In the first part of the paper a short review of the computer meteor stream searching techniques is given. Different fractions of the stream component obtained amongst radio, photographic and TV data are partially due to the use of different methods applied in the stream search. A new objective approach is proposed in order to obtain the threshold value D_c_ of the orbital similarity corresponding to the probability of chance occurrence of the stream. The 502 Canadian TV data has been used to test this approach. It appears, that the values of D_c_ given by Southworth & Hawkins (1963, Smith. Cont. Aph. 7, 261) and Lindblad (1971, Smith. Cont. Aph. 12, 14) formulae are too high to warrant sufficient reliability of the identified streams. Indeed using these formulae the result is that the probability to obtain by chance at least one stream of 4-5 members goes from 21% to 68%. The effect of the three different distance functions used to measure the orbital similarity (Southworth & Hawkins (1963, Smith. Cont. Aph. 7, 261), Drummond (1979, 1981Icar...45..545D), and Jopek 1993) has been investigated. For all functions taking the same 95% reliability level, the number of streams detected is considerably less than in the case of the traditional approach (Jopek 1993Icar..106..603J). However, results are different using different distances. Finally, we give the list of the orbits of eight streams identified at the reliability level W_M_=95% using the Jopek distance.