The input and output samples for the network are determined through random function and preprocessed with scaling in interval (0.2, 0.8). The neutral units in the hidden layer is decided by the trialanderror method and an a model of artificial neural network(ANN) with a structure of 7∶8∶1 are established for the assessment of environment quality of urban soils. BPANN model is simulated through programming in Matlab 6.0. It is found that the BPANN model is reliable for assessing accurately and objectively the environment quality of the urban soils. With Changchun as a case study, it is concluded that the environment quality of urban topsoil is generally good and about 81% of the area reached the first level of national environmental quality standard for soils. It should be mentioned that the heavy metals pollution in local area is serious and should not be neglected, and some effective control measures of reducing heavy metal pollutants discharges should be put forward to ensure the sustainable use of urban soil resources.