Aiming at the problem that heterogeneous software modules are difficult to integrate due to high overhead and complexity of traditional middleware, a JSON-based lightweight bridging architecture (LJ-Bridge) is proposed. The key technologies adopted by the architecture, including asynchronous file I/O and predefined JSON data format, are studied, and the realization of loose-coupling communication of heterogeneous modules by combining atomic writing, polling mechanism and timestamp verification is analyzed. The technological innovations are reflected in constructing a file system-level lightweight communication paradigm, designing multiple fault-tolerant mechanisms to ensure data integrity, and adapting to the needs of resource-constrained embedded scenarios. With the UAV obstacle avoidance system on Raspberry Pi 4B platform as a case, experiments on bridging Python 3 vision module and Python 2 control module are carried out. Through simulation tests, compared with TCP/IP communication, the architecture shows advantages in memory efficiency, communication reliability and delay stability. Practical application verification indicates that it meets the heterogeneous integration needs of resource-constrained embedded scenarios with high robustness requirements, and provides a low-overhead solution for the integration of software modules with multiple programming languages and multiple runtime environments.