For an improvement of the accuracy of object detection, an image object detection algorithm has thus been proposed based on attention mechanism and multiple feature fusion. On the basis of the Cascade R-CNN model, the algorithm uses RseNet50 as the backbone network, with a simple attention module (SAM) embedded so as to improve the discrimination ability of the network. Secondly, a multi-level feature fusion module (MFFM) is designed by using the deep separable convolution to improve the feature pyramid network (FPN), followed by a fusion of the multi-scale features to enrich the information of feature maps, with the corresponding weights given to the feature maps of different levels to balance the feature information of different scales. Finally, combined with the region proposal network (RPN) structure in the target detection method, the candidate regions of the target can be obtained for classification and regression processing to determine the location and category of the detection target. Experimental results show that compared with Cascade R-CNN target detection algorithm, the detection accuracy of the proposed algorithm has been improved by approximately 2.0%.