The task of object detection poses a significant challenge in the field of computer vision, with vast implications across various domains, including military operations. The ability to accurately detect ground targets, especially in long-range battlefield scenarios, is crucial for successful reconnaissance missions. In this paper, we introduced the process for constructing our homemade dataset, which targeted at army vehicles observed from horizontal view of military equipment, and improved YOLOv8 algorithm to tackle this challenge. Our semi-automatic process of data collection included filtering criteria of similarity and bounding box size to obtain our desired objects from diverse sources. The detection algorithm was enhanced by comprising a large feature map extracted from initial convolutional blocks in the backbone, blending these feature maps with both horizontal and vertical connections, and applying Distance IoU method for post-process instead of traditional IoU metrics. The efficacy of our proposed approach on our dataset was demonstrated by notable improvements of 88.8% for mAP50 and 71.5% for mAP50-95. Specifically, our method yields a 0.6% enhancement in mAP50 and a 0.9% improvement in mAP50-95 compared to the original YOLOv8 model.