Abstract:
هدف اصلی از این تحقیق پایش تغییرات کاربری اراضی دشت خانمیرزا با استفاده از الگوریتمهای مختلف است که از تصاویر ماهوارة لندست 5، 7، و 8 و سنجندههای TM، ETM، و OLI برای سه دورة 1996، 2006، و 2016 استخراج شد و نقشة کاربری اراضی دشت با استفاده از چهار الگوریتم حداکثر احتمال، شبکة عصبی مصنوعی، حداقل فاصله، و فاصلة ماهالانویی با استفاده از ضریب کاپا ارزیابی شد. نتایج حاصل از ارزیابی دقت این دو روش با استفاده از تعیین ضریب کاپا نشان داد الگوریتم شبکة عصبی مصنوعی نسبت به الگوریتم حداکثر احتمال با ضریب از دقت بیشتری برخوردار است. همچنین، دو الگوریتم شبکة عصبی مصنوعی و حداکثر احتمال با دقت کلی 29/90 و 79/86 در شش کلاس کاربری (کشاورزی، مرتع، مسکونی، اراضی سنگی و لخت، باغ، و اراضی پست نمدار) طبقهبندی شد. تجزیهوتحلیل حاصل از تغییرات نشان داد کاربریهای کشاورزی و مسکونی روند افزایشی داشتهاند؛ بهطوریکه میزان این افزایش بهترتیب برابر با 5/62 و 5/3درصد بوده است و از اراضی پست نمدار، مراتع، و اراضی سنگی و لخت کاسته است. بیشترین تغییر کاربریها مربوط به تبدیل کاربری اراضی سنگی و لخت به کاربری کشاورزی است که 1673 هکتار از اراضی سنگی و لخت در سال 2006 به اراضی کشاورزی در سال 2016 تبدیل شده است. از دیگر تغییر کاربریهای مشهود در منطقه تغییر کاربری اراضی سنگی و لخت و مراتع به اراضی مسکونی است؛ بهطوریکه 7/65 هکتار از اراضی سنگی و لخت و 8/40 هکتار از اراضی مرتع به کاربری مسکونی تبدیل شده است.
Accurate and real time information on land use and land cover and their changes is very important in urban management decisions, ecosystem monitoring and urban planning. In recent decades, widespread changes in land use of the Khan mirza plain as one of the northern Karun watersheds have occurred, that need to monitoring these changes.In this study, Landsat 5, 7 and 8 satellite images and TM, ETM, and OLI sensors for the period of 1996, 2006, and 2016 were used to produce of land use and land cover map of Khan mirza plain by four methods: maximum likelihood, artificial neural network, minimum distance and Mahalanobis distance and theirs Kappa coefficient were evaluated.The results of the evaluation of the accuracy of these two methods by using Kappa coefficients have shown that the artificial neural network algorithm is more accurate than the maximum likelihood algorithm. Also, by results of two algorithms of artificial neural network and maximum likelihood with an overall accuracy of 90.29 and 86.79, all of land cover maps were classified in six classes (agriculture, rangeland, residential area, rocky and bare lands, gardens and flatlands).The analysis of the classifications showed that agricultural and residential classes had a rising trend, 62.5% and 3.5%, respectively, and rangeland, rocky and bare lands and flatlands were decreased.The largest change is related to the conversion of rocky and bare lands class to the agricultural class, which 1673 hectares of rocky and bare lands in 2006 changes into agricultural lands in 2016. Another obvious land use change in this area, are change of rangelands into residential areas, which 40.8 ha of rangelands changed into residential area.In overall, this research showed that the best way to produce of land use map in the study area is to use artificial neural network algorithm. According to the results, it is suggested using this method to produce of land use change map for this region.Accurate and real time information on land use and land cover and their changes is very important in urban management decisions, ecosystem monitoring and urban planning. In recent decades, widespread changes in land use of the Khan mirza plain as one of the northern Karun watersheds have occurred, that need to monitoring these changes.In this study, Landsat 5, 7 and 8 satellite images and TM, ETM, and OLI sensors for the period of 1996, 2006, and 2016 were used to produce of land use and land cover map of Khan mirza plain by four methods: maximum likelihood, artificial neural network, minimum distance and Mahalanobis distance and theirs Kappa coefficient were evaluated.The results of the evaluation of the accuracy of these two methods by using Kappa coefficients have shown that the artificial neural network algorithm is more accurate than the maximum likelihood algorithm. Also, by results of two algorithms of artificial neural network and maximum likelihood with an overall accuracy of 90.29 and 86.79, all of land cover maps were classified in six classes (agriculture, rangeland, residential area, rocky and bare lands, gardens and flatlands).The analysis of the classifications showed that agricultural and residential classes had a rising trend, 62.5% and 3.5%, respectively, and rangeland, rocky and bare lands and flatlands were decreased.The largest change is related to the conversion of rocky and bare lands class to the agricultural class, which 1673 hectares of rocky and bare lands in 2006 changes into agricultural lands in 2016. Another obvious land use change in this area, are change of rangelands into residential areas, which 40.8 ha of rangelands changed into residential area.In overall, this research showed that the best way to produce of land use map in the study area is to use artificial neural network algorithm. According to the results, it is suggested using this method to produce of land use change map for this region.Accurate and real time information on land use and land cover and their changes is very important in urban management decisions, ecosystem monitoring and urban planning. In recent decades, widespread changes in land use of the Khan mirza plain as one of the northern Karun watersheds have occurred, that need to monitoring these changes.In this study, Landsat 5, 7 and 8 satellite images and TM, ETM, and OLI sensors for the period of 1996, 2006, and 2016 were used to produce of land use and land cover map of Khan mirza plain by four methods: maximum likelihood, artificial neural network, minimum distance and Mahalanobis distance and theirs Kappa coefficient were evaluated.The results of the evaluation of the accuracy of these two methods by using Kappa coefficients have shown that the artificial neural network algorithm is more accurate than the maximum likelihood algorithm. Also, by results of two algorithms of artificial neural network and maximum likelihood with an overall accuracy of 90.29 and 86.79, all of land cover maps were classified in six classes (agriculture, rangeland, residential area, rocky and bare lands, gardens and flatlands).The analysis of the classifications showed that agricultural and residential classes had a rising trend, 62.5% and 3.5%, respectively, and rangeland, rocky and bare lands and flatlands were decreased.The largest change is related to the conversion of rocky and bare lands class to the agricultural class, which 1673 hectares of rocky and bare lands in 2006 changes into agricultural lands in 2016. Another obvious land use change in this area, are change of rangelands into residential areas, which 40.8 ha of rangelands changed into residential area.In overall, this research showed that the best way to produce of land use map in the study area is to use artificial neural network algorithm. According to the results, it is suggested using this method to produce of land use change map for this region.
Machine summary:
در چند دهۀ اخير مطالعات متعددي در زمينۀ به کارگيري سنجش از دور و تصاوير ماهواره اي براي استخراج نقشه هاي کاربري اراضي و بررسي تغييرات ايجادشده در پوشش زمين در داخل و خارج از کشور انجام گرفته و مدل هاي به کاربرده شده در تحقيقات يادشده ، با توجه به نوع الگوريتم هاي مورداستفاده ، نتايج متفاوتي داشته که به چند مورد از آن ها اشاره ميشود.
هدف از اين پژوهش پايش تغييرات کاربري اراضي دشت خان ميرزا با استفاده از چهار روش حداکثر احتمال ٢، شبکۀ عصبي مصنوعي ٣، حداقل فاصله ٤، و فاصلۀ ماهالانويي ٥ با استفاده از تصاوير چندزمانه و چندطيفي ماهوارة لندست سال هاي ١٩٩٦، ٢٠٠٦، و ٢٠١٦ است .
Mahalanobis Distance در اين مطالعه به منظور تهيۀ نقشۀ تغييرات کاربري اراضي از چهار روش (حداکثر احتمال ، شبکۀ عصبي مصنوعي ، حداقل فاصله ، و فاصلۀ ماهالانويي) در دشت خان ميرزا و از تصاوير لندست ٥، ٧، و ٨ در سه دورة زماني (١٩٩٦، ٢٠٠٦، و ٢٠١٦) استفاده شده است .
Monitoring of land use change in Kashan Plain using remote sensing data, Zamin pouya International journal, Vol. 4, N.
Kumar, 2018, Land use classification and change detection by using multi-temporal remotely sensed imagery: The case of Chunati wildlife sanctuary, Bangladesh, The Egyptian Journal of Remote Sensing and Space Science, Vol. 21, No. 1, PP.
, 2010, Land use and land cover change detection through remote sensing approach: A case study of Kodaikanal taluk, Tamil nadu, International Journal of Geomatics and Geosciences, Vol. 1, No. 2.