Based on the deep foundation pit project of a high-rise building, this paper analyzes the construction technology of vertical excavation cantilever type supporting cast-in-place pile of deep foundation pit in built-up area by combining numerical simulation with field practice, and introduces the construction technology principle and construction flow of the steel composite supporting system inserted at the top of deep foundation pit. The engineering results show that: The construction technology of steel composite support system inserted at the top of deep foundation pit piles successfully solves the difficult problem of safety in the process of foundation pit excavation, and effectively shortens the construction period, and the technology has a good application prospect. The study of construction technology can provide better guidance for similar construction of vertical excavation cantilever supporting cast-in-place pile, and help to promote the progress of foundation pit supporting construction technology

In this paper, we propose a new object detection model, YOLOv7-TinyObject, which is an upgrade of YOLOv7. Two technologies, SE (Squeeze-and-Excitation) and NWD (Normalized Gaussian Wasserstein Distance), are introduced into YOLOv7 to improve the detection ability of tiny objects. Although YOLOv7 has a great improvement over the previous YOLO series, there are still many shortcomings for detecting tiny objects. The detection of tiny objects has always been a difficult and challenging task in object detection tasks, because tiny objects usually have low resolution and fuzziness, and are easily disturbed by noise and occlusion. To address these issues, we propose a novel optimization method that combines the SE attention(SE) with the NWD technique. First, we analyze the limitations of YOLOv7 in tiny object detection. Specifically, YOLOv7 is prone to the problem of positioning error and missing targets in tiny object detection. To overcome these problems, we introduce the SE. The SE improves the representation ability of important features by adaptively learning the relationship between feature channels. By adjusting the weights of feature channels, the SE can enhance the modeling of important features, thereby improving the accuracy of object detection algorithms. Based on this, we further introduce the NWD to enhance the perception ability of the algorithm for tiny targets. The NWD provides richer semantic information by modeling the contextual information around the target. It considers context relations such as domain information, texture features and edge information of the object and is added to the object detection model to enhance the localization and detection accuracy of tiny objects. By increasing the contextual semantic understanding ability of the target, NWD helps to reduce the problem of localization error and missed detection, and improves the performance of tiny target detection. To validate our proposed optimization method, we use a wide range of datasets for experiments and evaluation. Experimental results show that compared with the traditional YOLOv7 model, our proposed improved method achieves significant performance improvement in tiny object detection. By introducing the SE and NWD, our algorithm exhibits higher accuracy and robustness in tiny target scenarios. Our study shows that the SE attention mechanism can improve the representation ability of important features and the NWD can enhance the perception of tiny targets. With these optimization methods, our algorithm achieves higher accuracy and robustness in the task of tiny object detection. This is of great significance for improving the application performance of object detection algorithms in complex scenes. Future research can further optimize and extend these methods to address challenges in more real-world applications.

Flowers were collected from a chosen tree as they dropped without the influence of any human activity. Petals were manually detached using hands while the calyxes were sun dried to 5.5%wb and reserved for the study. The dried samples were pounded and sieved using sieve of 0.25mm aperture to achieve uniform particle size. 100ml of distilled water was used along with 10g of sample to form slurries of average consistency for all viscosity experiments. Potassium carbonate was reduced into powder and sieved. Dosages of potassium carbonate preservative were computed as percentage of Bombax costatum sample (0%, 3%, 6%, 9%) in grams and blended with the sample, which was package into ceramic, metal, plastic and glass containers and hermitically stored at room temperature, for nine months. The viscosity of samples was determined using the Brookfield viscometer. The result showed that viscosity of Bombax costatum at 5.5%wb decreases with increase in potassium carbonate, with 0% sample presenting the best performance. The result also showed that all containers performed effectively. Both metal and glass recorded 92% performance efficiency. Metal container was rated above glass due to its low coefficient of variation, compared to glass. Plastic and ceramic presented 90% and 87% performance efficiencies respectively, with control exhibiting the least performance efficiency of 36%. Room temperature and relative humidity of the stored environment were recorded throughout the storage duration