Process of forecasting trips is one of the most applicable sectors of traortation planning. According to efficient information and accurate analysis, such processes can be used as an efficient attribute during future developments. Data required to forecast trips is stored in a matrix called Origin-Destination matrix. In fact, this matrix consists of the origin-destination data of the traortation network which can be obtained via two direct and indirect methods. Direct methods to construct origin-destination matrix obtain the data directly from origin of the trip, destination of the trip, and during of the trip by asking questions from passengers or observing trips being made. Since construction of the origin-destination matrix using direct methods is so much costly and time consuming, hence implementation of such methods is not reasonable. Indirect methods try to estimate origin-destination matrix for the current year on the basis of existing matrices of some years before using existing traffic volume of some network links and mathematical methods. These methods are mostly in the form of mathematical programming problems. Since 70s up to now, indirect methods have been enormously developed. In this research, estimation methods of Origin-destination matrices were considered by counting traffic of some of the links. Then, one of the most applicable methods called Spiess’s gradient was chosen from the set of methods being considered. Also in this research a model is proposed to estimate demand matrix by counting traffic of some of the links. This model incorporates the idea of partial allocation. Partial allocation method allocates origin-destination matrix to the traortation network while the presented estimation technique of this research operates in a reverse manner. In the presented technique, a matrix is estimated in each stage in order to reproduce the volume of the counted links, gradually. Since number of unknowns (number of elements of the origin-destination matrix) are more than knowns (number of counted links), then one can find lots of matrices capable to reproduce volume of the counted links when being allocated to the network. Also there exists an old matrix for this network and such a matrix includes the total framework of the trips being made between origin and destination. Hence, confinement of the estimated matrix to the old one can be a logical assumption. In the proposed method the estimated matrix can be confined to the old matrix by defining some goals in such a way that a unique and reliable matrix can be achieved. Also this method can utilize additional information such as amount of trips being made and absorbed in each region. .