measurement of Map Accuracy Measurement of Map Accuracy
Error Matrix : matrik perbandingan image Error Matrix : matrik perbandingan image reference dengan image yang akan
reference dengan image yang akan g g g y g g y g
dianalisa berdasarkan kelompok klasifikasi dianalisa berdasarkan kelompok klasifikasi pixel
pixel--pixel yang sama dalam image pixel yang sama dalam image--image image
pp p p y g y g g g gg
tersebut.
tersebut.Î Î dari Error Matrik dapat dari Error Matrik dapat dihitung % correct,
dihitung % correct, g g , ,
% correct = sum agreement pixel between reff & image(jumlah
% correct = sum agreement pixel between reff & image(jumlah diagoal pada error matrix)/total pixel
diagoal pada error matrix)/total pixel
Meas ement of Map Acc ac Meas ement of Map Acc ac Measurement of Map Accuracy Measurement of Map Accuracy
Error Matrix Error Matrix
Meas ement of Map Acc ac Meas ement of Map Acc ac Measurement of Map Accuracy Measurement of Map Accuracy
Compiling Error matrix Compiling Error matrix
Image direpresentasikan dengan pixel2 Image direpresentasikan dengan pixel2 Image direpresentasikan dengan pixel2 Image direpresentasikan dengan pixel2
Hitung jumlah pixel untuk tiap klasifikasi Hitung jumlah pixel untuk tiap klasifikasi
Yg perlu diperhatikan : Klasifikasi reference Yg perlu diperhatikan : Klasifikasi reference
Yg perlu diperhatikan : Klasifikasi reference Yg perlu diperhatikan : Klasifikasi reference dengan image yg akan diklasifikasi harus dengan image yg akan diklasifikasi harus compatible
compatibleÎ pp Î turunan klasifikasi harus masih turunan klasifikasi harus masih sesuai dengan kategori pada reference
sesuai dengan kategori pada reference
Meas ement of Map Acc ac
Meas ement of Map Acc ac
Measurement of Map Accuracy
Measurement of Map Accuracy
Meas ement of Map Acc ac Meas ement of Map Acc ac Measurement of Map Accuracy Measurement of Map Accuracy
Omission & Commission Error Omission & Commission Error
Omission : jumlah pixel pada reference image yang tidak sesuai Omission : jumlah pixel pada reference image yang tidak sesuai dengan kategori kalsifkasi pada image yg dievaluasi
dengan kategori kalsifkasi pada image yg dievaluasigg gg pp g ygg yg
Commission : jumlah pixel pada image yang dievaluasi yang Commission : jumlah pixel pada image yang dievaluasi yang tidak sesuai dengan keadaan sebanarnya/klasifkasi pada tidak sesuai dengan keadaan sebanarnya/klasifkasi pada reference
reference
CA (Customer Accuracy) & PA (Produsen Accuracy) CA (Customer Accuracy) & PA (Produsen Accuracy)
CA : jumlah pixel pada image yg dievaluasi, yang sesuai dengan CA : jumlah pixel pada image yg dievaluasi, yang sesuai dengan kondisi reference dibandingkan jumlah total pixel pada image yg kondisi reference dibandingkan jumlah total pixel pada image yg
d l k kl f k b
d l k kl f k b
dievaluasi untuk klasifikasi tsb.
dievaluasi untuk klasifikasi tsb.
PA : jumlah pixel pada image yg dievaluasi, yang sesuai dengan PA : jumlah pixel pada image yg dievaluasi, yang sesuai dengan kondisi reference dibandingkan jumlah total pixel pada image kondisi reference dibandingkan jumlah total pixel pada image reference
reference reference.
reference.
Meas ement of Map Acc ac
Meas ement of Map Acc ac
Measurement of Map Accuracy
Measurement of Map Accuracy
http://rst.gsfc.nasa.gov/Sect13/Sect13_3.html
Interpretation of the Error Interpretation of the Error pp
matrix matrix
Percentage Correct (PC) Percentage Correct (PC)
Ukuran yg sering dipakai Ukuran yg sering dipakai
Beberapa rekomendasi : Beberapa rekomendasi :
PC = 85 % dibutuhkan untuk landPC = 85 % dibutuhkan untuk land--use data resource use data resource management (Anderson et al, 1976)
management (Anderson et al, 1976) management (Anderson et al, 1976) management (Anderson et al, 1976)
FitzpatrickFitzpatrick--Lins(1978) akurasi dari USGS landLins(1978) akurasi dari USGS land--cover map cover map
untuk central Atlantic coastal : 85 % (untuk skala 1:24.000), untuk central Atlantic coastal : 85 % (untuk skala 1:24.000), 77% (1:100.000), 73% (1:250.000)
77% (1:100.000), 73% (1:250.000) 77% (1:100.000), 73% (1:250.000) 77% (1:100.000), 73% (1:250.000)
Untuk automasiUntuk automasi--interpretasi dari Landuse menggunakan interpretasi dari Landuse menggunakan hanya data MSS PC yang didapat = 38%, dan untuk MSS + hanya data MSS PC yang didapat = 38%, dan untuk MSS + ancillary data PC = 78 % (Tom et al. 1978)
ancillary data PC = 78 % (Tom et al. 1978) ancillary data PC 78 % (Tom et al. 1978) ancillary data PC 78 % (Tom et al. 1978)
http://rst.gsfc.nasa.gov/Sect13/Sect13_3.html
I t t ti f th E t i
I t t ti f th E t i
Interpretation of the Error matrix Interpretation of the Error matrix
Quantitative Assessment of the Error Matrix Quantitative Assessment of the Error Matrix kappa (k) = measured of difference between
kappa (k) = measured of difference between observed observed tt b t b t t t d d th th t th t t th t agreement
agreement between two map and between two map and the agreement that the agreement that might be attained solely by chance matching two map might be attained solely by chance matching two map..
k = (observed
k = (observed ––expected)/(1 expected)/(1-- expected) expected) Observed = percentage correct
Observed = percentage correct
Expected = product row & column,
Expected = product row & column,Î Î change agreement change agreement two categories when two images superimposed
two categories when two images superimposed (Fig (Fig 14 8)
14 8)
14.8)
14.8)
I t t ti f th E t i
I t t ti f th E t i
Interpretation of the Error matrix
Interpretation of the Error matrix
I t t ti f th E t i
I t t ti f th E t i
Interpretation of the Error matrix Interpretation of the Error matrix
k = 0 83 k = 0 83 Î Î accuracy = 83% better than accuracy = 83% better than
k 0.83 k 0.83 Î Î accuracy 83% better than accuracy 83% better than expected from chance assignment of pixel expected from chance assignment of pixel to cattegories
to cattegories to cattegories.
to cattegories.
k = +1, k = +1,Î Î accuracy = 100%, perfect accuracy = 100%, perfect classification table 14 6
classification table 14 6
classification, table 14.6
classification, table 14.6
Pen t p Pen t p Penutup Penutup
Accuracy dibutuhkan sebagai ukuran informasi yang didapatkan Accuracy dibutuhkan sebagai ukuran informasi yang didapatkan mendekati nilai standar/referensi tertentu/nilai sebenarnya
mendekati nilai standar/referensi tertentu/nilai sebenarnya
Untuk aplikasi tertentu direkomendasikan menggunakan suatu nilai Untuk aplikasi tertentu direkomendasikan menggunakan suatu nilai
Untuk aplikasi tertentu direkomendasikan menggunakan suatu nilai Untuk aplikasi tertentu direkomendasikan menggunakan suatu nilai accuracy tertentu. Selain itu accracy juga berdampak pada nilai accuracy tertentu. Selain itu accracy juga berdampak pada nilai legal dari data dan informasi yang dihasilkan.
legal dari data dan informasi yang dihasilkan.
Accuracy didapatkan dengan membandingkan dengan suatu image Accuracy didapatkan dengan membandingkan dengan suatu image referensi tertentu, yg dianggap benar, lebih akurat dst
referensi tertentu, yg dianggap benar, lebih akurat dst
Untuk mengukur accuracy digunakan alat bantu error matrix Untuk mengukur accuracy digunakan alat bantu error matrix
Untuk mengukur accuracy digunakan alat bantu error matrix, Untuk mengukur accuracy digunakan alat bantu error matrix,
dengan menghitung percentage correct, omission&comission error, dengan menghitung percentage correct, omission&comission error, PA & CA dan kappa, semuanya untuk melihat kerelatifan kebenaran PA & CA dan kappa, semuanya untuk melihat kerelatifan kebenaran klasifkasi yang telah dilakukan.
klasifkasi yang telah dilakukan.