Overview

• Increase frequency of image capture of wired video cameras in nestboxes to increase usefulness for biological applications.
• Collect data necessary to analyze and improve functionality of cyclops as biological sensors for avian studies. Expand existing system.
• Design and test cyclops system for use in herpetological pitfall sampling methodology.
• Utilize embedded network sensing of environmental variables, and networked video-based remote sensing approaches, determine microclimatic influences on nesting activity and nest success in secondary cavity nesting avifauna.
• Collect the data necessary to test video content analysis software approaches for automated classification of avian behavioral activities from nest box video images in real time at remote nest sites.

Approach

Avian Studies Avian senor studies being conducted at the James Reserve focus on species of birds that typically nest in tree cavities ("cavity nesters"). These species often occupy human-constructed nesting boxes when they are made available, and numerous avian research studies utilize data collected in nestboxes due to the ease with which the nest contents can be viewed. Currently, our work is focused on recording still images inside the nestboxes (using either wired or wireless camera systems) to record bird behavior during the breeding cycle. In addition, we are measuring environmental characteristics in the immediate nesting environment (i.e., the nestbox) including temperature, humidity, and dew point, as well as near the nesting environment (i.e.,outside of the nestbox). Light intensity (Photosynthetically Active Radiation, PAR) and soil moisture content are also measured near the nestboxes. The environmental data and associated nestbox images are being used to answer questions about bird breeding behavior and breeding success. The questions fall into the following areas.

Nest Site Selection. Selection of nest sites is an important area of study in avian ecology because of the consequences that the properties of the nest site have to the successful rearing of young. Poor nest sites and/or poor territories may lead to reduced success at reproduction. Photographs by cameras inside the nestbox allow us to determine what species are choosing which nestboxes, including whether multiple species "investigate" a box before it is occupied continuously by a single breeding pair. We can also establish the date that breeding is initiated overall, as well as when specific stages of the breeding cycle occur. Environmental characteristics inside and outside of the boxes can be correlated with occupancy to give an indicator of what environmental conditions are selected by the birds (e.g., test whether there are similarities among boxes that are selected for nesting versus those that are not). For example, we are able to evaluate whether boxes that are chosen for nesting are colder, warmer, or not significantly different in temperature from boxes that are not chosen. If birds are selecting boxes with particular temperature conditions, we would expect to see a significant difference between selected and non-selected nest boxes in some aspect of temperature (e.g. daily minimum temperature, daily maximum temperature, daily mean temperature). The continuous monitoring capability of CENS technology enables us to test hypotheses relating to nest site selection.

Microclimatic Influences On Adult Breeding Behavior And Nest Success. Environmental sensors in occupied nestboxes provide information about various characteristics of the nesting environment experienced by the birds using them. This information alone can be used to evaluate the relationship between environmental factors and reproductive success. Reproductive success can be measured in a variety of ways, including: (1) the number of eggs hatched; (2) whether a reproductive attempt produces any young to the age at which they are independent from the nest (called "fledging"); and/or (3) the precise number of young that survive to fledging. Environmental data (e.g., temperature) can be used in conjunction with photographic images within the nestbox to evaluate whether relationships exist between environmental conditions and bird behavior (e.g., incubation patterns and feeding rates), which can vary among breeding adult birds. Variation in environmental characteristics among boxes results in part from differences in where boxes occur in the landscape. For example, a nestbox in a relatively shady area will experience colder temperatures. Since incubating adult birds spend a substantial percentage of time on the nest keeping eggs warm so that embryonic development proceeds, a parent in a relatively cold box may spend more time incubating eggs. Similar relationships are expected to exist with respect to the investment of the female bird in keeping the nestlings warm soon after hatching ("brooding"), and/or the rates at which adults bring food to the nestlings in the box.

Nest predation. Nest predation is one cause of nesting failure. However, in typical nesting studies where field workers check the nest every several days, it can be difficult to ascertain late in the nesting cycle whether a nest has been lost to a predator, or whether the young fledged the nest, since the exact day that the young are developed enough to depart from the nest is variable. This uncertainty can confound estimates of avian productivity. The use of photographic images in nestboxes allows us to determine whether a nest was actually lost to a predator consuming the nestlings, or whether the young departed successfully.

Herpetological Studies

Use of Pitfall Traps to Determine Diversity and Abundance. Reptiles and amphibians can be difficult to sample at a community level due to the size, crypticness, and activity patterns. Recent studies by Ted Case (UCSD) and Robert Fisher (UCSD, USGS), among others, has determined that drift fences in combination with pitfall buckets and funnel traps are the most effective means for sampling herptofauna diversity and abundance while minimizing observer bias.

The standard pitfall array consists of three drift fence arms with 5-gallon buckets buckets at each end and midpoint. One snake funnel trap is located along each drift fence arm. The pitfall buckets are dug into the earth so that their lips are flush with the soil surface. When a herp’s path (and often small mammal or arthropod) intersects with the silt fence, the animal tend to turn aside and travel along the silt fence until it falls into the bucket for later collection by the biologist.

Graphic from Draft Herpetological Monitoring, Using a Pitfall Trapping Design in Southern California, in prep. USGS

Graphic from Draft Herpetological Monitoring, Using a Pitfall Trapping Design in Southern California, in prep. USGS

Herptofauna Sampling at James Reserve. USGS established seven of these pitfall arrays at the James Reserve as part of a greater southern California study of herptofauna in 1995. Sampling was conducted between November 1995 and March 1997. Sampling occurred for a 4-day period and was repeated about every two months and provided the reserve with baseline data on herp diversity and abundance onsite. Additional sampling at all seven arrays is anticipated to begin in 2007.

Added Value of Cylops for Herptofauna Sampling. Although an effective means for sampling herptofauna, there are some drawbacks to pitfall trapping that could be alleviated through use of biological sensors such as Cyclops. These include:

• Time traveling to remote arrays that may have no captures
• Mortality of organisms if not removed soon after capture (environmental conditions, predation by other captives)
• No control for documenting escapees

Utilizing cyclops on the pitfall arrays for herptofauna studies has the benefit of testing their efficacy in an application and microenvironment uniquely different from their usage in nestboxes for avian studies. The pitfall arrays have a low profile and therefore seem more prone to connectivity issues. The pitfall arrays are also more intensely handled and therefore are more prone to physical connections coming loose and misalignment of the view field.

Systems/Experiments

Systems

During this last year we have utilized sensor systems to increase our knowledge of birds, reptiles and amphibians (herptofauna) as part of TEOS. Since 2003 the James Reserve and CENS have utilized wired micro-climate sensor systems and video cameras at nestboxes, which now total 13, to initiate biological studies on avian nesting behavior. Twelve additional nestboxes were outfitted with wireless camera systems (Cyclops) in early 2006 to expand the avian study sample size and field test the cyclops system. An additional expansion of 13 cyclops nestboxes, and outfitting all 25 cyclops nestboxes with sensors, is anticipated for 2007. We also identified the potential use of the cyclops system for herpetological pitfall trap studies and plan to test a prototype array with seven pitfall trap buckets in Summer 2007.

Distribution of nestboxes and associated cameras and sensors in the James Reserve.

We discuss the wired and wireless systems below and their application to ongoing and expanded avian and herpetological studies.

Wired Cameras in Nestboxes. As discussed in detail in the 2003 and later Annual Reports, we have [JS1]13 nestboxes outfitted with video cameras in the interior of the box, which transmit images (480 x 700 pixels) via hardwire to our axis video servers and to our website. Each nestbox is outfitted with a Hobo Microstation datalogger and collects information on interior box temperature and humidity, exterior box temperature and humidity, PAR, and soil moisture. Images of the interior of the nestbox are captured every 15 minutes between March-October. During October-March images are captured every 30 minutes as nestbox use by birds is limited to infrequent overnight roosting.

Example nestbox with wired camera and microclimate sensors.

During May 2006-April 2007 we have continued to operate this system year-round without any major modifications. We have also sought to minimize data loss associated with weather events (lightning strikes, wind storms) and streamline the data upload process for review and use by researchers and the public on the James Reserve webpage. We have also identified that increasing the frequency of image capture is likely to increase the value of wired nestboxes for answering questions about behavior. For example, images at 15-minute intervals provide information on nesting initiation, stages of juvenile development (hatching, fledging), while increased image capture frequency would allow observation of finer scale activities (nest competition, feeding behaviors and duration).

Wireless Cyclops. Cyclops is a low-power imaging platform that was designed by UCLA CENS in collaboration with Agilent Technologies. This technology demonstrated the feasibility of imaging for wireless sensor networks, which are generally constrained by power and bandwidth. Cyclops couples with Mote class devices to acquire and process reduced sized images over a wireless channel. Currently Cyclops is commercially available (http://cyclopscamera.com/) and in use by a number of external research groups. One of the greatest benefits of using Cyclops is the flexibility of placing them in locations lacking wired infrastructure. To further demonstrate the practicality of such a wireless imaging system as a scientific instrument, James Reserve was chosen as deployment site. James Reserve was primarily chosen because it not only lacks wired infrastructure in most locations but that it would allow us to compare our system to a pre-existing wired camera system as described in the previous section.

A typical nestbox setup with a cyclops/mote unit.

As discussed in the 2006 Annual Report, Cyclops/mote units were installed in a total of 12 nestboxes. However due to connectivity issues, we were able to obtain images from only 11. Image collection was initiated in early May and lasted until August, corresponding to the nesting cycle of most birds at the Reserve.

Our image collection system worked in the following way:

1. Each Mote was queried by its respective microserver every 15 minutes.
2. Upon receiving the query, the Mote would signal the Cyclops to capture an image while illuminating the box with an infrared (IR) LED.
3. The image would be sent back to the microserver via the Mote.
4. Upon receiving images from all the Mote/Cyclops pairs, the microserver would proceed to upload them to an external server.
5. The external server was responsible for uploading the images to an external database called SensorBase as well hosting a web front-end used to browse the images.

We organized all wireless camera data so viewable by day and multi-day periods on a webased GUI as shown below.

View of images captured for avian studies.

Biological Research

Avian Studies. The James Reserve currently has 53 nestboxes scattered throughout the reserve. This includes 13 with wired cameras and sensors, 12 with wireless cyclops, and 28 without cameras or sensors). The vast majority were designed with study of the Western Bluebird (WEBL) in mind and have a 1.5-inch diameter entry hole. We are continuing to target this and other species that have historically occupied nestboxes onsite. The predominant species include WEBL and Violet-green Swallows (VGSW). Smaller species such as the Mountain Chickadee (MOCH) can also enter. The greatest occupancy to date at James Reserve over all nestboxes is by VGSW and WEBL, and we expect that in the short-term occupancy will continue to be greatest for these two species.

Analyzing Existing Data. In 2006, six of the 12 nestboxes with wireless Cyclops cameras were occupied by birds that initiated nest building (3 by VGSW, two by WEBL, and one by MOCH. Of these, 4 pairs successfully fledged young. For two of these four nests there exist images throughout the entire nesting cycle once nest building was initiated (Box 13, Box 71).

Camera Type	Nest Box #	Year	Bird Species	Breeding Stage Captured by Images
				Roosting/Prebuilding	Nest Building	Egg Laying	Incubation	Nestling	Fledgling
Wireless Cyclops	1	2006	WEBL				+(1-4 day data loss before hatching)	(1-4 day data loss near hatching; 6 day data loss mid-nestling)	+
	13		VGSW		+	+	+ (image quality issue; id of hatch time problematic)	+	+
	33		MOCH		+
	40		None	(some data loss)
	41		None	--MOCH? Unknown woodpecker spp.	uncertain
	42		WEBL		+	+ (2 day data loss at start of laying)	Nest fails
	46		None
	52		None
	53		VGSW		Several days of data loss	Several days of data loss	Several days of data loss	Several days of data loss	+
	71		VGSW		+	+	+	+	+
Wired	3	2004	None
		2005	None	(WEBL, MOCH, NUWO visit)
		2006	WEBL	+	+ (MOCH builds some)	+	+	+	+
	8	2003	VGSW	+	+	+	+	+	Nest fails
		2004	VGSW	+	+	+	+	+	+
		2005	VGSW	+	+	+	+	+	+
		2006	VGSW	+	+	+	+	+	+
Wired, cont'd	11	2004	(technical failure)
		2005	None
		2006	None	(Note: technical failure April 2006)
	14	2004	None
		2005	None	(WEBL & MOCH visit; MOCH lays eggs w/o next 5/16)
		2006	MOCH	+	(abandoned during building or laying; WEBL visits after this)
	21	2003	WEBL	+	+	+	+	+	+
		2004	None	(WEBL visits)
		2005	WEBL	+ (MOCH visits)	+	+	+	+	+
		2006	WEBL	+	+	+	+	+	+
	22	2004	WEBL	(technical failure)	+	+	+	+	+
		2005	WEBL	+	Building aborted
		2006	None	Several species visit; MOCH and WEBL do some buiding?	Building aborted
Wired, cont'd	27	2004	WEBL	(some following technical failure)	Building aborted
		2005	None	(NUWO visits 7/1/2005, presumed post-breeding)
		2006	None	Several species visit; MOCH does some buiding?	Building aborted
	31	2003	WEBL	+ (MOCH visits)	+	+	+	+	+
		2004	WEBL	+	+	+	+	+	+
		2005	WEBL	+ (MOCH visits)	+	+	+ (some eggs disappear around hatching date)	+	+
		2006	WEBL	(NUWO roosting)	+	+	+	+	+
	45	2004	None	(VGSW visits)
		2005	None	(VGSW visits)
		2006	VGSW	+	Nest abandoned during building (human disturbance?)
	47	2004	None	(VGSW visits)
		2005	MOCH	+ (VGSW, WEBL visit. VGSQ starts building before MOCH)	+	+	+	+	+
		2006	VGSW	+	+	Nest abandoned during building (human disturbance?)
Wired, cont'd	48	2004	None
		2005	None
		2006	None
	54	2004	None
		2005	WEBL	+	+	+	+	+	+
		2006	None
	55	2004	None	(VGSW visits)
		2005	VGSW	+ (MOCH visits)	+	+	+	+	+
		2006	WEBL	+ (MOCH visits)	+ (MOCH building stops; WEBL building starts)	+	+	+	+

Notes:

(1) “+” symbol indicates image data exist for indicated stage of breeding cycle with data losses or other issues indicated. Gray shading indicates data not available for that period at all, or are notably incomplete.

(2) For boxes where no nesting was initiated, roosting and/or pre-building observations may exist, but were not precisely confirmed or denied due to present time constraints on viewing all images.

(3) Image frequency: For wired cams: every 30 minutes in 2003 breeding season; every 15 minutes 2004-2006 breeding season, decreasing to every 30 minutes in off-season. For wireless cams in 2006, every 15 minutes.

(4) Resolution: For wired cams was lower in 2003 relative to later years.

In 2005, thirteen nestboxes contained wired cameras. Of these 13, five nestboxes had birds breeding in them for the entire nesting cycle and the associated image data were collected (Boxes 8, 21, 31, 47, 55). One nestbox contained a pair of birds that initiated breeding but failed at one of the nesting stages (Boxes 22). Lastly, a set of five boxes were not occupied at all (Boxes 3, 14, 27, 45, 48), although some of these boxes were visited by various species.

In 2006, the same thirteen nestboxes contained wired cameras as in 2005. Of these 13, five nestboxes had birds breeding in them for the entire nesting cycle and the associated image data were collected (Boxes 3, 8, 21, 31, 55); three of these were the same boxes occupied for the entire nesting cycle in 2005. Thus, there was not an overall increase from 2005 in the number of boxes with wired cameras that obtained image data for the entire nesting cycle. In 2006, five nestboxes contained pairs of birds that initiated breeding but failed at one of the nesting stages (Boxes 14, 22, 27, 45, 47) compared to just one in 2005. Several of these failures may have been associated with human disturbance at the box. Lastly, three boxes were not occupied at all (Boxes 11, 48, 54).

Images recorded during the "roosting/pre-building" stage of the breeding cycle indicate that for several boxes, multiple species enter a box (e.g., Box 27 in 2006). In other cases, a species visits but the box is never ultimately used for breeding (e.g., Box 45 in 2004 and 2005, Box 47 in 2004). In addition, one species may start building (e.g. VGSW in Box 47 in 2005) but the box is subsequently occupied by a different species (in this case, MOCH) that ultimately completes successful breeding.

System Assessment. While the implementation of a wireless system to convey images from nestbox to microserver location expands the number of nestboxes that can be targeted to locations where a significant source of power does not exist (as is required of the wired cameras), the ability to record images in nestboxes using the Cyclops wireless cameras is nonetheless limited to those areas that are within the range of connectivity. Not only is there a distance limit between nestbox and microserver, but features within the landscape, such as sloping hillsides and vegetation, can obstruct wireless signals even to some nestboxes that are relatively close to the high-gain antennas located at the Stargate microserver locations. The Cyclops units communicate via a "single-hop" system, which may mean that there is no alternative path for a cyclops unit to transmit its image if connectivity is blocked by a topographc obstruction or other factors. As a result, understanding the area where connectivity is most likely to occur and testing this for each box as it is installed is vital to maximizing data collection. Figure xx above shows the anticipated connectivity polygons available in the James Reserve.

Obtaining data sufficient to adequately address biological questions is limited by the number of boxes that have contained wired and/or wireless cameras that also have been occupied by breeding birds within the same year. In any given year, an unpredictable number of nestboxes will be unoccupied by any breeding pair. Of the boxes where nesting is initiated, some proportion of pairs will not complete the nesting cycle. Lastly, technical failures, while expected to be reduced in the future, further reduce sample size. Overall, the total number of boxes with image data for the complete nesting cycle, and/or to answer many types of biological questions covering only a specific stage of the cycle, is relatively low relative to many studies published in scientific journals. As mentioned above, only two of the ten boxes possessing Cyclops cameras in 2006 actually contain images for the complete nesting cycle (from the start of the nesting cycle through to the fledging phase). Overall, increasing the number of boxes with cameras will improve the number of pairs for which breeding behavior can be collected and subsequently increase the opportunity to answer significant biological questions.

The view of the cameras within the nestbox are occasionally obstructed by insects and debris that collect on the plexiglass barrier between the camera and the nest. This can require personnel to visit the box and remove the obstruction.

System Automation. Although Cyclops has a dedicated processor allowing it to perform image processing tasks on the node, most of our efforts so far has been geared towards post-processing of images once aggregated at the back-end server. We feel that this is a necessary step since any automation process should be evaluated with ground truth data, hence the need to visually inspect the images that were processed.

In our initial study we have used images from wired cameras since they offer better resolution and picture quality with respect to exposure. These images have been analyzed to determine two things:

• Presence/absence of the bird inside a nestbox
• State determination of the nextbox activities: occupancy, nest building, egg laying, incubation, hatching, and raising young

In occupancy determination we do not make a distinction between multiple or single birds. In out learning phase these two instances are of the same class. As for the state determination we hope to be able to distinguish between the various phases by looking for a visual cue, which in this case is the eggs. By counting the number of eggs and seeing how it varies over time we would be able to determine in which state the bird is in. So for example, if the egg total remains constant over a period of 2 or more days we can say that bird is in the incubation phase. If the number of eggs decreases dramatically from four to none we can say that the eggs have hatched. By leveraging the deterministic nature of the birds we can reduce the complexity of our vision task to the counting of eggs. These conclusions are being tested with input from project biologists.

Currently we can determine the presence and absence of the bird with at least 90% accuracy. State determination via egg counting has been more difficult since eggs can easily occlude one another or be occluded by other nesting material like twigs or feathers. From our experiments so far, we have found that by leverage results over a sequence of images we can more accurately count the number eggs and hence identify the state of the box.

System Interface. Our 2006 deployment used a database developed at CENS UCLA called SensorBase along with a web-server to allow the browsing of images. Images were regularly uploaded to SensorBase that were then queried by web front-end to display them. The web front-end organized the images in rows, with each row corresponding to a particular Cyclops (nestbox) location. Although we had 3 separate sub-networks, the website displayed images from all the nodes. Each row on the page contained 1-day snapshots for an entire week; this allowed the user to navigate through the images on a weekly basis making for faster browsing. If a particular day of interest was found, the user could then click on the 1-day snapshot to see all the image sequences from that day.

Expanding Study in 2007. In 2007, the focus on the avian nestbox sensor system is being expanded with the involvement of a PhD student in Biology at UC Riverside (Ms. Sharon Coe). Existing biological questions will be expanded to include additional research in the areas of over-winter use of nestboxes, roosting behavior of birds prior to egg laying, incubation behavior as it relates to hatching order and hatching success, brooding behavior, and brood reduction. Specifically, relationships between behavior and environmental variables (especially temperature) will be examined.

For Spring 2007, the wireless Cyclops camera system will be expanded to a total of 25 cameras (one per nestbox) from the 12 installed in Spring 2006 and up to 25 additional nestboxes without cameras or sensors. This increase in Cyclops camera placement in nestboxes will occur by placing them in existing nestboxes that have a history of occupancy by breeding birds but have not previously contained any camera. Furthermore, additional nestboxes are being established at the James Reserve prior to the 2007 breeding season in areas that have been shown to both possess wireless connectivity, and are considered suitable habitat for WEBL and/or VGSW and thus have a reasonable probability of becoming occupied. Nestboxes with a poor history of occupancy will not contain Cyclops cameras in 2007. Furthermore, in Spring 2007 we will have the ability to move any Cyclops from a box that is unoccupied once the breeding season is well underway to an occupied box that lacks a Cyclops. In February 2007, we conducted a broad test of wireless connectivity surrounding the existing Stargate microserver locations to aid in expanding the wireless camera system into additional nestboxes not previously outfitted with Cyclops cameras. This graphic was shown above.

For the wired cameras, we are will increase the frequency that images are collected from the once-every-15 minutes in Spring 2006 to 1 minute or less in Spring 2007. Preliminary tests at a frequency of <1 min. (eg 15 sec) indicate storage and/or ftp handling problems. The resolution currently from the wired cameras is 480 x 700 pixels; there are no plans to alter this resolution for Spring 2007.

For the wireless Cyclops cameras, two different resolutions are being considered: (1) the CIF resolution of 320 x 280 pixels which takes 1 minute to transfer an image from a nestbox to the Stargate microserver; and (2) the same resolution as was used by Cyclops in 2006, which is 128 x 128 pixels. This lower resolution is estimated to be capable of taking as little as ~15 secs to send; however, in 2006, an image of this resolution was sent once every 15 minutes to the Stargate microserver (to match the frequency of the wired cameras). We may program the microservers to request images of low resolution at interval X (higher frequency of images but less information in the image) and request images of higher resolution at interval Y (that could/would contain more visual information).

In Spring 2007, we will utilize the existing three Stargate microserver locations ("Acoustic Tower," "Trailfinder Lodge," and "AMARRS tower"). We expect that there will be 8 nestboxes sending images to a single Stargate so that we will have the capability of a high-resolution (CIF) image every 8 minutes, or a low-resolution image every ~2 minutes.

Herpetological Studies. One of the seven herptofauna pitfall trap arrays installed by the U.S. Geological Service (UCSG) at the James Reserve in 1995 was repaired and modified to support the cyclops/mote units within each of the seven bucket lids. The goal of the prototype design was to minimize interference with animal movement while having a structure that was robust to frequent handling and easily broke down for storage between sampling periods.

Test deployment of pitfall trap array at James Reserve using Cyclops.

The modified bucket lid design included three PVC leg supports that inserted into three 1-inch diameter PVC pipes mounted equidistant on the interior of the bucket. The Cyclops/Mote units were mounted in a plastic clear box and anchored to the lid with screws. Two infrared LED lights were soldered on the Cyclops board in parallel using copper wires and 100 ohm resistors. A 433 Mhz antenna was mounted three feet above the lid using 3/4'' PVC pipe. [JS2] Images were sent over a wireless link to a microserver located approximately 100 meters away in the James Reserve Trailfinder Lodge. Transmission was single-hop. Power was provided to each cyclops mote unit by 2 D cell batteries. A laminated sheet with grid lines was installed in the bucket bottom to provide scale. See the table below for the specifications and make of the materials utilized.

Cross sectional view of modified pitfall trap lid with cyclops/mote unit.

In order to determine the optimal combination of image quality for the purposes of species identification versus transmission/processing speed, several preliminary images were taken at varying resolutions as shown below. A 240 X 240 pixel resolution was ultimately chosen as higher resolution was determined to be more useful to species identification than color. Our current system allows images to be sent once every minute to microserver at this resolution. At this frequency, battery change-outs are anticipate to be required every 2 weeks.

We completed two sampling bouts during August and November 2006. Each lasted four days to mimic standard USGS sampling protocol. Buckets were opened in the morning of Day 1 and closed during the evening of Day 4. The 240 x 240 pixel black and white images were sent every 15 minutes to the stargate microserver located in the James Reserve Trailfinder Lodge and were uploaded into a database by the project engineer for review by the project biologist to determine whether any specimens had been captured, and if they were identifiable from the imagery. The biologist also checked the buckets daily to confirm imagery and process any animals captured.

Time series of images captured with the cyclops unit.

Overall, the majority of cyclops/mote units transmitted viewable images during the trial period. As is typical for a prototype design, several challenges were encountered during the trial run. These and the interim solutions or areas for further exploration in 2007 are summarized in the table below.

Problem Areas	Solutions/ Improvements Areas
Lid supports stick with variable weather causing connections to become loose and the images to become unaligned	Lid is being redesigned to include a pivot point and guide lines.
Occasional freezing up of the motes during transmission	A watchdog module has been implemented on cyclops/mote that regularly checks for "hangups" and resets the units if necessary
Connection between antenna and mote loosened with frequent handling	Problem solving in upcoming year
Reflection from laminated grid paper decreased image quality	Other weather resistant materials will be explored
GUI interface would be improved by showing sequential images from one bucket in a row to facilitate comparison.	GUI interface to be modified to be more user friendly.
Exposure issue: overexposed when sun directly on buckets, underexposed at night	From our experience exposure will always be a problem due to the limitations of current camera technology. What we hope to do is use pre-defined exposure parameters that were obtained during pre-deployment tests.
Image processing	If image processing can be done on the Cyclops this provides the benefits of increased uptime of network as well as increased sampling rate.

Additional sampling periods for the modified pitfall arrays are scheduled for Summer 2007 after the above refinements are made to the prototype design. One pitfall array will be outfitted with cyclop/motes systems and will be compared to the traditional array design for differences in sampling efficacy. We will also analyze the additional information that the cyclops units provide to determine usefulness in identification of species captured, studying animal interactions within the buckets, percent escapees, and minimizing captive mortality.

Accomplishments

Avian Studies

We have successfully implemented the Cyclops wireless camera systems in nestboxes (several of which were occupied in 2006 by breeding birds), have delineated new areas that we can achieve wireless connectivity to nestboxes not previously possessing wireless cameras, and are increasing the number of Cyclops cameras used in nestboxes. We are increasing the resolution of images taken by Cyclops to increase their biological utility, as well as increasing the image frequency. We are incorporating environmental sensors in nestboxes containing wireless cameras so that they collect the same environmental data as the existing wired cameras. We will have 25 nestboxes with wireless cameras and 13 boxes with wired cameras, for a total of 38 boxes with image capability and environmental data. We will also have the ability to transfer the cyclops units and sensors from nestboxes that do not show occupancy to those that do, thus increasing the potential sample size.

As previously mentioned, in 2007 we have expanded the system of wireless Cyclops cameras in nestboxes to 25 in 2007 (from 12 in 2006). This is a critical step to further testing of Cyclops as a useful tool to biological researchers in the real-world field setting. Further, this expansion is important to increasing the sample size of data that allows us to address important biological questions concerning avian reproduction. The addition of environmental sensors to all Cyclops for 2007 (none were present in Cyclops boxes in 2006) allows us to expand our ability to answer questions that relate environmental characteristics to avian breeding behavior. The increase in resolution of the Cyclops camera images to 320 x 280 pixels in 2007 (from 128 x 128 pixels in 2006) is also an important improvement, as low resolution images make interpretation of behavior more difficult. For example, in some of the low resolution images, it is difficult to determine precisely when hatching occurs (eg whether a nestling has hatched or not) or the exact number of nestlings in the nest (even at relatively late stages of the nestling phase).

We have increased the image frequency of wired cameras to one every minute in 2007 (from one every 15 minutes in 2006). The current resolution (480 x 700 pixels) is quite good for many purposes, including identifying the exact number of nestlings in the nest. The increase in image frequency in Cyclops cameras to one every 8 minutes in 2007 (from one every 15 minutes in 2006) is a notable improvement and will increase the usefulness of the data in answering biological questions.

Herpetological Studies

We have designed and tested twice a prototype pitfall array outfitted with a cyclops/mote units. Additional fine tuning and utilization of the design will occur in Summer-Fall 2007.

Future Directions

Further increases in resolution in the Cyclops cameras to match that of the wired cameras (480 x 700 pixels) is highly desirable. The current resolution of the wired cameras is excellent; however, future increases in resolution should be explored to allow for improvements in visibility of nest contents that relate directly to our ability to interpret key events in the breeding cycle of the birds (eg hatching date and time).

Additional tools could be developed in the future, such as motion-detecting sensors to prompt an image to be collected. Motion-detectors could be used to activate the video camera within the box since current sampling protocols (every 8-15 minutes) likely miss birds entering the box and departing soon after. Other methods previously explored that may be re-evaluated include an entry-hole photo-interrupter, thermopile, IR thermometer, motion detector, nest temperature probe, and strain-gage scale.

We are considering expanding our nestbox system to target another abundant breeding species at James Reserve, MOCH. Relatively few of our occupied nestboxes have occupied for an entire breeding cycle by this species. One possible explanation is that this species is a poorer competitor for nestboxes relative to other cavity-nesting species that we commonly do record for an entire breeding cycle (VGSW, WEBL). By reducing the hole size on existing nestboxes (including those that have not been previously occupied by nesting birds), or establishing new boxes with this smaller hole size, it is possible that we can target MOCH as an additional species of study and effectively expand the scope of the research.

Another important factor involved in automating image processing especially on the node is the resolution of the image. Although we would like higher quality images at higher resolutions, current hardware limitations make it impossible or very difficult due to the power constraints of the wireless sensors. If we can map resolution to accuracy for a given algorithm, then we can do a better job of allocating resources (power/memory) on the node. Our two nestbox related algorithms are prime examples of these tradeoffs. Our intuition tells us that presence/absence determination in the nestbox does not require high-resolution images since we are essentially looking for dark "blobs" in the image. However, to count the number of eggs much higher resolution images are needed since the eggs are much smaller objects and similar looking objects could be found in the nestbox. Our current experiments validate these intuitions allowing us to better determine the minimal resolution at which to capture images on the node and hence prolonging the life of the sensors.

Once we feel that such algorithms meet the accuracy requirements of the end-users (scientists/biologists), we can then proceed to port them to Cyclops, which has several benefits: 1) allows for a more scalable system since we are not stressing the bandwidth limitations and 2) longer lifetime since the amount of transmissions are reduced from entire images (thousands of bytes) to meta-data results (a couple of hundred bytes) 3) Increased sampling rate, since the Cyclops does not have to wait until an entire image is sent over the wireless but can capture another image upon completion of the image processing task.

People

Faculty:

• Deborah Estrin (UCLA)
• Michael Hamilton (UCR)
• John Rotenberry (UCR)

Staff:

• Kevin Browne (UC NRS)
• John Hicks (CENS)
• Jamie King (James Reserve)
• Mohammed Rahimi (UCLA)
• Michael Taggart (James Reserve)
• Tom Unwin (James Reserve)

Graduate Students:

• Shaun Ahmadian (UCLA)
• Sean Askay (UCR)
• Sharon Coe (UCR)

External Research Partnerships

Cornell Laboratory of Ornithology (in discussion)