To solve the problem of ineffective monitoring indoor air quality in the environment of numerous and uneven distributed polluting gases with a single sensor, and the issue of indoor obstacles affecting the sensor deployment, the improved Northern Goshawk optimization (INGO) algorithm was used to study the deployment of heterogeneous sensor networks. Firstly, the SPM chaotic mapping was used to initialize the population to solve the problems of low diversity, low coverage, and high redundancy in the initialized population of the original Northern Goshawk algorithm. Secondly, the Lévy flight strategy was improved by using non-linear step weights to update the population location. Finally, the problem that the population tends to fall into local optimum at the later stage of the algorithm was solved by fusing Cauchy variation and backward learning. The results show that the proposed optimization algorithm achieves coverage rates of 942% and 930% in barrier-free and obstructed environments, respectively, and the coverage is improved by 08%, 12%, 28%, and 71%, respectively, compared to algorithms proposed by other scholars in barrier-free environments. Therefore, the INGO algorithm can optimally deploy air quality monitoring sensors in indoor obstacle environments, providing a scientific basis for heterogeneous sensor deployment in complex environments such as indoor air quality detection.