Limits...
Robust vehicle detection under various environments to realize road traffic flow surveillance using an infrared thermal camera.

Iwasaki Y, Misumi M, Nakamiya T - ScientificWorldJournal (2015)

Bottom Line: However, the first method decreases the vehicle detection accuracy in winter season.The second method did not achieve high detection accuracy for vehicles on left-hand and right-hand lanes except for two center-lanes.Therefore, by combining the first method and the new method, high vehicle detection accuracies are maintained under various environments, and road traffic flow surveillance can be realized.

View Article: PubMed Central - PubMed

Affiliation: Faculty of Industrial and Welfare Engineering, Tokai University, 9-1-1 Toroku, Higashi-ku, Kumamoto 862-8652, Japan.

ABSTRACT
To realize road traffic flow surveillance under various environments which contain poor visibility conditions, we have already proposed two vehicle detection methods using thermal images taken with an infrared thermal camera. The first method uses pattern recognition for the windshields and their surroundings to detect vehicles. However, the first method decreases the vehicle detection accuracy in winter season. To maintain high vehicle detection accuracy in all seasons, we developed the second method. The second method uses tires' thermal energy reflection areas on a road as the detection targets. The second method did not achieve high detection accuracy for vehicles on left-hand and right-hand lanes except for two center-lanes. Therefore, we have developed a new method based on the second method to increase the vehicle detection accuracy. This paper proposes the new method and shows that the detection accuracy for vehicles on all lanes is 92.1%. Therefore, by combining the first method and the new method, high vehicle detection accuracies are maintained under various environments, and road traffic flow surveillance can be realized.

No MeSH data available.


An example of four ROIs which are indicated in green lines.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4339787&req=5

fig10: An example of four ROIs which are indicated in green lines.

Mentions: (9) Gamma transformation with γ is employed for the image after histogram equalization. Traffic image we used has four lanes. The image has four ROIs (Regions of Interest) to detect vehicles on four lanes. The four vertices of the rectangle measurement area (ROI) on each lane can be freely determined with mouse clicks in accordance with the road shape. Figure 10 shows an example of four ROIs which are indicated in green lines. Traffic volume and occupancy affect warming up degrees on lanes. The mean of temperature on each lane is different because traffic volume and occupancy in each lane are different. In order to emphasize effectively the tires' thermal energy reflection areas from the road surface, the optimum γ value of gamma transformation for each lane is not same. This algorithm chooses 0.1, 0.2, 0.3, or 0.4 as γ value for each lane. The process of choosing γ value for each lane is described below. Four images are obtained after gamma transformation with four different γ values. We call the four images as “IMG-A” in this section. Pixel values in a narrow area near the left or right edge are extracted from IMG-A. The distance between the x-coordinate of the narrow area and that of edge is shown in Figure 11. The distance depends on y-coordinate. The distance between points A and B is [0.02y + 1] and that between points A and C is [0.07y + 2]. These geometric parameters [0.02y + 1] and [0.07y + 2] are affected by the camera angle, the camera height, and the road shape. We decide these parameters by an image processing experiment in advance. The means of pixel values in the narrow areas are calculated. We assume that high pixel values are over 80% of the means, and pixel values except for high pixel values are low pixel values. This algorithm chooses the γ value when the ratio of the number of pixels with high pixel values to that of pixels with low pixel values is closest to 0.5. Processes (1)–(9) are repeated 60 times to decide the final optimum γ value for each lane. Most selected γ value for each lane in 60 times is decided as the final optimum γ value. Vehicle detection is not done while this procedure is continued. Therefore, this procedure is usually employed when traffic lights are red. Figure 12 shows an image in lane 2 after final optimum gamma transformation.


Robust vehicle detection under various environments to realize road traffic flow surveillance using an infrared thermal camera.

Iwasaki Y, Misumi M, Nakamiya T - ScientificWorldJournal (2015)

An example of four ROIs which are indicated in green lines.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4339787&req=5

fig10: An example of four ROIs which are indicated in green lines.
Mentions: (9) Gamma transformation with γ is employed for the image after histogram equalization. Traffic image we used has four lanes. The image has four ROIs (Regions of Interest) to detect vehicles on four lanes. The four vertices of the rectangle measurement area (ROI) on each lane can be freely determined with mouse clicks in accordance with the road shape. Figure 10 shows an example of four ROIs which are indicated in green lines. Traffic volume and occupancy affect warming up degrees on lanes. The mean of temperature on each lane is different because traffic volume and occupancy in each lane are different. In order to emphasize effectively the tires' thermal energy reflection areas from the road surface, the optimum γ value of gamma transformation for each lane is not same. This algorithm chooses 0.1, 0.2, 0.3, or 0.4 as γ value for each lane. The process of choosing γ value for each lane is described below. Four images are obtained after gamma transformation with four different γ values. We call the four images as “IMG-A” in this section. Pixel values in a narrow area near the left or right edge are extracted from IMG-A. The distance between the x-coordinate of the narrow area and that of edge is shown in Figure 11. The distance depends on y-coordinate. The distance between points A and B is [0.02y + 1] and that between points A and C is [0.07y + 2]. These geometric parameters [0.02y + 1] and [0.07y + 2] are affected by the camera angle, the camera height, and the road shape. We decide these parameters by an image processing experiment in advance. The means of pixel values in the narrow areas are calculated. We assume that high pixel values are over 80% of the means, and pixel values except for high pixel values are low pixel values. This algorithm chooses the γ value when the ratio of the number of pixels with high pixel values to that of pixels with low pixel values is closest to 0.5. Processes (1)–(9) are repeated 60 times to decide the final optimum γ value for each lane. Most selected γ value for each lane in 60 times is decided as the final optimum γ value. Vehicle detection is not done while this procedure is continued. Therefore, this procedure is usually employed when traffic lights are red. Figure 12 shows an image in lane 2 after final optimum gamma transformation.

Bottom Line: However, the first method decreases the vehicle detection accuracy in winter season.The second method did not achieve high detection accuracy for vehicles on left-hand and right-hand lanes except for two center-lanes.Therefore, by combining the first method and the new method, high vehicle detection accuracies are maintained under various environments, and road traffic flow surveillance can be realized.

View Article: PubMed Central - PubMed

Affiliation: Faculty of Industrial and Welfare Engineering, Tokai University, 9-1-1 Toroku, Higashi-ku, Kumamoto 862-8652, Japan.

ABSTRACT
To realize road traffic flow surveillance under various environments which contain poor visibility conditions, we have already proposed two vehicle detection methods using thermal images taken with an infrared thermal camera. The first method uses pattern recognition for the windshields and their surroundings to detect vehicles. However, the first method decreases the vehicle detection accuracy in winter season. To maintain high vehicle detection accuracy in all seasons, we developed the second method. The second method uses tires' thermal energy reflection areas on a road as the detection targets. The second method did not achieve high detection accuracy for vehicles on left-hand and right-hand lanes except for two center-lanes. Therefore, we have developed a new method based on the second method to increase the vehicle detection accuracy. This paper proposes the new method and shows that the detection accuracy for vehicles on all lanes is 92.1%. Therefore, by combining the first method and the new method, high vehicle detection accuracies are maintained under various environments, and road traffic flow surveillance can be realized.

No MeSH data available.