امکان‌سنجی شناسایی و مطالعۀ آسیب خیابان‌های شهری با استفاده از تصاویر پهپادی و ماهواره‌ای (مطالعۀ موردی: شهر یزد)

راه‌های ارتباطی جزو مهم‌ترین و اصلی‌ترین ساختارهای موجود در سطح یک شهر هستند؛ بنابراین پایش و نگهداری از شبکه‌های ارتباطی شهری و بین‌ شهری همواره جزو موارد پرکاربرد مدیریت شهری است. در پژوهش پیش رو از روش‌های مختلف نظارت‌شده، شیءگرا و فیوژن تصاویر ماهواره‌ای و پهپادی با استفاده از الگوریتم گرام-اشمیت برای بررسی آسیب‌های آسفالت ازجمله ترک‌خوردگی و فرسودگی آسفالت استفاده شده است تا بهترین روش برای تحلیل ارائه شود. نتایج نشان‌دهندۀ آن بود که امکان استخراج آسیب‌های مربوط به آسفالت با تصاویر پهپادی و ماهواره‌ای با استفاده از روش‌های سنجش از دوری وجود دارد. در بررسی روش‌های متفاوت استخراج آسیب‌ها، روش‌های نظارت‌شدۀ ماشین‌بردار پشتیبان با ضریب کاپای 87 و دقت کلی 90 درصد بیشترین و روش کمترین فاصله با ضریب کاپا و دقت کلی به‌ترتیب 57 و 67 درصد کمترین دقت را در طبقه‌بندی روش‌های نظارت‌شده از خود نشان داده‌اند؛ همچنین بین روش‌های شی‌ءگرا، الگوریتم ماشین بردار پشتیبان با ضریب کاپای 86 و دقت کلی 91 درصد خروجی دقیق‌تری نسبت‌به سایر الگوریتم‌های موردمطالعه داشته و کمترین دقت نیز مربوط به الگوریتم نزدیک‌ترین همسایه با ضریب کاپا 78 و دقت کلی 80 درصد بوده است. در خروجی فیوژن پهپاد با سنتینل -2، طبقه‌بندی با بهینه‌ترین الگوریتم بررسی‌شده، ماشین ‌بردار پشتیبان در روش شیءگرا انجام شد که نتایج نشان‌دهندۀ افزایش دقت طبقه‌بندی به ضریب کاپای 91 و دقت کلی 93 درصد بود؛ همچنین روش آستانه‌گزاری با ضریب کاپای 90 درصد نشان‌دهندۀ بهترین نتیجه برای تشخیص فرسودگی آسفالت است. نتایج این پژوهش به‌منظور پایش وضعیت آسفالت‌ راه‌های شهری با هدف افزایش امنیت جاده‌ها و نیز پایداری شهرها و رفاه شهروندی برای سازمان‌های شهرداری‌ و راه‌سازی با صرف هزینه، زمان و نیروی انسانی کمتر مناسب است.

feasibility of identifying and studying the damage of inner-city streets using drone and satellite images (case study: a part of yazd city)

abstractsustainability in the field of pavements is one of the subsets of sustainability topics in sustainable development. in the upcoming research, various supervised and object-oriented methods and fusions of satellite and drone images by using the gram-schmidt algorithm were used to investigate asphalt damage, including asphalt cracking and wear, in order to provide the best method for investigation. the results showed that the supervised methods of support vector machine with a kappa coefficient of 87% and overall accuracy of 90% provided the best and shortest distance method with the kappa coefficient and overall accuracy of 57% and 67%, respectively, while showing the lowest accuracy in the classification of supervised methods. also, among the object-oriented methods, the support vector machine algorithm with a kappa coefficient of 86% and an overall accuracy of 91% had a more accurate output compared to the other studied algorithms. the lowest accuracy was related to the nearest neighbor algorithm with a kappa coefficient of 78% and an overall accuracy of 80%. in the uav fusion output with sentinel-2, the classification was done by using the most optimal algorithm and the support vector machine in the object-oriented method. the results showed an increase in classification accuracy up to  the kappa coefficient of 91% and  overall accuracy of 93%. furthermore, the thresholding method with a kappa coefficient of 90% showed the best result for detecting asphalt wear.keywords: asphalt, remote sensing, drone, classification, road introductionsustainability in an environment highly depends on the sustainability of the components of that environment. sustainability in the field of pavements is also one of the subsets of sustainability topics in sustainable development. ways of communication are among the most important structures in a city; thus, monitoring and maintenance of urban and intercity communication networks are always among the most used cases of urban management. materials methodsin the upcoming research, various supervised classification methods, such as k-nearest neighbors, artificial neural network, support vector machine, maximum likelihood, and minimum distance, and object-oriented classification methods, such as random forest, decision tree, naïve bayes, and fusion of sentinel-2 images and drone images, by using the gram-schmidt algorithm were used to investigate asphalt damage, including asphalt cracking and wear, in order to provide the best method for investigation. also, to study the lifespan of asphalt, first, the spectral profile of different points of asphalt was drawn. then, it was compared with the existing spectral libraries and classified via classification based on the threshold limit. research findingsthe obtained results indicated that in the supervised method, the algorithm of support vector machine achieved the highest accuracy with an overall accuracy of 90% and a kappa coefficient of 87%. similarly, in the section of object-oriented algorithms, the method of support vector machine achieved the best accuracy with an overall accuracy of 91% and a kappa coefficient of 86%. the higher accuracy of the algorithm of support vector machine was probably due to the preparation of a more optimal decision boundary compared to other algorithms. also, the minimum distance method obtained the lowest accuracy among all the classification algorithms. the inaccuracy of the shortest distance algorithm in road classification was proven in the research conducted by li et al. (2020). the results also showed that the most optimal algorithm investigated in the current research, i.e., the support vector machine, in the fused image of the uav with satellite images, increased the overall accuracy and kappa coefficient up to 93% and 91%, respectively. this increase indicated the effective role of fusion in increasing classification accuracy, which was probably the result of merging the higher radiometric power of the sentinel-2 image with an image that had a high spatial resolution of the uav. moreover, the results of this research revealed the higher accuracy of the threshold method with an overall accuracy of 80% and a kappa coefficient of 90% compared to the object-oriented classification of the support vector machine for distinguishing worn asphalt from less worn asphalt.discussion of results conclusionthe results demonstrated that the supervised methods of support vector machine with a kappa coefficient of 87% and overall accuracy of 90% provided the best and shortest distance method with the kappa coefficient and overall accuracy of 57% and 67%, respectively, thus showing the lowest accuracy in the classification of supervised methods. also, among the object-oriented methods, the algorithm of support vector machine with a kappa coefficient of 86% and overall accuracy of 91% had a more accurate output compared to the other studied algorithms, while the lowest accuracy was related to the nearest neighbor algorithm with a kappa coefficient of 78% and an overall accuracy of 80%. in the uav/sentinel-2 data fusion, the classification was done by the most optimized algorithm, which was the support vector machine in the object-oriented method and the results showed an increase in classification accuracy with the kappa coefficient of 91% and  overall accuracy of 93%. mansourmoghaddam et al. (2022) also proved an increase in classification accuracy after fusing the landsat-8 image with images of higher spatial resolution. also, the thresholding method with a kappa coefficient of 90% showed the best result for detecting asphalt wear. the efficiency of using the the threshold method to separate asphalt into two defined classes was proven in the research conducted in this field by kim et al. (2012) and amelian et al. (2014) as well.