Figure 4.(a) Figure 5(b) Figure 6(c)
(a)ISODATA Classification Method Land Use Classification Map
(b)Mahalanobis Distance Classification Method Land Use Classification Map
(c) Maximum Likelihood Classification Method Land Use Classification Map
B. Land Use / Cover Classification Accuracy Comparision
1) Random Sample Size
When accuracy for land types’ evaluation, the number of actual reference validated samples is an important considered factor and the number of random samples directly impacts on the accuracy. Recommended to use the N samples which based on the capacity of binomial probability theory to evaluat the accuracy of land use classified map[9]. The computational formula of sample size N as follows:
Z2pq
N =
E2
p stands for the accuracy percentage of the whole image, q = 100-p, E is the allowable error, Z = 2 indicates the normal standard error of 1.96, 95% is the bilateral confidence. By the formula, the lower the expected accuracy, the greater the allowable error and the less number of refered ground simples to estimate the classified accuracy.
The hypothetical samples’ accuracy expectation in this research is 85% and the allowable error is 5%. The points which reliable results needed get at least 203 points. Thus, in the ERDAS IMAGINE 9.1 accuracy evaluation has selected 256 random points to participate in the valuation.
2) Classification Accuracy Statistics
The accuracy evaluation in ERDAS IMAGINE 9.1 is mainly applied in the Classification module which belongs to the accuracy assessment command. After the classified image’s random points come out, comparing the visual interpretation maps and classified images, outputing the classified evaluation reports and according to each evaluation index to analyse the quality of classification results, which include the error matrix evaluation overall accuracy, producer accuracy, user accuracy and Kappa index. There are three types of land use and land cover classification accuracy assessment table.
According to the standard of land use classification, the classification names and the corresponding evaluation results are shown in Table Ⅰ, Table Ⅱand Table Ⅲ.
รูปที่ 4 ( ก ) ( ข ) รูปที่ 5 รูปที่ 6 ( c )
( ) isodata การจำแนกวิธีการการใช้ที่ดินประเภทแผนที่
( b ) mahalanobis ระยะทางการใช้ที่ดินประเภทแผนที่ )
( C ) วิธี Maximum Likelihood การจำแนกประเภทการใช้ที่ดินประเภท B .
แผนที่การใช้ที่ดิน / ครอบคลุมการเปรียบเทียบความแม่นยำในการจำแนก
1 ) สุ่ม ขนาดตัวอย่าง
เมื่อความถูกต้องการประเมินประเภท ' ที่ดิน the number of actual reference validated samples is an important considered factor and the number of random samples directly impacts on the accuracy. Recommended to use the N samples which based on the capacity of binomial probability theory to evaluat the accuracy of land use classified map[9]. The computational formula of sample size N as follows:
Z2pq
N =
E2
p stands for the accuracy percentage of the whole image, q = 100-p, E is the allowable error, Z = 2 indicates the normal standard error of 1.96, 95% is the bilateral confidence. By the formula, the lower the expected accuracy, the greater the allowable error and the less number of refered ground simples to estimate the classified accuracy.
The hypothetical samples’ accuracy expectation in this research is 85% and the allowable error is 5%. The points which reliable results needed get at least 203 points. Thus, in the ERDAS IMAGINE 9.1 accuracy evaluation has selected 256 random points to participate in the valuation.
2) Classification Accuracy Statistics
The accuracy evaluation in ERDAS IMAGINE 9.1 การใช้เป็นหลักในการจำแนกโมดูลซึ่งเป็นของความถูกต้องการประเมินคำสั่ง หลังจากแบ่งภาพเป็นจุดสุ่มออกมา เทียบกับแผนที่การตีความภาพและภาพย่อย outputing จัดประเมินผลและรายงานตามแต่ละประเมินดัชนีวิเคราะห์คุณภาพของผลลัพธ์หมวดหมู่ which include the error matrix evaluation overall accuracy, producer accuracy, user accuracy and Kappa index. There are three types of land use and land cover classification accuracy assessment table.
According to the standard of land use classification, the classification names and the corresponding evaluation results are shown in Table Ⅰ, Table Ⅱand Table Ⅲ.
การแปล กรุณารอสักครู่..
