Logical Modeling And Application Of Image Target Detection

Because of the difficulties in detecting target events in a large amount of video data by human,the target detection technology of image objects is widely used in various fields.Image object detection technology is supported to detect target events quickly and correctly in different backgrounds.According to the above requirements,this paper proposes a logical modeling method of target events in images,which is suitable for scenarios that are multi-factor related and can be recognized by human.By simulating the process of human cognition,we use a variety of factors such as color,texture,shape,space-time relationship and edge to make a complete description of the target object based on prior knowledge,that is,the logical model of the target object.The logic model includes a variety of factors and their metrics.We can quickly and correctly detect the target events by computing the factors' metric variables of the input image and then making logical decisions.In this paper,the logical modeling method of target events in images is applied to the detection of finger joint events.According to the characteristics of finger joint events and environmental conditions,the hand was fully described by color and shape,which were distinguished from the background.Then,the logical model of metacarpophalangeal joint was defined by shape of hand and the spatial relationship between fingers and palm.The logical model of proximal interphalangeal joints is defined using factors such as edge and shape.We choose appropriate image processing methods to extract and calculate the model factor indicator variables,and then detect and position the metacarpophalangeal joints and proximal interphalangeal joints by making logical decisions.This paper also applies the logic modeling method of target events in images to video smoke detection in different monitoring scenarios.We construct the logic model o f the video smoke event by factors such as color and the temporal and spatial relations of smoke in continuous frame images.The saturation and color histogram were used to define the color indicator.The temporal and spatial relation factors mainly describe the spatial positional relationship of smoke in adjacent frame images.At last,we quickly and correctly finish the detection of the smoke event by comparing the indicator variables of the factors.Finally,the logical modeling of target events in images is implemented and tested on the image finger joint events and video smoke events.We analyze the experimental results and note the defects and the aspects to improve.