The excessive iterative transmission capacity of abnormal information within the IoT video acquisition terminal can lead to resource constraints caused by the overwhelming information parameters, ultimately affecting the performance and accuracy of the detection system. To address this issue, we have designed an IoT video acquisition terminal-based abnormal weak signal detection system leveraging depthwise separable convolution. This design involves establishing a real-time connection between the IoT video acquisition terminal, the intrusion information embedded processor, and the ARM detection unit, thereby completing the hardware design of the abnormal data and weak signal detection system. The depthwise separable convolution model is defined, and based on the layout of the node objects, the deployment of abnormal weak signal detection nodes using depthwise separable convolution is implemented. By acquiring abnormal data packets and standardizing the processing, we refine the specific detection process for abnormal weak signals, thus completing the design of the detection system. Experimental results demonstrate that the proposed method effectively suppresses the excessive iterative transmission of abnormal information within the IoT video acquisition terminal, proving that the designed system exhibits superior performance in detecting abnormal data and weak signals.