For energy detection you should know the variance of noise but for matched filter you have to know the CSI of channel or environment. Matched filter is more complex wit respect to the ED.
Spectrum sensing techniques are general processes used to detect some characteristics about the spectrum. Energy detector , matched filter detector, cyclostationary detector and, pilot based coherent detection are algorithms for spectrum sensing.
The energy detector (ED) is the simplest case where, the detector doesn't have any information about the incoming signal, so, its hypotheses are only based on the energy level seen at its interface. So, ED will put a threshold for detecting the presence or absence of others's systems signal and to differentiate it from noise. So, if the signal SNR seen by ED is low, then, it might fail in detecting it.
The matched filter detector is more complex detector , it requires channel side information about the incoming signal and then, once its arrival, it tries to compare to find the existence of the known patter (it is a pilot) and if it exists then, the detector assumes that a signal is there and vice versa. This type has better performance because depending on pattern detection, it can works even at low SNR values where the ED can fails. The drawback in its complexity.
For clarification, I recommend the book: Cognitive Radio, Software Defined Radio, and Adaptive Wireless Systems, Edited by HÜSEYIN ARSLAN, University of South Florida, Tampa, FL, USA, ISBN 978-1-4020-5541-6 (HB), ISBN 978-1-4020-5542-3 (e-book), Published by Springer, www.springer.com. Chapter 9 Spectrum Sensing for Cognitive Radio Applications