It depends ....but from my experience, I used V-BOX! Very impressive one but come with costs.

Bluetooth has gained popularity, although from my experience, the percentage matching is much lower than reported in most literature. Apple and Samsungs new phones are scrambling MAC addresses. Although, this is newer technology, it will catch ground especially in the US.

V-Box is both impressive and expensive!

wavetronix radar works great for freeway applications when speed and volume are of interest.

arterial collection methods are hybrid measurements and traffic models such as in adaptive signal systems. Look into various systems and how they are collecting data. Controllers now store and transmit through ethernet or wireless to a TMC and soon to vehicles through I2V communications and new 5.9 GHz 

Point sensors remain the most widely deployed traffic measurement technologies.  These include inductance loops, magnanometers, and cameras.  Expressways (particularly those with ramp meters), actuated traffic signals, and adaptive traffic signals all have one type of point sensor or another.  However, whether all this point sensor data is collected and/or archived is a completely different story.  Other types of technologies used include point-to-point (e.g., Bluetooth and electronic tool tag readers) and trajectory data (e.g., taxi data, commercial vehicle data, and crowd-sourced GPS/smartphone data); the former widely used for travel time measurements, the latter less common in practice (with exception to companies like Google and Yahoo, which presumably use this type of data for their real-time route prescription applications).  Sometimes agencies use more than one technology/data source as well. 

Driver-less vehicles are all the buzz now, though.  These are expected/hoped to be the main source of traffic data in the near future.

Do you know TRID DB? trid.trb.org

Perhaps with this combined terms, you find useful information:

Keywords:"traffic data" AND (collection* or analysis)

Published/started date range:2014-2015

good luck

The leading edge of transportation data has for long been streaming data coming for a variety of sensors (loop detectors, video cameras, weather stations etc). What has changed over the years is the cost of new monitoring systems (more economic ways of producing streaming data, such as the passive data produced by personal GPS), the data granularity (very detailed information collected in real time) and the availability of new sources of unstructured or semi-structured data, such as logs, clickstreams, and social media data (tweets, Facebook posts etc).

Traffic analyses especially those conducted in real time are usually data intensive and, consequently, are directly dependent on the availability of systems and technologies for data collection. Departing for the classical loop detector data collection that is well-documented and researched, there are currently a variety of sources to collect traffic data such as video based technologies. Wireless communication infrastructures and navigation technologies have revolutionized the manner by which we conceive data collection and data coverage. These technologies: (i) collect vehicle positions, (ii) infer relevant information concerning vehicular kinematic characteristics and congestion, and (iii) provide congestion information to drivers. Research on integrating new data collection technologies is still growing and the entire spectrum of new technologies has not yet been evaluated; a good example is mobility pattern information obtained from social media.

Some interesting references can be found in http://www.sciencedirect.com/science/article/pii/S0968090X14000096

mark the new hype: fusion algorithms for big data 

See these websites, maybe they can help you: http://www.jamartech.com/ or http://metrocount.com/

If you need to reconstruct the trajectories of small areas you'll want to use the cameras, but not because NGSIM very expensive, you can use this

https://www.researchgate.net/publication/268217374_Un_algoritmo_per_lacquisizione_delle_traiettorie_veicolari_da_immagini_video

or for individual vehicles, if you need the engine parameters VideoV-BOX or normal smartphone

Article Un algoritmo per l’acquisizione delle traiettorie veicolari ...