Hardware Level

The figure 1 illustrates the hardware components of the system. The operation on hardware level commences when the radar detects a bird and sends information of the blip including the parameters to the video head steering server. The video head steering server reads the coordinates from the data sent by the radar. The system reacts only if the data has the altitude information of the object.

We have used the PT-1020 Medium Duty video head of the 2B Security Systems. This head has five different speeds of panning: 12, 18, 24, 32 and 48 as well as two of tilting: 12 and 18, in degrees/s. The maximum speed is 1499 m/s (i.e. 48 degree/s) at the distance of 1350 m, where the pilot turbine is located, and the minimum speed is 287 in m/s (i.e. 12 in degree/s) . The maximum ground speed of a flying bird in the test area is estimated to be 110 km/h = 30.5 m/s [3] added with the maximum average wind of 30 m/s [4] resulting in 60.5 m/s. The minimum ground speed of a flying bird in the test area is roughly 6 m/s, which is based on the radar. Soaring gulls and terns are not included in

4 Juha Niemi and Juha T. Tanttu

The too-fast-minimum-speed problem (i.e. the minimum speed of steering the video head is not slow enough in order to track a flying bird at desired distance) is solved as follows: the video head steering system calculates the most probable trajectory of the target bird and adds a lead to the calculated position so that when the head will be driven to that final position it would be ahead of the target bird. The calculations are based on the flight speed (measured by the radar) of a target and the right-angled triangle and they ignore any possible sudden deviations off the flight path. The curvature of the horizon is also ignored due to the relatively short distances in question. When the final position is reached the head stops. This solves the minimum speed problem and also maximizes the probability that the target bird is within the frame of the camera at distance in question, see the Fig. 1. After the head is driven to the final position the camera takes series of images and sends them to the classification software that runs on the same server.

Radar. We use a radar system supplied by Robin Radar Systems B.V. because they provide an avian radar system that is able to detect birds. They also have tracker algorithms for tracking a detected object over time i.e. between the blips. The model we use is the ROBIN 3D FLEX v1.6.3 and it is actually a combination of two radars and a software package for implementation of various algorithms such as the tracker algorithms. The two radars are: a horizontal scanning S-band radar, and a 3D tracking frequency-modulated continuous wave (FMCW) vertical X-band radar. The 3D tracking FMCW radar supports 2-axis scanning mode for 3D coverage and tracking mode with either manual or automatic track selection. The ROBIN 3D FLEX v1.6.3 radar system is capable to provide parameters, such as velocity and bearing of the detected object, for our system. The S-band radar enables a long range detection up to 10 km of flying birds and it provides the longitude and the latitude of a target bird. The X-band radar enables higher resolution i.e. it can detect smaller objects such as small birds up to distance of 5 km. The X-band radar also provides the altitude of the target [5,6,7].

The S-band radar can operate the whole 360 degrees but it is adjusted to operate 180 degrees in the test site since the objects are always in the sector of south via west to north. The X-band radar operates roughly in 20 degree sector that can be configured to lie in a constant position or it can be configured to multiplex between two separate positions. The two possible operation modes of the X-band radar are presented in the Fig. 2. and the Fig. 3., respectively. The ability to multiplex is important because it enables to adjust the minimum dis-tance to the wind turbine of the approaching bird and therefore it gives sufficient time to shut down the wind turbine. The cross section of the two radar beams is the key feature since the intersection of this is the area from where all the three coordinates (i.e. latitude, longitude and altitude) are available [5,6,7]. The

Automatic Bird Identification 5

Fig. 2.The vertical X-band radar operation at a constant position.

Camera System. The resolution of the camera sensor measured by the total number of pixels and the focal length of the lens are important qualities because of the long distance to birds of which images are to be taken. We use a term

6 Juha Niemi and Juha T. Tanttu

called the effective number of pixels (ENP) defined by the number of pixels representing a bird. The remaining number of pixels are considered noise. Image classification is achieved only with ENP as birds will be very small (i.e. they consist of only a small number of pixels) in the images. ENP depends on the focal length of the lens and can be increased by choosing a long (i.e. in terms of focal length) telephoto lens. Because of these facts we have chosen to use the Canon EOS 7D II camera with 20.2-megapixels sensor and the Canon EF 500/f4 IS lens. Correct focusing of the images relies on the autofocus system of the lens and the camera. Automatic exposure is also applied. The operation of the proposed system is not restricted to this combination of the camera and the lens but a combination of any standard DSLR camera with any standard lens suitable for that camera can be utilized.