A Case Study on Green Areas Change-Detection in Baghdad Using Artificial Intelligence

Experts predict that the size of urban areas will rise by around three times between the years 2000 and 2030 [1]. It is well documented that the "structures of our cities" have a major impact on the occurrence of severe or extreme weather in the surrounding regional environment [2]. In rapidly expanding cities, especially in developing nations, these urban morphologies are characterized by impermeable surfaces, rapid loss of green area, and habitat fragmentation [3]. Green areas in cities, known collectively as urban forests, help lessen regional and local storm-related flooding and water pollution [4], improve air and water quality, moderate temperature, and promote nutrient cycling in soil, all while sequestering carbon [5]. So, Massive land transformations in urban areas—from green to concrete – result in an ever-increasing number of impermeable surfaces, resulting in an unnatural environment [6]. The green area (or leaf area) is frequently used to measure the ratio of green content in urban areas [7]. laying a scientific basis for optimizing green areas. The rates at which green spaces are planned within the spatial space of the urban environment depend on the local conditions of each city, and each of its residents is allocated a specific area of green areas; these rates vary from one country to another. For example, the percentage of green areas in German cities is 37%, in England it is 26% [8], and the American National Organization for Recreation and Parks considers that 25% is a minimum for modern cities [9], while 17.5% of land in Iraqi cities has been allocated for green spaces that have begun to shrink as a result of urban sprawl to fill them [10]. AI techniques are now becoming increasingly essential in categorization challenges [11]. Given the significance of the green cover within cities, the researcher examined this topic by comparing the green area in 2008 to the existing green area in 2022 [12].

To determine the green area, artificial intelligence techniques were employed to convert the green area from the satellite imagery [13]. Machine learning is insignificantly exploited for environmental needs. So should focus on AI to fight negative environmental impact as a major step to fight climate change [14]. The process involves converting the image's color format from RGB to HSV in order to exploit the image's color properties. This process extracts the image's green area features, as HSV is a cylinder color model that converts the RGB primary colors into aspects that are more intuitive to humans [15]. These contain hue, saturation, and value, similar to the Munsell Color System. After that the range of selected color shaped into a mask, the green area percentages are calculated and the necessary analyses are conducted. The accuracy of the study's calculations for the green area and the percentages for each year reached 92%, which is very remarkable.

Adhamiya is one of the oldest areas of Baghdad. The Adhamiya region first appeared during the Abbasid dynasty. The city was named after the Imam who was buried there, Abu Hanifa al-Numan. It is situated on the east bank of the Tigris River, about to the north of Baghdad. Figure 1 shows the impact on the Adhamiya area of just the loss of green areas and the replacement of gardens with homes.

1.png

Figure 1. Field survey of the case study area, Adhamiya, performed by the authors [17]

This section explains the dataset utilized as well as the method used to construct the framework. The dataset used in the paper was created with the Adhamiya neighborhood Satellite images. The process is summarized by converting the images from RGB to HSV, then mask generation, and finally calculating the percentage of the green areas.

3.1 Dataset

The dataset utilized in this study was developed by the researcher particularly for this purpose; it consists of 208 high-resolution satellite images of the Adhamiya neighborhood in Baghdad. For accurate change detection, machine learning techniques such as Convolutional Neural Network (CNN) can be utilized to relieve the difficulty [24].

2.png

Figure 2. Baghdad-Adhamiya area map history. Shows a decline in green space between 2008 and 2022

Table 1. Dataset images specifications

It comprises of 104 images of various parts of the region taken in 2008, and the same number of images were taken in the same locations in 2022. Figure 2 depicts a map of Baghdad-Adhamiya area map history. Shows a decline in green space between 2008 and 2022, highlighting the selected portions as datasets. Table 1 show the images specifications used in dataset.

3.2 Methodology

This method relies heavily on the inventory of the color spectrum, and it is vital to follow these stages to achieve effective results. Figure 3 depicts the method's steps:

3.png

Figure 3. Methodology workflow

1. Reading all of the images is the initial stage of the approach.

2. Convert RGB image to HSV image:

1. Hue specifies the color's angle on the RGB color circle. A hue of 0 degrees is red, 120 degrees is green, and 240 degrees is blue.

2. Saturation determines how much color is utilized. A color with a saturation of 100 percent is the clearest color conceivable, while a saturation of 0 percent produces grayscale.

3. Value regulates the color's luminance. A color with zero percent brightness is pure black, while a color with one hundred percent brightness contains no black. The HSV color model is also referred to as HSB, including in P5.js, because this dimension is frequently referred to as brightness.

The conversion is done from RGB to HSV, because as we learned that HSV has to deal with gradients separately, and it is the best way to get a gradient to deal with it, in order to study the gradients, we need to make the threshold. Figure 4 shows the RGB to HSV example. This ensures that you get the best possible result. HSV ranges for the year (2022) is:

$(\mathrm{H} \rightarrow 70-105, \mathrm{~S} \rightarrow 25-255, \mathrm{~V} \rightarrow 25-255)$

and for (2008) is:

$(\mathrm{H} \rightarrow 33-75, \mathrm{~S} \rightarrow 25-255, \mathrm{~V} \rightarrow 25-255)$

3. Mask Generation: The second stage of the image analysis procedure started when the range of the green areas' tonalities was identified after studying them for both years. This range was then turned into a mask. The input and source image are the converted image and the inputs are two values, with the lowest value representing the lowest gradient and the greatest value representing the widest color spectrum. In the Figure 5, it clear that the mask is made of two colors: black and white. The white parts of the mask represent the green parts, while the black parts represent the rest of the ingredients.

4. Percentage Calculating: After making a mask, it turns out that the white is the green area. To find the percentage of green areas in each image, we divide the number of white points in the mask by the total number of pixels. It is easier to find similarities and differences between the two years. The percentage of green area Figure 6 is calculated by the equation:

Green Percentage = White area / Size of image          (1)

where:

• White area represents the green areas;

• Size of image is 2000 * 2000 px.

4.png

Figure 4. RGB to HSV example (a) 2022 (b) 2008

5.png

Figure 5. Mask generation example (a) 2022 (b) 2008

6.png

Figure 6. Green area percentage (a) 2022 (b) 2008

The following conclusions were reached as a result of the preceding information and experiments:

1. The method employed in this study greatly reduced the time required to generate masks.

2. The method assisted in determining the proportion of green areas and comparing them between the two years, so enabling numerous studies to be conducted on this basis.

3. By analysing numerous temperature scales, it was shown that the percentage of green regions is inversely proportional to the temperature, since locations with green areas have a greater temperature than those with few or no green areas Figure 7 and 8.

4. The superiority of the results is due to the fact that the image in 2022 was obtained with newer and more advanced equipment; consequently, the resolution is better and the detecting technique is more exact and efficient.

5. The housing problem played a significant role in Baghdad's loss of green spaces, particularly in the Adhamiya region, extensive buildings with large gardens vanished in the lack of a serious plan to address the situation, as a result of the housing crisis and rising real estate costs, houses were divided and their gardens were converted into small houses without trees.

7.png

Figure 7. Green area percentage chart

8.png

Figure 8. Green area average percentage change

Baghdad has seen a significant increase of green spaces since the 1930s, beginning with public gardens and parks and progressing through squares and planting walkways and open spaces overlapping between residential districts. And increasing the aesthetic value of the city, until this growth began to veer in the opposite direction since the ninth decade of the twentieth century, when the delay began in completing the implementation of development projects plans for the development of the city, particularly with regard to green spaces, while the period that followed the year 2008 was characterized by stagnation in the expansion of green spaces due to what The country witnessed turbulent conditions, weak oversimplification. All of this encouraged citizen to encroach on green spaces and change their use at random for purposes in an attempt to solve the problem of increasing population growth, rapid and uncontrolled urban expansion, which damaged the city's green spaces and lost an important aesthetic element, and caused an imbalance in the local environment's balance that it maintained. For decades, it has had a deleterious impact on the local climate.

As a result, it is necessary to pay attention to the expansion of green spaces within the city and work to remove and control the slums that occupy them, employing advanced artificial intelligence techniques that reduce the effort and time required for urban monitoring of the city while enabling adherence to the planned basic design.