The poverty alleviation targets have downscaled to the farming household unit in China. Developing and improving the measurement method for multidimensional poverty at the farming household scale has become the key to accurately allocate the resources and improve the efficiency of poverty alleviation. In this study, a comprehensive framework of ability-assets-environment multidimensional poverty assessment was constructed based on human-environment-activities relationship theories. Taking the Qinling-Daba Mountains area as an example, this study measured the multidimensional poverty of surveyed households. Multidimensional poverty households were identified and compared with the designated poor households. The impact mechanism of multidimensional poverty was also explored. Finally, the multidimensional poverty households were divided into different types according to the deprivation dimensions. The research results show that: 1) There were 245 multidimensional poverty households and 239 non-multidimensional poverty households. With regard to the scores of the capacity dimension, the assets dimension, and the environment dimension, there were great differences between multidimensional and non-multidimensional poor households. 2) The majority (84.08%) of the multidimensional poverty households overlapped with the designated poor households. Furthermore, the degree of poverty of multidimensional poor households was deeper in both comprehensive and individual dimensions. 3) The occurrence of multidimensional poverty was mainly influenced by the weakness and deprivation of development ability and production ability of the households, financial and physical assets for activities, geographical locations and conditions of the environment. 4) There were four types of multidimensional poverty households, namely, development deficiency type, compound poverty type, severe living environment type, and comprehensive poverty type.