1 US10719940B2 - Target Tracking Method and Device Oriented to Airborne-Based mostly Monitoring Scenarios - Google Patents
Lowell Vogel edited this page 2025-09-28 02:16:53 +08:00


Target detecting and monitoring are two of the core duties in the field of visible surveillance. Relu activated totally-connected layers to derive an output of four-dimensional bounding box data by regression, whereby the four-dimensional bounding field knowledge consists of: horizontal coordinates of an higher left corner of the first rectangular bounding box, vertical coordinates of the upper left corner of the first rectangular bounding box, a size of the primary rectangular bounding field, and a width of the primary rectangular bounding box. FIG. 3 is a structural diagram illustrating a goal tracking device oriented to airborne-primarily based monitoring eventualities in response to an exemplary embodiment of the present disclosure. FIG. 4 is a structural diagram illustrating one other goal tracking device oriented to airborne-based mostly monitoring situations in keeping with an exemplary embodiment of the current disclosure. FIG. 1 is a flowchart diagram illustrating a target tracking method oriented to airborne-primarily based monitoring situations in accordance with an exemplary embodiment of the present disclosure. Step one zero one acquiring a video to-be-tracked of the goal object in actual time, and performing frame decoding to the video to-be-tracked to extract a first body and a second frame.


Step 102 trimming and capturing the first body to derive an image for first curiosity area, and iTagPro shop trimming and capturing the second body to derive a picture for target template and an image for second interest area. N times that of a length and width knowledge of the second rectangular bounding box, respectively. N may be 2, that is, the length and width information of the third rectangular bounding box are 2 times that of the size and width data of the primary rectangular bounding box, respectively. 2 instances that of the unique knowledge, acquiring a bounding field with an space four times that of the original data. According to the smoothness assumption of motions, it is believed that the place of the goal object in the first body must be found within the curiosity area that the realm has been expanded. Step 103 inputting the image for target template and the image for first interest area into a preset look tracker network to derive an appearance monitoring position.


Relu, and the variety of channels for iTagPro shop outputting the feature map is 6, 12, 24, 36, 48, and sixty four in sequence. Three for the remainder. To make sure the integrity of the spatial position data in the characteristic map, the convolutional network does not include any down-sampling pooling layer. Feature maps derived from completely different convolutional layers in the parallel two streams of the twin networks are cascaded and integrated utilizing the hierarchical function pyramid of the convolutional neural network whereas the convolution deepens repeatedly, respectively. This kernel is used for performing a cross-correlation calculation for dense sampling with sliding window type on the characteristic map, which is derived by cascading and integrating one stream corresponding to the image for first interest area, and a response map for appearance similarity is also derived. It may be seen that in the looks tracker network, the tracking is in essence about deriving the place the place the target is positioned by a multi-scale dense sliding window search in the curiosity region.


The search is calculated based on the goal look similarity, that's, the appearance similarity between the goal template and the picture of the searched position is calculated at every sliding window position. The position the place the similarity response is massive is extremely probably the position the place the goal is positioned. Step 104 inputting the image for iTagPro shop first curiosity area and the picture for second curiosity area into a preset movement tracker community to derive a motion monitoring position. Spotlight filter frame difference module, a foreground enhancing and background suppressing module in sequence, iTagPro shop wherein each module is constructed based on a convolutional neural network structure. Relu activated convolutional layers. Each of the variety of outputted characteristic maps channel is three, whereby the feature map is the distinction map for the input image derived from the calculations. Spotlight filter body distinction module to obtain a body distinction motion response map corresponding to the interest areas of two frames comprising previous frame and subsequent body.


This multi-scale convolution design which is derived by cascading and secondary integrating three convolutional layers with different kernel sizes, iTagPro geofencing goals to filter the movement noises caused by the lens motions. Step 105 inputting the looks tracking place and the motion monitoring position into a deep integration community to derive an integrated final tracking place. 1 convolution kernel to restore the output channel to a single channel, thereby teachably integrating the monitoring outcomes to derive the final monitoring position response map. Relu activated absolutely-related layers, and a four-dimensional bounding box data is derived by regression for outputting. This embodiment combines two streams tracker networks in parallel within the strategy of tracking the target object, whereby the target object's look and movement information are used to perform the positioning and tracking for the target object, and the final monitoring place is derived by integrating two times positioning data. FIG. 2 is a flowchart diagram illustrating a target tracking technique oriented to airborne-primarily based monitoring situations in accordance to a different exemplary embodiment of the present disclosure.